
ADIPOSITY AND CVD RISK FACTORS: 
A COMPARISON BETWEEN ETHNICITIES


A thesis submitted for the degree of Doctor of Philosophy


By

 
NONSIKELELO MATHE


Research Centre for Society and Health
Buckinghamshire New University 
Brunel University


October 2010

ABSTRACT
Background: The prevalence of overweight, obesity and obesity-related disease, mainly cardiovascular disease (CVD), is increasing in both developed and developing countries. Ethnic differences have been reported in the prevalence of overweight, obesity and CVD. However, measures used to define overweight and obesity, and identify increased risk of CVD were developed and validated in predominately Caucasian populations in developed countries. Consequently, these measures may not accurately define disease risk in all population groups. Therefore the specific aims of this programme of study were:
To establish the relationship between adiposity and cardiovascular risk factors in different ethnic groups.
To identify field measures of adiposity, relating to cardiovascular risk in different ethnic groups. 
To compare the relationship of adiposity and cardiovascular risk factors in a single ethnic group, that of a rural and an urban population in Zimbabwe.
To identify risk factors for CVD related to adiposity in a population of African origin.
Study design:
Three empirical studies were undertaken. In study one, 312 adult subjects from three ethnic groups (Afro-Caribbean (n=106), Caucasian (n=165) and South Asian (n=41)) were recruited from a University. Twenty-six (26) of each group were individually matched for age (?3 years) gender and BMI (?2 kg/m2) to allow for comparability. Measures of body composition included height, weight, waist and hip circumferences, skinfold thickness measures, body density and percentagebody fat. 
In study two, 81 subjects from two ethnic groups (Afro-Caribbean (n=39) and Caucasian (n=42)) were recruited and tested. They were matched for age, gender and BMI using the same criteria as study one. In addition to the body composition measures taken in study one, random non-fasting blood glucose, total cholesterol, triglycerides and blood pressure were taken.
In study three, 55 men and 108 women from rural Zimbabwe, 8 men and 17 women from an urban low-density suburb in Harare Zimbabwe, and 28 male and 16 female students from the University of Zimbabwe were recruited and tested. In addition to all measures of body composition in studies one and metabolic analysis in study two, participants? dietary intake was assessed by food frequency questionnaire and 24hour recall and physical activity was assessed by a physical activity questionnaire.
Main findings:
The relationship between BMI and %BF was not the same in all ethnic groups. (aim 1)
There were ethnic differences in the cardiovascular risk predictors between Afro-Caribbean and Caucasian men and women. (aim 1)
It is not recommended that BIA is used as a substitute for TBW estimation in multi-compartment models. (aim 2)
In three groups of Zimbabweans from urban, rural and university locations, a pattern emerged. Amongst women, urban women were at greatest risk, reporting highest values for all variables, followed by rural then university women. Amongst men, urban men were at highest risk, however there were few differences between rural and university men. (aim 3).
Finally, increased WC and dyslipidemia are associated with increasing BMI in populations of African origin. (aim 4)
Conclusions:
The relationships between overweight, obesity and risk of obesity-related disease differ between different ethnic groups. Moreover, in the groups from Zimbabwe, differences in obesity-related risk were associated with being female and living in urban areas. Therefore, application of universal measures for defining obesity and related diseases may not be applicable to all ethnic groups.


TABLE OF CONTENTS

ABSTRACT	ii
TABLE OF CONTENTS	v
LIST OF FIGURES	xv
LIST OF TABLES	xvi
ABBREVIATIONS	xix
ACKNOWLEDGEMENTS	xxi
AUTHORS DECLARATION	xxii
chapter One:	INTRODUCTION	1
1.1	Context	1
1.2	Aims	3
1.2.1.	Study objectives:	4
1.2.1.1.	Study one: Measurement of adiposity in three ethnic groups	4
1.2.1.2.	Study two: Adiposity and cardiovascular risk factors in Afro-Caribbean and Caucasian men and women	4
1.2.1.3.	Study three: Adiposity and cardiovascular risk factors in rural and urban Zimbabweans	4
1.3	Thesis narrative	5
1.4	Thesis outline	7
1.5	References	8
chapter Two:	ETHNIC DIFFERENCES IN OVERWEIGHT AND OBESITY	11
2.1 Introduction	11
2.2 Classification of obesity	12
2.3 Global prevalence of overweight and obesity	13
2.3.1 Ethnic differences in overweight and obesity prevalence	15
2.4 Overweight and obesity-related disease	18
2.4.1 Global prevalence of obesity-related disease	21
2.4.1.1 Cardiovascular Disease	21
2.4.1.2 Type II diabetes mellitus	22
2.4.1.3 Hypertension	22
2.4.2 Ethnic differences in obesity-related disease	23
2.5 Central obesity	27
2.5.1 Classification of central obesity	27
2.5.2 Ethnic differences in central obesity	29
2.5.3.1 Ethnic differences in visceral adiposity	30
2.6 Ethnic differences in obesity measurement	33
2.6.1 Body mass index	33
2.6.2 Waist Circumference	36
2.7 Conclusion	38
2.8 References	39
chapter Three:	AN HISTORICAL REVIEW OF BODY COMPOSITION ASSUMPTIONS AND METHODS	44
3.1 Introduction	44
3.2 Cadaver studies	45
3.3 Characteristics of body components	47
3.3.1 Fat mass	47
3.3.2 Fat free mass	48
3.3.2.1 Hydration of fat free mass	49
3.3.2.2 Protein content of fat free mass	50
3.3.2.3 Mineralisation of fat free mass	50
3.4 Body composition models	51
3.4.1 The two-compartment model of body composition	51
3.4.1.1 Assumptions	51
3.4.1.2 Limitations and sources of error in the two-compartment model	52
3.4.1.3 Ethnic considerations	53
3.4.2 Three-compartment models	53
3.4.3 The four-compartment model	54
3.4.4 Multi- compartmental models	55
3.6 In vivo body composition methods	57
3.6.1 Air displacement plethysmography	59
3.6.1.1 Validity of air displacement plethysmography for estimating percentagebody fat	61
3.6.1.2 Air displacement plethysmography in different ethnic groups	66
3.6.2 Bioelectric Impedance Analysis	69
3.6.2.1 Validity of BIA of estimation of body fat in different ethnic groups	70
3.6.3 Anthropometry	71
3.6.4.1 Skinfold thickness measurement	72
3.7 Conclusion	75
3.8 References	76
chapter Four:	BODY MASS INDEX AND ADIPOSITY IN THREE ETHNIC GROUPS	88
4.1 Introduction	88
4.2 Methods	90
4.2.1 Subjects	90
4.2.1.1 Ethical Approval	91
4.2.1.2 Selection Criteria	91
4.2.1.3 Ethnic Origin	91
4.2.2 Outcome Measures	92
4.2.2.1 Anthropometry	92
4.2.2.3 Air displacement plethysmography	93
4.2.3 Statistical Analysis	93
4.3 Results	95
4.3.1 Unmatched group comparisons	95
4.3.2 Matched group comparisons	96
4.3.2.1 Relationship between body mass index and % body fat	97
4.3.2.2 Age, gender, body mass index and ethnicity as predictors of percentagebody fat	97
4.4 Discussion	99
4.4.1 Effect of individual matching for age, gender and body mass index	100
4.4.2 Use of different methods for estimating % body fat in different ethnic groups	101
4.4.3 Implications for the use of BMI as a proxy for adiposity	102
4.4.4 Study limitations	103
4.5 Conclusion	103
4.6 References	104
chapter Five:	A COMPARISON OF TWO COMPARTMENT DENSITOMETRY EQUATIONS FOR THE ESTIMATION OF percentageBODY FAT IN AFRO-CARIBBEANS	108
5.1 Introduction	108
5.2 Methods	113
5.2.1 Subjects	113
5.2.2 Outcome Measures	113
5.2.2.1 Anthropometry	114
5.2.2.2 Body density	114
5.2.2.3 Bioelectric Impedance Analysis	115
5.2.2.4 Two?compartmental densitometry equations	115
5.2.2.5 The reference model	116
5.2.3 Statistical analysis	116
5.3 Results	117
5.3.1 Comparison of two-compartment equations with reference in Afro-Caribbean men and women	118
5.3.2 Comparison of density of FFM in Afro-Caribbean and Caucasian group	121
5.4 Discussion	122
5.4.1 Ethnic specific versus generalised equations in Afro-Caribbean men and women	122
5.4.2 Modified three compartment reference model	124
5.4.5 Study limitations	125
5.5 Conclusion	125
5.6 References	126
chapter Six:	A STUDY OF CVD RISK FACTORS IN AFRO-CARIBBEANS AND CAUCASIANS	128
6.1 Introduction	128
6.2 Methods	130
6.2.1 Subjects	130
6.2.1.1 Ethical approval and selection criteria	131
6.2.1.2 Ethnicity	131
6.2.2 Outcome measures	131
6.2.2.1. The Framingham risk score	131
6.2.3 Statistical analysis	132
6.3 Results	133
6.3.1 Ten-year cardiovascular risk	134
6.3.2 Predictors of CVD risk in Afro-Caribbeans compared with Caucasians	135
6.4 Discussion	136
6.4.1 Study limitations	137
6.5 Conclusion	138
6.6 References	139
chapter Seven:	THE NUTRITION TRANSITION: A CASE STUDY OF ZIMBABWE	142
7.1 Introduction	142
7.2 Description of the nutrition transition	144
7.3 Drivers of the nutrition transition in developing countries	146
7.3.1 Urbanisation and diet	147
7.3.2 Urbanisation and physical activity	148
7.4 Consequences of nutrition transition	149
7.4.1 Double burden of malnutrition and disease	150
7.5 Zimbabwe	152
7.5.1 The population	153
7.5.2 The economy	155
7.5.3 Urbanisation	157
7.5.4 Evidence of dietary transition in Zimbabwe	157
7.5.5 Evidence of changes in physical activity	161
7.5.6 Double burden of disease in Zimbabwe	164
7.5.6.1 Selected risk factors for CVD in Zimbabwe	166
7.6 Conclusion	170
7.7 References	171
chapter Eight:	COMPARISON OF ANTHROPOMETRIC AND DEMOGRAPHIC PROFILES OF A GROUP OF URBAN AND RURAL ZIMBABWEANS	176
8.1 Introduction	176
8.2 Methods	178
8.2.1 Subjects	178
8.2.2 Ethical approval	178
8.2.3 Study sites	178
8.2.4 Recruitment strategy	179
8.2.5 Outcome measures	179
8.2.5.1 Demographics and socio-economic variables	179
8.2.5.2 Anthropometry	180
8.2.6 Statistical analysis	181
8.3 Results	182
8.3.1 Demographics and socio-economic variables	182
8.3.1.1 Mean age	183
8.3.1.2 Marital status	184
8.3.1.3 Education	184
8.3.1.4 Employment status	184
8.3.1.5 Number of dependents	185
8.3.2 Anthropometric measures of adiposity	185
8.4 Discussion	187
8.4.1 Demographic and socio-economic differences	187
8.4.2 Anthropometric differences	188
8.4.3 Study limitations	189
8.5 Conclusion	190
8.6 References	191
chapter Nine:	ENERGY BALANCE IN A SAMPLE OF RURAL AND URBAN ZIMBABWEANS	193
9.1 Introduction	193
9.2 Methods	196
9.2.1 Outcome measures	196
9.2.1.1 Energy intake	196
9.2.1.2 Total energy expenditure (TEE)	197
9.2.2 Statistical analysis	197
9.3 Results	198
9.3.1 Comparison of daily energy intake and dietary patterns in urban and rural Zimbabweans.	198
9.3.2 Comparison of energy expenditure and physical activity patterns in rural and urban Zimbabweans	200
9.3.3 Comparison of energy balance in rural and urban Zimbabweans	201
9.4 Discussion	202
9.5 Conclusion	204
9.6 References	205
chapter Ten:	BLOOD PRESSURE, BLOOD GLUCOSE AND LIPID PROFILES IN RURAL AND URBAN ZIMBABWEANS	208
10.1 Introduction	208
10.2 Methods	210
10.2.1 Outcome measures	211
10.2.1.1 Health history and Smoking status	211
10.2.1.2 Blood pressure	211
10.2.1.4 Blood lipids and glucose	212
10.2.2 Statistical analysis	212
10.3 Results	213
10.3.1 Previous health history and history of health checks	214
10.3.1.1 Last blood pressure check	215
10.3.1.2 Last blood cholesterol check	215
10.3.1.3 Last blood glucose check	215
10.3.1.4 Smoking status	216
10.3.1.5 Previous diagnosis of disease	216
10.3.2 Blood pressure, lipids and glucose	217
10.3.2.1 Blood pressure	218
10.3.2.2 Non-fasting blood glucose	218
10.3.2.3 Non-fasting total cholesterol	218
10.4 Discussion	219
10.4.1 Risk factor screening	220
10.4.2 Differences in blood pressure, lipids and glucose	220
10.4.3 Study limitations	223
10.5 Conclusion	223
10.6 References	224
chapter Eleven:	ADIPOSITY AND CARDIOVASCULAR RISK FACTORS IN ADULTS OF AFRICAN ORIGIN	228
11.1 Introduction	228
11.2 Methods	230
11.2.1 Subjects	230
11.2.2 Outcome measures	230
11.2.1.1 Anthropometric measures	230
11.2.1.2. CVD risk factors	230
11.2.2 Statistical analysis	231
11.3 Results	232
11.4 Discussion	234
11.4.1 Study limitations	236
11.5 Conclusion	237
11.6 References	237
chapter Twelve:	THESIS SUMMARY, LIMITATIONS AND RECOMMENDATIONS	239
12.1 Summary	239
12.1.1 Summary of chapter two	240
12.1.2 Summary of chapter three	240
12.1.3 Summary of chapter four	240
12.1.4 Summary of chapter five	241
12.1.5 Summary of chapter six	242
12.1.6 Summary of chapter seven	242
12.1.7 Summary of chapter eight	243
12.1.8 Summary of chapter nine	244
12.1.9 Summary of chapter ten	245
12.1.10 Summary of chapter eleven	245
12.2 Limitations	246
12.3 Recommendations for future work	247
12.4 Conclusion	249
12.4.1 Public health implications	250
APPENDICES	252
APPENDIX 2.1 :Ethnicity	252
APPENDIX 3.1: THE BRUSSELS CADAVER ANALYSIS SERIES	256
APPENDIX 4.1: UK ETHNIC CATEGORIES, CENSUS (2001)	259
APPENDIX 8.1 FIELDWORK BOOK	260


LIST OF FIGURES
Figure 1. 1 Graphic representation of thesis	7

Figure 2. 1 Distribution of deaths, globally by leading cause Source: WHO (2004)	21
Figure 2. 2 Prevalence of CVD by ethnic group and age in men and women	24
Figure 2. 3 Mortality risk versus BMI in US men and women	33
Figure 2. 4 WHO obesity action points for determining public health and clinical action points in Asian populations.	35

Figure 3. 1 The five level body composition model developed by Wang et al., (1992)	55
Figure 3. 2 Cross sectional representation of the BOD POD	59
Figure 3. 3 Skinfold thinkness analysis	73

Figure 5. 1 A-E  Bland-Altman analysis of the individual residual %BF scores for 2-C equations	120

Figure 7. 1 Map of Zimbabwe	152
Figure 7. 2 Zimbabwean population pyramid (2005-2006)	153
Figure 7. 3 Food groups plotted against year (Source: FAO 2010)	159
Figure 7. 4 Total energy availability (kcal/capita/day) from 1961-2007 in Zimbabwe	160
Figure 7. 5 Causes of death in Zimbabwean males (2004)	164
Figure 7. 6 Causes of death in Zimbabwean females (2004),	165


LIST OF TABLES
Table 2. 1 World Health Organization body mass index classifications for adults	12
Table 2. 2 Prevalence of overweight and obesity among people aged 20-60 years in the USA by ethnic group (%)	15
Table 2. 3 Prevalence of overweight and obesity among people aged 16 and over in England by ethnic group (%)	17
Table 2. 4 Waist circumference and waist-to-hip ratio cut-off points for central obesity, reflecting increased risk of CVD	27
Table 2. 5 Prevalence of central obesity in minority ethnic groups in the UK	29
Table 2. 6 Ethnic specific cut-off points for WC from International Diabetes Federation	37

Table 3. 1 Classic cadaver studies between 1945 and 1968	46
Table 3. 2 Composition of fat mass	47
Table 3. 3 Properties of fat free mass	48
Table 3. 4 Multi-compartment models of body composition	56
Table 3. 5 In vivo body composition methods	58
Table 3. 6 Formulae used to convert body density to percentagebody fat in the BOD POD system	61
Table 3. 7 Validity of air displacement plethysmography against hydrodensitometry in adults	63
Table 3. 8 Validity of air displacement plethysmography against dual energy x-ray absorptiometry in adults	65
Table 3. 9 Advantages and disadvantages of bioelectric impedance analysis	70
Table 3. 10 Advantages and disadvantages of anthropometric measures	72

Table 4. 1 Unmatched subject characteristics (mean (SD))	95
Table 4. 2 Classification of weight using World Health Organization BMI categories (N, % of group in this category)	96
Table 4. 3 Matched subject characteristics (mean (SD))	96
Table 4. 4 Summary of model to predict percentagebody fat	97
Table 4. 5 Evaluation of the contribution of independent variables to the variance and prediction of percentagebody fat	98
Table 4. 6 Summary of prediction of model of ethnicity variables	99

Table 5. 1 Body composition prediction equations	109
Table 5. 2 Subject characteristics (mean (SD))	117
Table 5. 3 percentagebody fat (mean (SD))	118
Table 5. 4 Agreement between 3-C reference and 2-C equations	119
Table 5. 5 Comparison of body density and density of FFM, and between Afro-Caribbeans and Caucasians (mean (SD))	121

Table 6. 1 Established risk factors for CVD	129
Table 6. 2 Subject characteristics (mean (SD))	133
Table 6. 3 Fat distribution measured by WC and waist-to-hip ratio	133
Table 6. 4 Blood glucose, lipid profile and blood pressure	134
Table 6. 5 Predictors of Framingham risk score in Afro-Caribbeans compared with Caucasians	135

Table 7. 1 Stages of the nutrition transition	144
Table 7. 2 Annual population growth rates between 1901-2002	154
Table 7. 3 Birth, death rates and life expectancy in Zimbabwe	155
Table 7. 4 Food supply in Zimbabwe between 1980 and 2007 (kcal/capita/day)	158
Table 7. 5 Levels of physical inactivity in leisure, transport and work in Zimbabwean adults, in rural and urban areas	161
Table 7. 6 Difference in occupation between rural and urban men and women (%)	162
Table 7. 7 Exposure to mass media in rural and urban Zimbabwe	163
Table 7. 8 Overweight and obesity prevalence in Zimbabwean adults (%)	166
Table 7. 9 Blood pressure (BP) and prevalence of hypertension in adults aged 25-100years, combined rural and urban data	167
Table 7. 10 Diabetes prevalence in Zimbabwean men and women (2005) (%)	168
Table 7. 11 Cholesterol level and prevalence of raised cholesterol, 25-100years	169

Table 8. 1 Demographic and socioeconomic characteristics	183
Table 8. 2 Comparison of body weight, height, BMI and body fat in rural and urban residents	185
Table 8. 3 Anthropometric measures of fat distribution	186

Table 9. 1 Commonly consumed foods in urban and rural Zimbabweans	198
Table 9. 2 Comparison of daily energy intake and sources of energy in rural and urban Zimbabweans (mean (SD))	199
Table 9. 3 Comparison of common physical activities in rural and urban Zimbabweans	200
Table 9. 4 Physical activity levels in rural and urban Zimbabweans	201
Table 9. 5 Energy balance in rural and Urban men and women	201

Table 10. 1 Ranges for blood lipids in Zimbabwean men and women (mmol/l)	209
Table 10. 2 Previous blood pressure, lipid and glucose checks	214
Table 10. 3 Blood lipids, pressure and glucose in rural and urban men and women	217
Table 10. 4 Comparison of total cholesterol, high density lipoprotein and triglycerides means with ranges suggest by Gomo (1985)	222

Table 11. 1 Subject characteristics of African men and women	232
Table 11. 2 Odds ratios for cardiovascular risk factors by body mass index category	233
Table 11. 3 Comparison of odds ratios for CVD risk factors in different ethnic groups	236

ABBREVIATIONS
AA- African American
ADP- Air Displacement Plethysmography
ATP III- Adult Treatment Panel Three
BIA- Bioelectric Impedence Analysis
BMI- Body Mass Index
CA- Caucasian American
CHD- Coronary Heart Disease
CVD- Cardiovascular Disease
CT- Computed Tomography
DFFM ? Density of Fat Free Mass
DFM ?Density of Fat Mass
ESRF- End-Stage Renal Failure
FAO- Food and Agriculture Organization
FFA- Free Fatty Acids
FFM- Fat Free Mass
FM- Fat Mass
HD-Hydrodensitometry
HDL-C-High Density Lipoprotein Cholesterol
HFFM- Hydration of Fat Free Mass
HSE- Health Survey for England
IDF- International Diabetes Federation
IHD- Ischemic Heart Disease
LDL-C ? Low density Lipoprotein Cholesterol
LPL- Lipoprotein Lipase
METS- Metabolic Equivalents
MetS- Metabolic Syndrome
MFFM ?Mineralisation of Fat Free Mass
MI- Myocardial Infarction
NCEP- National Cholesterol Education Programme
NHLBI- National Heart Lung and Blood Institute
NIH- National Institute of Health
PALS-Physical Activity Levels
SAA- Surface Area Artifact
SAT- Subcutaneous Adipose Tissue
SD- Sagittal Diameter
TBW- Total Body Water
TEE- Total Energy Expenditure
T2DM- Type 2 Diabetes Mellitus
TGV- Thoracic Gas Volume
VAT- Visceral Adipose Tissue
WHO- World Health Organization
WC- Waist Circumference
WHR- Waist-to-Hip Ratio
%BF- Percentage body fat


ACKNOWLEDGEMENTS

I am eternally indebted to my supervisors Professor David Brodie and Dr Joan Gandy for their tireless efforts, patience, knowledge and guidance throughout these years.
I am grateful for the advice and guidance I received from the late Professor Lucie Malaba and colleagues at the University of Zimbabwe. 
To Dr Paul Amuna and Dr Francis Zotor at the University of Greenwich who assisted with software for dietary analysis.
Many thanks to Fortune Maduma, Lydia Muzangwa, Mthulisi Dube, and Cynthia Matare who assisted me with data collection in Zimbabwe.
I am grateful to Paul Sharp from Point of Care Services for his ingenious idea for a battery charger which was invaluable in my fieldwork in rural Zimbabwe.
To Dumisani Nyoni, Mrs Sithembiso Nyoni and staff members of ORAP Zimbabwe who assisted with accommodation and transportation during the fieldwork in rural Nkayi, Zimbabwe.
To all those who participated in the studies in the UK and Zimbabwe.
To my friends who?ve provided useful diversion at times of high stress
To Dr David Shaw for his moral support throughout the years
To my family, especially my uncle Kidwell and Aunt Ann and my cousins who have been very supportive.
To my sister, Sibonile for her support
Finally and most importantly, I am especially grateful to my mother, Reginah Mathe, for her unconditional support, encouragement and love, without which none of this would have been possible.

I dedicate this thesis to my father Fortune Mathe, who is missed dearly.
AUTHORS DECLARATION
I take responsibility for all the material contained within this thesis and confirm this is my own work.

Nonsikelelo Mathe


This thesis has been supplied on condition that anyone who consults it is understood to recognise that its copyright rests with the author under terms of the United Kingdom Copyright Acts. No quotation from this thesis and no information from it may be published without proper acknowledgement.
 INTRODUCTION

Context

In 1997, the World Health Organization (WHO) declared a global obesity epidemic (WHO 1998). This was in response to the increasing prevalence of overweight and obesity reported in all regions of the world. In 2005, 1.6 billion people were overweight with at least 300 million of them of obese. It was estimated that by 2015, 2.6 billion people would be overweight with 700million obese (WHO 2006). Obesity is associated with a large number of medical problems, including sleep apnoea, osteoarthritis, psychological problems, diabetes, cardiovascular diseases (CVD) (such as coronary heart disease, hypertension) and certain types of cancer (Bray 2004).
Some ethnic groups appear to be disproportionately affected by obesity and its comorbidities, therefore this thesis focuses on the ethnic differences in obesity and its comorbities. In the USA, the highest rates for overweight and obesity were reported in non-Hispanic black women; more than half of these women aged 40 years or older were obese and more than 80% were overweight in 1994-2006 compared with Caucasian groups (Wang and Beydoun 2007). Globally, the Pacific Islands have the highest concentration of overweight and obese people compared with the rest of the world. Rates as high as 75% for overweight and obesity combined have been reported in Nauru, Samoa, American Samoa, Cook Islands, Tonga and French Polynesia (WHO, West Pacific Region 2003). Ethnic differences have been reported in the prevalence of obesity-related disease. In the UK, compared with the general population, the prevalence of type II diabetes mellitus was highest in people of South Asian origin (from the Indian sub-continent), while people of African origin (Black African and Caribbean) had the highest prevalence of hypertension and its sequeale (McKeigne et al., 1991, Primatesta et al.,2004, Scarborough et al., 2010).

Differences in the prevalence of obesity-related diseases among ethnic groups have been attributed to differences in body composition and particularly, differences in the distribution of body fat (Rush et al., 2009). It is well established that the accumulation of fat in the abdominal area increases the risk of CVDs more than its accumulation in other regions of the body (Donahue et al., 1987, Rexrode et al., 1998). However, the majority of body composition measures were developed in predominately Caucasian populations and as such may not be appropriate for identifying increased body fat, differences in distribution of body fat and increased risk for obesity-related disease in non-Caucasian ethnic groups (Deurenberg and Duerenberg-Yap 2003). 
Body composition assumptions, that govern the development of body composition methods, were derived from the analysis of a limited number of cadavers. Between 1945 and 1968 the classic cadaver studies by Mitchell et al., (1945), Widdowson et al., (1951), Forbes et al., (1956) and Moore et al., (1968), reported the chemical analysis of eight cadavers. The cadavers were predominately male and of Caucasian origin, therefore it has been questioned whether the assumptions derived from these analysis can be generalised to other population groups (Deurenberg and Duerenberg-Yap 2003). Thus a specific focus of this thesis is the differences in body composition methods and the implications of the assumptions in three ethnic groups in the UK, South East Asians, Afro-Caribbeans and Caucasians.
Although previously thought to affect mostly populations in high-income countries, overweight and obesity prevalence has increased in middle- and low-income countries (Prentice 2006). It co-exists alongside under-nutrition in what has been called the double burden of malnutrition (Food and Agriculture Organization (FAO) /WHO 2002). The increase in the prevalence of overweight and obesity is a result of changes in diet and physical activity patterns occurring in developing regions that have been described as the nutrition transition (Popkin 2002). The nutrition transition describes the shift from traditional diets to those high in fats, salt, sugars and low in fibre, and increasing physical inactivity as a result of mechanisation of work and leisure activity, and the increased use of motorised transport (Popkin 2009). Thus the prevalence of obesity-related diseases is increasing in developing countries (Amuna and Zotor 2008).


Aims 

The aims of this programme of study were.

To establish the relationship between adiposity, (as measured by air-displacement plethysmography (ADP), bioelectrical impedence analysis (BIA) and anthropometry) and CVD risk factors in different ethnic groups.
To identify field measures of adiposity, relating to CVD risk in different ethnic groups. 
To compare the relationship of adiposity and CVD risk factors in a single ethnic group, that of a rural and an urban population in Zimbabwe.
To identify risk factors for cardiovascular disease related to adiposity in a population of African origin.


Study objectives:

Study one: Measurement of adiposity in three ethnic groups

Measurement of body composition using ADP, BIA, and anthropometry in Caucasians, South Asians and Afro-Caribbean subjects.
Comparison of the relationship between BMI and %BF in three ethnic groups

Study two: Adiposity and cardiovascular risk factors in Afro-Caribbean and Caucasian men and women

Exploration of the relationship between measured adiposity and cardiovascular risk factors in two ethnic groups (Afro-Caribbeans and Caucasians) and identification of the strongest predictors of CVD risk which relate to adiposity in each ethnic group.

Study three: Adiposity and cardiovascular risk factors in rural and urban Zimbabweans
Comparison of measures of adiposity and cardiovascular risk in a rural and urban population in Zimbabwe.
Identification of predictors of CVD risk in Africans, which relate to adiposity.


Thesis narrative

Following the introduction (chapter one), chapter two discussed the global prevalence of overweight and obesity with a particular focus on ethnic differences in prevalence. It can be seen from this chapter that there are ethnic differences in the prevalence of obesity and obesity-related diseases. However, there appears to be a disconnect between measures of obesity such as BMI and WC in identifying increased risk of obesity and obesity-related disease in non-Caucasian ethnic groups. Thus, body composition measurement is a key theme in this thesis. Chapter three is an historic review of body composition assumptions and methods. This chapter is a critique of the application of body composition assumptions in different ethnic groups. It is clear from this chapter that the assumptions developed from the classic cadaver studies may not be applicable to all population groups and in particular non-Caucasian groups. This has important implications for the use of proxy measures of adiposity, such as BMI for the identification of obesity and obesity-related disease in all population groups.  The details of the first study are reported in chapters four and five. In chapter four the relationship between BMI and adiposity estimated using ADP and the generalised two-compartment densitometry equation of Siri (1961) was compared. This comparison was carried out in a group of Afro-Caribbean, Asian and Caucasian men and women, matched for age, gender and BMI. 
The equation of Siri (1961) is a generalised equation, which assumes that the density and composition of fat free mass is constant in all populations based on the classic cadaver studies.  However, this assumption has been challenged and alternative equations have been developed for use in non-Caucasian groups. In chapter five, three two-compartment densitometry equations for the estimation of percentage body fat (%BF) are compared with a three-compartment densitometry model in a group of Afro-Caribbean men and women. Three of the two-compartment equations were developed for use in African Americans and have been integrated into the BOD PODs? software. Therefore, this study was limited to Afro-Caribbean and Caucasian groups. In addition, body density and density of fat free mass in the Afro-Caribbean group are compared to a Caucasian group. This chapter serves to emphasis the complexities of ethnic differences in body composition and its measurement. 
Study one showed differences in body composition and differences in the measurement of body composition in different ethnic groups. Therefore, in study two the aim was to explore the differences in CVD risk between ethnic groups given the differences in body composition. The details of study two were reported in chapter six, where the relationship between measured adiposity and CVD risk factors in two ethnic groups (Afro-Caribbeans and Caucasians) was compared and the strongest predictors of CVD risk which relate to adiposity in each ethnic group were identified using the Framingham risk score as the dependant variable.
The third study is population study titled adiposity and cardiovascular disease risk factors in rural and urban Zimbabweans. The study is described in three comparative studies. In chapter eight the demographic and anthropometric differences are compared between three groups, rural, urban and university men and women. Chapter nine describes differences energy expenditure, and chapter ten describes differences in blood pressure, lipids, and glucose between three groups of Zimbabweans.  
Chapter eleven aimed to identify predictors of cardiovascular risk in adult Africans associated with an increasing BMI. A diagrammatic representation of the thesis outline is shown in figure 1.1


Thesis outline






		
	







	



	


References
Amuna, P.A., and Zotor, F., (2008). Epidemiological and nutrition transition in developing countries: impact on human health and development. Proceedings of the Nutrition Society, 67, pp.82?90.

Bray, G.A., (2004). Medical consequences of obesity. Journal of Clinical Endocrinology and Metabolism, 89,  pp.2583-2589.
Brozek, J., Grande, F., Anderson, J.T., Keys, A.,(1963). Densitometric analysis of body composition: revision of some quantitative assumptions. Annals of New York Academy of Science, 110, pp.113-140.
Deurenberg, P., and Deurenberg-Yap, M., (2003). Validity of body composition methods across ethnic population groups. Acta Diabetologica, 40, pp.s246-s249.
Donahue R. P.,  Bloom, E., Abbott R. D.,. Reed, D., , Yano, K., (1987). Central obesity and coronary heart disease in men. The Lancet,  329, pp.821? 824.

Forbes, R.M., Cooper, A.R., Mitchell, H.H., (1953). The composition of the adult human body as determined by chemical analysis. Journal of Biological Chemistry, 203, pp.359-366.
Forbes, R.M., Mitchell, H.H., Cooper, A.R., (1956). Further studies on the gross composition of the adult human body. Journal of Biological Chemistry, 223, pp.969-975.
McKeigue, P.M., Shah, B., Marmot, M.G., (1991). Relation of central obesity and insulin resistance with high diabetes prevalence and cardiovascular risk in South Asians. The Lancet, 337, pp.382 ? 386.
Mitchell, H.H., Hamilton, T.S., Steggerda, F.R., Bean, H.W., (1945). The chemical composition of the adult human body and its bearing on the biochemistry of growth. Journal of Biological Chemistry, 158, pp.625-637.
Popkin, B.M., (2006). Global nutrition dynamics: the world is shifting rapidly toward a diet linked with non-communicable disease. American Journal of Nutrition, 84, pp.289-298. 
Popkin, B. M. and Gordon-Larsen P., (2004). The Nutrition Transition: Worldwide Obesity Dynamics and Their Determinants. International Journal of Obesity, 28, pp.S2-S9. 
Popkin, B. M., (2002). An overview on the nutrition transition and its health implications: the Bellagio meeting. Public Health Nutrition, 5, pp.93-103.
Primatesta, P., (2004). Health Survey for England Health of Ethnic Minorities available [online] at www.doh.gov.uk .[Accessed January 2008]
Prentice, A.M., (2006). The emerging epidemic of obesity in developing countries. International Journal of Epidemiology, 35, pp.93-99.

Rexrode, K.M., Carey,V.J., Hennekens, C.H.,. Walters, E.E., Colditz, G.A., Stampfer, M.J., Willett, W.C., Manson, J.E., (1998). Abdominal Adiposity and Coronary Heart Disease in Women. Journal of the American Medical Association. 280, pp.1843-1848.
Scarborough, P., Bhatnagar, P., Kaur, A., Smolina, K., Wickramasinghe, K., Rayner, M., (2010). Ethnic differences in CVD. British Heart Foundation Statistics Database available [online] at: www.heartstats.org [Accessed July 2010]

Siri, W. E., (1961) Body composition from Fluid spaces and density: analysis of methods. In: Brozek J, Hensche A (eds). Techniques for measuring body composition. National Academy of Sciences: Washington DC pp.223 ? 244.

Wang, Y., Beydoun, M.A., (2007). The Obesity Epidemic in the United States?Gender, Age, Socioeconomic, Racial/Ethnic, and Geographic Characteristics: A Systematic Review and Meta-Regression Analysis. Epidemiological Reviews, 29, pp.6-28.
World Health Organization (1998). Obesity: Preventing and managing a global epidemic. Geneva: WHO
World Health Organization and Food and Agricultural Organization (2002). Diet, nutrition and the prevention of chronic diseases; Report of a joint WHO/FAO expert commission. WHO Technical Report Series, 916. Geneva : WHO
World Health Organisation, West Pacific Region (2003). Diet, food supply and obesity in the Pacific. Geneva:WHO 
World Health Organization (2006). Obesity fact sheet: available [online] http://www.who.int/mediacentre/factsheets/fs311/en/index.html [Accessed March 2007
ETHNIC DIFFERENCES IN OVERWEIGHT AND OBESITY

2.1 Introduction

Ethnic differences have been reported in the prevalence of obesity and its comorbidities (Scarborough et al., 2010). However, ethnicity is a complex epidemiological variable with imprecise and fluid characteristics, changing over short periods of time and with the environment (Chaturvedi 2001, Bhopal 2009). It includes social, cultural, economic and political facets and therefore it is difficult to know if differences in disease prevalence are a result of ethnicity per se or its many facets (Bhopal 2009). It is beyond the scope of this chapter to discuss the use of ethnicity as an epidemiological variable, however a brief summary of the confounding points is shown in appendix 2.1.
The aim of this chapter was to discuss the prevalence of overweight, obesity worldwide and the associated co-morbidities with particular emphasis on ethnic differences. In addition, in this chapter, the use of anthropometric proxy measures used to identify excess adiposity and increased risk of obesity-related disease were analysed critically. The analysis focused on the influence of ethnicity on the widely used Body Mass Index (BMI) and Waist Circumference (WC).


2.2 Classification of obesity

Body mass index is used to classify overweight and obesity. It is a measure of weight in relation to height, calculated by the equation: 
Weight (kg)/ height2 (m)
Table 2.1 shows the classification of overweight and obesity using BMI according to the WHO (2000).

Table 2. 1 World Health Organization body mass index classifications for adults
Normal weightOverweight/ Pre-obeseObese IObese IIBMI (kg/m2)18.5-24.925.0-29.930.0-34.935.0-39.9Source: WHO (2000)

The advantages of using BMI are that it is simple to use, does not require expensive or cumbersome equipment and gives a quick and reliable estimate of nutritional status. However, BMI does not show whether excess weight is a consequence of muscularity or adiposity and therefore very muscular but lean individuals such as athletes, can be classified as overweight or obese (Ode et al., 2007).  More importantly, BMI gives no information on fat distribution, in particular central obesity or abdominal adiposity. The visceral compartment of abdominal adipose tissue is strongly related to increased risk of non-communicable disease particularly CVD (Prentice and Jebb 2001). 


2.3 Global prevalence of overweight and obesity

Overweight and obesity prevalence rates higher than 50% are typically reported in developed, high-income countries. Recently, prevalence and trend estimates for the USA showed that in the period 1999-2004, approximately 57.1% of adults aged 20-39 years and 73.1% of those aged 40-59 years were overweight, of which 28.5% and 36.8%, respectively, were obese (Ogden et al., 2006). The prevalence of obesity among men increased significantly during the six-year period, whereas there was no overall increase among women in the same period. It was estimated that by 2015 75% of adults in the USA would be overweight with 41% obese (Odgen et al., 2006). 
In the UK, a greater proportion of men than women are overweight. In 2008-2009, 68.8% of men were overweight, including obese, of whom 24.2% were obese, and 58.8% of women were overweight including obese, and 31.8% were obese (Mindell and Hiran 2009). The highest prevalence of overweight and obesity was in reported in those aged 65 years and over in both men and women. Mean population BMI in the UK was 27.2kg/m2 in men and 28.0kg/m2 amongst women (Butland et al., 2007). 
The prevalence of overweight and obesity is increasing in LMC (Prentice 2005). Particularly, in middle-income countries such as China, India and Brazil the prevalence of obesity is increasing rapidly (Popkin 1998, Wu 2003, Yoon et al., 2006).  The large size of the populations in these countries mean a greater number of people are overweight and obese. For example, about one fifth of the one billion overweight or obese people in the world are Chinese. Although this is a prevalence rate of 14.7% overweight and 2.6% obese in China, this represents 184 million overweight, and a further 31 million obese people, in China, out of a total population of 1.3 billion (Wu 2003). 
The data on overweight and obesity in some low-income regions, in particular sub-Saharan Africa are scarce. Where available a large amount is for women, between the ages of 15-49 years, and there is a paucity of data in men. Using the available data it is difficult to gain a representative picture of the prevalence and trends of overweight and obesity in these regions. However, some studies have reported increasing trends in these regions. Mendez et al., (2005) investigated the nutritional status data of urban and rural women in 36 middle- and low-income countries, 19 of which were in Sub-Saharan Africa. The prevalence of overweight in urban women ranged from 10.3-69.9% with a median of 32.4%. Overweight prevalence was greater than 20% in 33 countries and greater than 50% in 10 countries. Rural overweight ranged from 3.6-65.9% (median 19.4%) and in 18 countries rural overweight prevalence was greater than 20%.  A marked increase in overweight was associated with a gross national income (GNI) of above US$3000 and urbanisation in greater than 32% of the population. Overall, it was found that a greater number of women were overweight than were underweight in both urban and rural settings in these middle- and low-income countries (Mendez et al., 2005). However, these findings are limited to women.
Abubakari et al., (2007) reviewed 36 studies examining obesity prevalence and trends in adult men and men from the West African countries of Benin, Burkina Faso, Cape Verde, Cote d?Ivoire, Gambia, Ghana, Guinea, Guinea-Bissau, Liberia, Mali, Mauritania, Niger, Nigeria, Senegal, Sierra Leone and Togo and found a mean BMI range from 20.1-27.0 kg/m2; the prevalence of obesity was estimated at 10%. Women were more likely to be obese than men, and urban residents more obese than rural residents. Time trend analysis indicated that obesity in urban West Africa had more than doubled in the last 15 years; women accounted almost entirely for the increase (Abubakari et al., 2007).
Although the prevalence of obesity is rising in all regions, it is unequally distributed within and between countries. Some populations are disproportionately affected by overweight and obesity compared with others. In countries with multiple ethnic groups, differences in the prevalence of obesity have been reported.


2.3.1 Ethnic differences in overweight and obesity prevalence

There are ethnic differences in the prevalence of overweight and obesity and this is perhaps best illustrated by surveys from the USA and the UK. Table 2.2 shows age, gender and ethnicity analysis in the USA from the National Health and Nutrition Examination Surveys (NHANES). 


Table 2. 2 Prevalence of overweight and obesity among people aged 20-60 years in the USA by ethnic group (%)
1999-2002Overweight (25 kg/m2Obese(30 kg/m2Non-Hispanic WhiteNon-Hispanic BlackMexican AmericanGeneral populationNon-Hispanic WhiteNon-Hispanic BlackMexican AmericanGeneral populationGenderAge (years)Men >2069.462.973.168.828.227.927.327.620-396055.464.860.322.924.723.42340-5976.165.080.574.731.329.731.330.7>6074.872.275.473.932.230.527.430.5Women >2057.277.271.761.630.749.038.433.220-3949.070.361.854.524.946.631.229.140-5959.981.580.964.934.950.647.736.7>6066.782.273.968.433.750.335.834.7Source: Flegal et al., (2002) and Wang and Beydoun (2007)


The general population in the USA, against which all minority groups are compared, consists of mainly White or Caucasian people who make up 77.1% of the USA population (US Census Bureaux 2001). Compared with the general population of men the highest rates for overweight were in Mexican American men aged 40-59 years (80.5%), and the highest rates for obesity were in non-Hispanic white and Mexican American men aged >60years (32.2%). Non-Hispanic black men aged 20-39 years had the lowest rates for overweight  (55.4%) and Non-Hispanic white men aged 20-39 years had the lowest rates of obesity compared with the general population.

Among women, the highest rates for overweight were in non-Hispanic black women (81.5%) aged 40-59 years, and obesity was highest in this same group (50.6%) compared with the general population. Non-Hispanic white women aged >20 years had the lowest rates of overweight (57.2%). The same group, aged 20-39 years had the lowest rates of obesity compared with the general population.

Table 2.3 shows the age-standardised prevalence of overweight including obesity (BMI>25 kg/m2), severe obesity (BMI>30 kg/m2) and mean by ethnic group in the UK (Primatesta et al., 2004). The general population, against which all minority groups were compared, consists of mainly White Caucasian people who make up 92.1% of the UK population (UK Census 2001).


Table 2. 3 Prevalence of overweight and obesity among people aged 16 and over in England by ethnic group (%)
Black Caribbean Black African IndianPakistani Bangladeshi Chinese Irish General Population GenderMen Overweight (and obese)67.461.853.255.544.436.867.166.5Obese (and severely obese) 25.217.113.815.15.86.025.222.7Mean BMI (kg/m2)27.126.425.825.924.724.127.227.1Women Overweight (and obese)64.569.855.262.350.824.958.057.1Obese (and severely obese)32.138.520.228.117.27.621.223.2Mean BMI (kg/m2)2828.826.227.125.723.226.726.8Source: Health Survey for England (2004) Overweight: BMI(25 kg/m2, Obese: BMI(30 kg/m2

Among men, Black Caribbean men had the highest prevalence of overweight (and obese) (67.4%) and obese (and severe obese) (25.2%) compared with the general population. Chinese men had the lowest prevalence for overweight and obese 36% and 6.0% respectively. Among women, Black African women had the highest prevalence of overweight (69.8%) and obese (38.5%) while Chinese women had the lowest prevalence of overweight (24.9%) and obese (7.6%).
Among men, the mean BMI of Chinese (24.1 kg/m2), Bangladeshi (24.7 kg/m2), Indian (25.8 kg/m2) and Pakistani (25.9 kg/m2) men was lower than that of men in the general population (27.1 kg/m2). Among women, those of Chinese origin had a lower mean BMI (23.2 kg/m2) than those in the general population (26.8 kg/m2). In contrast, mean BMI was highest among Black Caribbean (28.0 kg/m2) and Black African (28.8 kg/m2) women. Indian women had a similar mean BMI to women in the general population (Primatesta et al., 2004). In the projections of the Foresight (2007) report, only slight increases in overweight and obesity are expected among Indian men and women. In contrast, levels of obesity are expected to rise among Black African women, Pakistani men and women compared with the general population (Butland et al., 2007). Ethnic differences in the prevalence of overweight and obesity result in differences in the prevalence of obesity-related disease. 

2.4 Overweight and obesity-related disease

Mortality and morbidity are substantially increased by obesity; however, the effects of overweight (BMI 25-29kg/m2) are controversial (Adams et al., 2006).
Must et al., (1999) described the relationship between weight status (measured by BMI) and prevalence of health conditions including T2DM, gallbladder disease, coronary heart disease, high blood cholesterol level, high blood pressure and osteoarthritis, by severity of overweight and obesity in the USA population. Data from the third National Health and Nutrition Examination Survey (NHANES III) for men and women was used. It was found that for both men and women, high blood pressure was the most common overweight- and obesity-related health condition and its prevalence showed a strong increase with increasing weight status. In addition, the prevalence of T2DM, gallbladder disease, and osteoarthritis increased sharply among both overweight and obese men and women corresponding with the increasing BMI category. 

Flegal et al., (2005) estimated relative risks of mortality associated with different levels of BMI from (NHANES) I, II and III. Compared with the normal weight category, obesity was associated with 111,909 excess deaths and underweight with 33,746 excess deaths. However, overweight was not associated with excess mortality. Using the same data Flegal et al., (2007) estimated cause-specific excess deaths associated with underweight, overweight and obesity using data from 2.3 million adults 25 years and older from 2004 vital statistics data for the USA who died from CVD (CVD), cancer, and all other causes (non-cancer, non-CVD causes). Underweight was associated with significantly increased mortality from non-cancer, non-CVD causes (23,455 excess deaths) but not associated with cancer or CVD mortality. Overweight was associated with significantly decreased mortality from non-cancer, non-CVD causes (?69,299 excess deaths) but not associated with cancer or CVD mortality. Obesity was associated with significantly increased CVD mortality (112,159 excess deaths) but not associated with cancer mortality or with non-cancer, non-CVD mortality. In further analyses, overweight and obesity combined were associated with increased mortality from diabetes and kidney disease (61,248 excess deaths) and decreased mortality from other non-cancer, non-CVD causes (?105,572 excess deaths). Obesity was associated with increased mortality from cancers considered obesity-related (13,839 excess deaths) but not associated with mortality from other cancers. Comparisons across surveys suggested a decrease in the association of obesity with CVD mortality over time. 

The finding of decreased association between overweight and excess mortality made by Flegal and colleagues (2005, 2007) has been challenged. It has been argued that reasons for low weight, often due to chronic disease, which may have lead to death, were not taken into consideration (Willet et al., 2005). Moreover, estimating excess mortality due to overweight is greatly exacerbated in elderly persons because of the high prevalence of chronic disease and loss of lean mass; analyses that simply used current BMI may be particularly misleading in this group. Secondly, it was argued that overweight persons have increased prevalence of lipid abnormalities, hypertension, and diabetes, and increased rates of coronary heart disease and many cancers. Thus, the lower mortality in the overweight group might have been due to either an undiscovered powerful beneficial physiological effect of overweight or be due to artifact (Willet et al., 2005).
In contrast to the findings of Flegal et al., (2005, 2007), Adams et al., (2006) examined BMI in relation to the risk of death from any cause in 527,265 USA men and women aged 50 to 71 years old. Over the course of a follow-up of 10 years, 1995-2005, 61,317 participants (42,173 men and 19,144 women) died. Initial analyses showed an increased risk of death for the highest and lowest categories of BMI among both men and women, in all racial or ethnic groups, and at all ages. To control for confounders introduced by a low BMI as a result of smoking or chronic disease, the analysis was restricted to people without chronic disease who had never smoked. It was found that the risk of death was associated with both overweight and obesity among men and women. In analyses of BMI at 50 years among those who had never smoked, the associations became stronger with the risk of death increasing by 20-40% among overweight persons and by two to at least three times among obese persons (Adams et al., 2006).


2.4.1 Global prevalence of obesity-related disease

This section will discuss the prevalence of specific obesity-related disease, including T2DM, hypertension and CVDs

2.4.1.1 Cardiovascular Disease

Globally CVDs (CVD) account for more deaths than any other disease (WHO 2004) (Figure 2.1). In 2004, an estimated 17.1 million people died from CVD representing 31.5% and 26.8% in men and women respectively. Of these deaths, an estimated 7.2 million were due to CHD and 5.7 million were due to stroke (WHO 2009b).


Figure 2. 1 Distribution of deaths, globally by leading cause Source: WHO (2004)


2.4.1.2 Type II diabetes mellitus

More than 220 million people worldwide have diabetes. In 2005, an estimated 1.1 million people died from diabetes (WHO 2009c). A recent estimate from the International Diabetes Federation (IDF 2010) show that the prevalence of diabetes in adults aged 20-79 years was 6.4% (285 million people) and is expected to increase to 7.8% (438 million people) by 2030. 
Almost 80% of diabetes deaths occurred in LMC. Almost half of diabetes deaths occurred in people under the age of 70 years; 55% of diabetes deaths were in women. The most important demographic change to diabetes prevalence across the world appears to be the increase in the proportion of people >65 years of age (Wild et al., 2004). Type 2 diabetes mellitus comprises 90% of people with diabetes around the world, and is largely the result of excess body weight and physical inactivity (WHO 2009, IDF 2010).

2.4.1.3 Hypertension

Globally, in 2000, 26?4% of the adult population had hypertension and this was projected to rise to 29?2% by 2025 (Kearney et al., 2005). The estimated total number of adults with hypertension in 2000 was 972million; 333million in developed countries and 639million in developing countries. The number of adults with hypertension in 2025 was predicted to increase by about 60% to a total of 1?56 billion. Men and women had similar prevalence of hypertension, and that prevalence increased with age consistently in all world regions. A particularly high prevalence of hypertension was reported in Latin America and the Caribbean, and Asia and the Pacific Islands had the lowest prevalence (Kearney et al., 2005).


2.4.2 Ethnic differences in obesity-related disease

Obesity-related disease varies by ethnic group.  Compared with people of Caucasian origin, people of black African origin have a higher incidence of stroke and end-stage renal failure (ESRF), whereas coronary heart disease (CHD) is less common (Gaillard et al., 2009). People of Indian sub-Continent origin (Asians or South Asians) have a much higher incidence of CHD. The high mortality from CHD amongst South Asians in their countries origin indicates that the high risk is a feature of these populations, globally. This risk remains even when the reference host population is at different risk. Similarly, the mortality from stroke in people of African origin is high, whether born in the migrated countries or in their countries of origin (Capuccio et al., 2003).

The theme of Health Survey for England (HSE) in 2004 was the health of minority ethnic groups. Figure 2.2 shows the results of analyses of the prevalence of CVD (HSE 2004). Cardiovascular disease referred to IHD, CHD and stroke.


Figure 2. 2 Prevalence of CVD by ethnic group and age in men and women 
Source: Health Survey for England (2004)

The graphs show that for all groups, CVD was highest among those over 55 years. It was reported that the prevalence of CVD was highest in Indian women (18.9%) and Pakistani men (41.1%) and lowest in the Chinese group (8.7% for men and 9.0% for women). Black Africans, aged 55 years and over, had a low prevalence of IHD -5.2% for men and 1.5% for women. Overall, Irish men were reported to have the highest prevalence of any CVD (14.5%) while Black African men had the lowest (2.3%). Women from the general population had the highest prevalence of any CVD (13.0%) and Bangladeshi women had the lowest (4.8%). 
In comparison with the 1999 HSE (Primatesta and Brookes), which also focused on the health of ethnic minorities, there was a general but not significant increase in the prevalence of CVD between 1999 and 2004 in most minority ethnic groups and in the general population. The only significant increase was among Pakistani men, where the prevalence of CVD almost doubled between the survey years (6.3% in 1999 to 12.0% in 2004). 
Recently updated statistics from the British Heart Foundation (2010) showed that certain ethnic groups had an increased prevalence of specific CVDs. For example the incidence of MI was higher in South Asians than in non-South Asians for both men and women. Afro-Caribbeans had a higher prevalence of stroke compared with Caucasians for both men and women. The prevalence of CHD was highest in Indian (6%) and Pakistani (8%) men. Moreover, Black Caribbean, Indian, Pakistani and Bangladeshi men have a considerably higher prevalence of diabetes than the general population (Scarborough et al., 2010).

It has been suggested that the causes of ethnic differences in obesity-related disease (in particular CVD), are a combination of environmental, cultural (including diet and physical activity, smoking and alcohol consumption behaviour) and socio-economic factors (Kurian and Cardarelli 2007). However, variations in these risk factors do not explain the excessive mortality of certain groups e.g. people of African origin, from certain conditions (e.g. stroke) and not others (e.g. CHD).
From the perspective of body composition, ethnic differences in the distribution of fat may explain differences in disease prevalence. It is well established that central obesity is a key determinant of obesity-related diseases. Central obesity has been identified as a key factor in the development of the metabolic syndrome (Alberti et al., 2006, Eckel et al., 2005). The metabolic syndrome (MetS) describes clustering of risk factors in order to identify increased risk of T2DM and CVD (Eckel et al., 2005). According to new standardised guidelines set by IDF, a key feature of the metabolic syndrome (MetS) is central obesity (increased abdominal VAT) (Alberti et al., 2006). In addition to central obesity, two of four of the following factors result in the diagnosis of MetS and increased risk of CVD and diabetes.

Raised triglyceride levelCVD and diabetes.

Raised triglyceride level: ? 1.7 mmol/l (150 mg/dl) 

Reduced HDL-cholesterol: < 1.03 mmol/l (40 mg/dl) in males and <1.29mmol/l (50 mg/dl) in females (or specific treatment for these lipid abnormalities) 

Raised blood pressure (systolic BP ? 130 or diastolic BP ? 85 mmHg) (or treatment of previously diagnosed hypertension) 

Raised fasting plasma glucose [FPG ? 5.6 mmol/l (100 mg/dl)] (or previously diagnosed type 2 diabetes).

Insulin resistance promotes each of the features ofaetiologic factor in the development of the MetS (Sumner 2009). Central obesity is associated with insulin resistance and thus it is the key feature of metabolic syndrome. However, the prevalence of MetS is very low in people of African origin compared with Caucasian origin populations. This is despite higher insulin resistance in populations of African origin (Sumner 2009). 


2.5 Central obesity

Central obesity describes the accumulation of adipose tissue in the upper body, particularly in the abdominal area. This type of fat distribution has been described as android obesity as opposed to gynoid obesity, which describes the accumulation of excessive fat in the gluteal region (Vague 1974). Abdominal fat is separated into subcutaneous adipose tissue (SAT) and visceral adipose tissue (VAT). Of the two, VAT is more strongly associated with obesity-related disease. Moreover, it is the differences in the size of this compartment that may explain ethnic differences in obesity-related disease (Lear et al., 2007). 

2.5.1 Classification of central obesity

Central obesity is classified using WC and waist-to-hip ratio (WHR). Waist-to-hip ratio is calculated by the following equation
Waist circumference (cm)/ Hip circumference (cm)
Table 2.4 shows the cut-off points for WC and WHR defined by the WHO and International Diabetes Federation (IDF). 

Table 2. 4 Waist circumference and waist-to-hip ratio cut-off points for central obesity, reflecting increased risk of CVD
WCWHRInc risk*Substantially inc*M (cm)F (cm)M (cm)F (cm)MFWHO/IASO/IOTF (2000)/ IDF (2006)9480102880.800.95*Inc risk= Increased risk of CVD at this level and above, *Substantially inc=Substantially increased risk of CVD at this level and above, WHO=World Health Organisation, IASO=International Association for the Study of Obesity, IOTF= International Obesity Task Force, IDF=International Diabetes Federation, WHR=Waist-to-hip ratio, Waist circumference =WC

Waist-to-hip ratio (WHR) is used to distinguish increased upper body fat distribution (android obesity) from increased lower body adipose tissue (gynoid obesity). However, WC is the most widely used anthropometric measure of central obesity and CVD risk. Recently the IDF (2006) recommended that WC alone was a sufficient indicator of metabolic syndrome, and is a central feature of the IDF?s definition of metabolic syndrome (Alberti et al., 2006). 
Although not a precise quantification of abdominal adiposity, WC and WHR are positively correlated with abdominal visceral adipose tissue (VAT) determined by computed tomography. In addition the measures do not require expensive or cumbersome equipment or technical expertise to obtain a reliable estimate of central obesity (Wang 2003, Kagawa et al., 2008).
However, no standardised protocol exists for the measurement of WC (Wang et al., 2003). The WHO recommends that WC be measured at the mid-point between the lowest rib and the iliac crest; while the National Institutes of Health (NIH) recommend its measurement just above the iliac crest (Wang et al., 2003). Other measurement sites include the minimal circumference of the torso, mid-way between the iliac crest and xiphoid process, at the level of the umbilical cord and at the level of the iliac crest. There is no scientific rationale for the use of one site over another (Wang 2003, Wang et al., 2003, Willis et al., 2007). 
Willis et al., (2007) compared the measurement of WC at the level of umbilical cord and minimal waist in men and women across different BMI ranges. In women, minimal WC and umbilical WC differed by on average 10cm while in men it differed by 4cm. The minimal WC site correlated highly with CVD risk factors and metabolic syndrome compared with the umbilical site in women only. It was concluded that the minimal WC site predicted CVD risk in women better than the umbilical site (Willis et al., 2007). In men, neither of the WC sites correlated highly with cardiovascular risk factors. This is the first study to provide scientific backing for the use of any WC over the others, although the differences were reported only in women (Willis et al., 2007). 
Waist circumference is complementary or superior to BMI in its association with CVD risk factors. A WC of 90cm for men and of 83cm for women has an equivalent risk of CVD similar to a BMI of 25kg/m2 whereas a WC of 100cm for men and of 93cm for women corresponds to CVD risk factors at a BMI of 30kg/m2 (Zhu et al., 2002). However, these cut-off points reflect associations in non-Hispanic whites. Ethnic differences in WC have been reported and are discussed in the following section.

2.5.2 Ethnic differences in central obesity

Table 2.5 shows the prevalence of central obesity and the risk of CVD, estimated by WC in different ethnic groups in the UK.

Table 2. 5 Prevalence of central obesity in minority ethnic groups in the UK
GenderBlack Caribbean Black African Indian Pakistani Bangladeshi Chinese IrishGeneral Population MenMean WC (cm)92.590.693.095.088.786.897.396.5% with WC over 102cm221920301283331WomenMean WC (cm)88.490.283.987.785.777.687.486.4% with WC over 88cm4753384843164341Source: Health Survey for England (2004)

Amongst men, Irish men had the largest mean WC (97.3cm) and the largest number of men at substantial risk of CVD defined by a WC of greater than 102cm. Chinese men had the smallest WC (86.8cm) and the lowest proportion of men at increased risk of CVD (8%). Amongst women, Black African women had the largest WC (90.2cm) and the highest proportion (53%) with substantially increased risk of CVD. 
2.5.3.1 Ethnic differences in visceral adiposity

It is well established that people of Asian origin are less lean but tend to carry more visceral adipose tissue compared with people of Caucasian origin (Deurenberg et al., 2002). This difference in VAT explains the greater prevalence of obesity-related disease in people of Asian origin compared to those of Caucasian origin (McKeigue et al., 1991). However, the difference in obesity-related disease and visceral adiposity in people of African origin compared with Caucasians remains unclear.

The majority of data available on fat distribution in populations of African origin has been collected in AAs. African Americans have a greater burden of obesity and obesity-related disease compared with Caucasian Americans (Wang and Beyoun 2007). However, substantial differences in environments and culture exist, such that African-American and other African populations cannot be grouped together, or assumptions made in AAs applied to other African populations (Agyemang et al., 2005, Lear et al., 2010). Grouping together all people of African origin may mask differences in body composition that may relate to CVD. For the purposes of illustrating ethnic differences in abdominal visceral adipose tissue, the available data on AAs will be discussed.

Early studies of AAs, in particular AA women have reported less abdominal VAT compared with their Caucasian counterparts. Conway et al., (1995) reported differences in VAT between obese age- and weight- matched AA (n=9) and Caucasian American (CA) women (n=11) before and after weight loss. Computed tomography (CT) scans were used to assess abdominal VAT and subcuteanous adipose tissue (SAT). African American women had 5% less VAT compared to CA women and 4-6% more SAT than CA women. The lower %VAT did not change after weight loss. Biochemical measures in this group of AA women showed lower plasma triglycerides, lower blood glucose, lower blood cholesterol, higher HDL cholesterol in AA compared with Caucasian women. Results did not support increased abdominal VAT as the explanation for higher risk for obesity-related morbidity in AA women nor did the biochemical measures differences show an increased risk in AA compared with Caucasian women. 
Lovejoy et al., (1996) compared abdominal VAT and insulin sensitivity in 37 AA women matched for age and weight to 22 CA women. African American women had significantly less VAT (98.0 cm2) compared with CA women (117.3cm2), despite similar BMI and WHR. Moreover, AA women were less insulin sensitive compared with CA women and WHR was significantly correlated with fasting insulin and serum cholesterol in CA but not AA women. It was concluded that body fat distribution and risk factors for CVD were different between AA and Caucasians.
Perry et al., (2000), compared abdominal VAT, and the ability of its anthropometric surrogates (WC and WHR) to predict serum lipoproteins, blood pressure and insulin resistance in 36 CA and 30 AA overweight (BMI of >27.3kg/m2) women.   Magnetic resonance imaging (MRI) showed higher VAT (128.2 cm2) in Caucasian than AA women (81.5 cm2) with similar BMI values, mean BMI was 34.8kg/m2 AA women and 35.1kg/m2 in CA women. 
These studies provided the first evidence of ethnic difference in abdominal VAT between AA and CA women. However, the sample sizes were small and the findings may not be applicable to males. Hill et al., (1999) obtained cross-sectional data from the CARDIA study in order to determine whether the amount of abdominal VAT in body fat differed in 391 men and women of which, 100 were AA men, 90 AA women, and 96 CA men and 105 CA women. For any given body fatness, AA men had less abdominal VAT (73.1cm2) than CA men (99.3 cm2), there was no difference in SAT.  Conversely, AA women who were more obese than CA women had higher abdominal VAT (75.1cm2) than CA women (58.6 cm2). This difference was apparent only when adjusted for WC or sagittal diameter not when adjusting for total body fat or BMI


In the largest study to date, which included 790 CA women, 435 AA women, 606 CA men and 136 AA, differences in abdominal VAT and SAT were investigated (Katzmarzyk et al., 2010). Abdominal VAT was significantly higher in CA men (148.6 cm2) than AA men (97.7 cm2), and higher in CA women (126.8 cm2) than AA women (96.7 cm2). Abdominal SAT was higher in CA men 319.3 (141.8cm2) than AA men (289.4 (156.6cm2). Conversely, abdominal SAT was higher in AA women (452.7 (141.6cm2) than CA women (412.0 cm2). However, after adjustment for age, total body fat mass, smoking and menopausal status, SAT was lower in higher in AA men than CA men, but remained unchanged in the women. This study reinforced the ethnic differences in abdominal VAT between AA and CAs.
The findings of lower abdominal VAT have been reported in other populations of African origin compared with Caucasian origin groups; for example between black and Caucasian women in urban South Africa (Van der Merwe et al., 2000, Crowther et al., 2006, Jennings et al., 2008) and Afro-Caribbeans and Caucasians in the UK  (Gailliard et al., 2009). However, the paradox of low visceral adiposity but higher risk of CVD in African origin compared with Caucasian origin populations remains. 
Anthropometric measures used in the identification of excessive adiposity and CVD risk may be implicated in the controversy related to ethnic differences in adiposity and cardiovascular risk. The following section discusses two commonly used measures of obesity in relation to ethnic differences that may influence their appropriateness for use in all ethnic groups.


2.6 Ethnic differences in obesity measurement

The basis of proxy measures of adiposity, such as BMI and WC, are based on laboratory measures developed in predominately Caucasian populations (Deurenberg et al., 2003). Thus their ability to identify overweight and obesity, and related disease risk in non-Caucasian groups is questionable. The following sections, will analyse critically the use of BMI and WC as measures of obesity in different ethnic groups.

 2.6.1 Body mass index

The normal range for BMI (18.5-24.9 kg/m2) was based on mortality estimates from USA insurance data. Lew and Garfinkel (1979) plotted BMI against mortality risk in 750,000 USA insurance holders as shown in figure 2.3. 



Figure 2. 3 Mortality risk versus BMI in US men and women 
(Lew and Garfinkel 1979)
The relationship between BMI and mortality was a ?U? shaped curve, which is indicative of increased mortality at very low BMI and very high BMI (Lew and Garfinkel 1979). The optimum range for BMI was between 18.5 kg/m2 and 24.9 kg/m2. This range was associated with the lowest mortality, therefore, the WHO adopted these cut-off points to define normal or desirable weight range, which is associated with the lowest risk of mortality. However, because these cut-off points were established in a predominately Caucasian population they may not represent a similar level of risk in all ethnic groups (Razak et al., 2005, Pan and Yeh 2008).
The use of mortality risk to identify BMI cut-off points in all ethnic groups may be inappropriate. There is an increase in mortality with increased weight, however, one condition, e.g. the risk of ischemic heart disease and haemorrhagic stroke increases progressively from a BMI of 20, this is not the case for all CVDs (Ezzati et al., 2002). Stevens et al., (2002) attempted to establish BMI cut-off points in a group of AA women against a group of CA women at a BMI of 30kg/m2, using the incidence of diabetes, hypertension and hypertriglyceridemia as indicators of disease. It was found that a large range in BMI values for AA women was associated with risks equivalent to those of the CA women at a BMI of 30kg/m2. In addition, BMI cut-off points differed for the AA women in diabetes, compared to hypertension compared to hypertriglyceridemia equivalent to the CA women at a BMI of 30kg/m2 (Stevens et al., 2002). 
Body frame size, composition of bone and muscle, and fat distribution varies among people of different ethnicities. People of Asian origin have a smaller frame size and increased abdominal fat compared with those of Caucasian or Asian Pacific origin (Rush et al., 2009). These differences in body composition result from complex interactions between genetic and environmental factors. Such that a measure of obesity should control for the differences in frame size.


Body mass index controls for frame size by dividing body weight (kg) by the square of the height (m) (Quetelet et al., 1842). However, due to variations in body shape and skeletal structure in different groups, BMI is not completely independent of height. Thus, the universal use of BMI to define overweight and obesity in all ethnic groups may be inappropriate because of differences in height (Rush et al., 2009). 

A higher incidence of obesity-related disease, in particular T2DM, has been reported in populations of Asian origin compared with their Caucasian counterparts at similar BMI (Dudeja et al., 2001, Duerenberg-Yap et al., 2002a). In 2004, the WHO recommended lower cut-off points for defining overweight and obesity in populations of Asian origin. These were 23 kg/m2 for overweight, 27.5 kg/m2 for obese I, 32.5 kg/m2 for obese II and 37.5 kg/m2 for morbid obesity. These additional cut-off points and original cut-off points are shown in figure 2.4.


Figure 2. 4 WHO obesity action points for determining public health and clinical action points in Asian populations. 
Source: WHO (2004)


The additional cut-off points have not been universally adopted in all populations of Asian origin.  It has been reported that people of Asian origin differ from each other in body composition and fat distribution (Gurrici et al., 1999, Deurenberg et al., 2002). Thus, it has been argued that the lower cut-off points shown in figure 2.4 may not be applicable to all populations of Asian origin (Pan and Yeh 2008). It was concluded that a wide range of BMI cut-offs could be used for Asian populations, dependent on the outcome being assessed and the population examined (WHO 2004).

2.6.2 Waist Circumference

Waist circumference may not represent the same level of abdominal fatness in all ethnic groups. There are no standardised and calibrated normal ranges for WC cut-off points associated with equivalent risks varying with age, gender and ethnicity (Wang 2003, Wu et al., 2007). Currently, the WHO recognises a range of 94-101.9cm in men and 80-87.9cm in women for increased risk corresponding to a BMI of 25-29.9 kg/m2 (WHO 2000). However, these cut-off points are derived in predominately Caucasian populations and thus the applicability in non-Caucasian groups is questionable. Table 2.6 shows the proposed IDF cut-off points for South Asians, Europeans, Japanese and Sub-Saharan Africans (Alberti et al., 2006).


Table 2. 6 Ethnic specific cut-off points for WC from International Diabetes Federation
Country/ ethnic groupWC 
(as a measure of central obesity)Europids* Male( 94cmFemale( 80cmSouth Asians**Male( 90cmFemale( 80cmChineseMale( 90cmFemale( 80cmJapanese***Male( 85cmFemale( 90cmEthnic South and Central AmericansUse South Asian recommendation until more specific data are availableSub-Saharan AfricansUse European recommendation until more specific data are availableEast MediterraneanUse European recommendation until more specific data are availableSource: The IDF worldwide consensus definition of the metabolic syndrome (Alberti et al., 2006)*The US ATP II values (102cm in males and 88cm women) continue to be used
**Based on Chinese, Malay and Asian Indian population
*** Subsequent data analyses suggest that Asian values (male, 90cm; female 80cm) should be used for Japanese populations until more data are available.

It is well established that despite low BMIs and body weight compared with Caucasians, people of South Asian origin have a greater amount of central fat and in particular visceral adipose tissue (WHO 2004). In addition, AAs have been found to have a lower VAT at any given BMI compared with Caucasians (Conway et al., 1995, Lovejoy et al., 1996, Katzmarzyk et al., 2010). Epidemiological studies are yet to establish cut-off points for South and Central Americans, Sub-Saharan Africans and East Mediterranean groups.
Waist circumference does not differentiate between abdominal SAT and VAT (Lear et al., 2010). Thus, WC measurements may not represent the same levels of abdominal fat. This has implications for defining CVD risk in different ethnic groups as it has been shown that the size of the visceral compartment varies with ethnicity (Perry et al., 2000, Carroll et al., 2008).

2.7 Conclusion 

The aim of this chapter was to discuss ethnic differences in obesity prevalence and the prevalence of obesity-related disease. It has been shown that the prevalence of obesity and obesity-related disease varies in different ethnic groups. Moreover, proxy measures of adiposity, namely BMI and WC developed in Caucasian origin populations may not appropriately identify increased risk in non-Caucasian groups. Therefore obesity and its comorbidities may be underestimated in non-Caucasian populations using these established proxy measures. 


2.8 References

Adams, K.F., Schatzkin, A., Harris, T.B., Kipnis,  V., Mouw, T., Ballard-Barbash, R., Hollenbeck, A., and. Leitzmann, M.F., (2006). Overweight, obesity, and mortality in a large prospective cohort of persons 50 to 71 years old. New England Journal of Medicine, 355 pp.763-78.

Alberti, K. G. M. M., Zimmet, P., Shaw, J., (2006). Metabolic syndrome?a new worldwide definition. A Consensus Statement from the International 
Diabetes Federation. The Lancet , 366, pp.1059-1062.
Bray, G.A., (2004). Medical consequences of obesity. Journal of Clinical Endocrinology and Metabolism.  89, pp.2583-2589.
Cappuccio, F.P., Barbato, A., Kerry, S.M., (2003). Hypertension, diabetes and cardiovascular risk in ethnic minorities in the UK. British Journal of Diabetes and Vascular Disease, 3, pp.286?293.

Carroll, J.F., Chiapa, A.L., Rodriquez, M., Phelps, D.R., Cardarelli, K.M., Vishwanatha, J.K., Bae, S., Cardarelli, R., (2008). Visceral fat, WC and BMI: Impact of race/ethnicity. Obesity, 16, pp.600-607.
Chaturvedi, N., (2001). Ethnicity as an epidemiological determinant- crudely racist or crucially important. International Journal of Epidemiology, 30, pp.925-927.
Chodhury, M., Falaschetti, E., Fuller, E., Mackenzie, H., Mindell, J., Nicholson, S., Pickup, D., Roth, M., Scholes, S., Tabassum, F., Thompson, J., Wardle, H., (2007). Health Survey for England 2007: Healthy lifestyles, knowledge attitude and behaviour. [online] available at: . [Accessed on January 2009]
Conway, J.M., Yanovski, S.V., Avila, N.A., Hubbard, V.S., (1995). Visceral adipose tissue differences in black and white women. American Journal of Clinical Nutrition, 61, pp.765-771.
Deurenberg, P., Duerenberg-Yap, M., & Guricci, S., (2002). Asians are different from Caucasians and from each other in their body mass index/body fat percentagerelationship. Obesity Reviews, 3, pp.141-146.
Eckel, R.H., Grundy, S.M., Zimmet, P.Z., (2005). The metabolic syndrome. The Lancet, 365, pp.1415-1428.
Flegal, K.M., Graubard, B.I.,Williamson, D.F., Gail, M. H., (2005). Excess deaths associated with under weight, overweight, and obesity. Journal of the American Medical Association, 293, pp.1861-1867.

Flegal, K.M., Graubard, B. I., Williamson, D.F., Gail, M. H., (2007). Cause specific excess deaths associated with underweight, overweight and obesity. Journal of the American Medical Association, 298, pp.2028-2037.

Gaillard, T., Schuster, D., Osei, K., (2009). Metabolic syndrome in black people of African diaspora: The paradox of current classification, definition and criteria. Ethnicity and Disease, 19, pp.S2-1 ? S2-7.
International Diabetes Federation (2010). Diabetes Atlas [online] available at http://www.diabetesatlas.org/ [Accessed on July 2010]

Kagawa, M., Bryne, N.M. Philips. A.P., (2008). Comparison of body fat estimation using waist:height ratio using different ?waist? measurements in Australian adults. British Journal of Nutrition, 100, pp.1135-1141.

Kearney, P.M., Whelton, M.,  Reynolds, K., Muntner, P.,  Whelton, P.K., He, J., (2005). Global burden of hypertension: analysis of worldwide data. The Lancet, 365, pp.217?223.
Katzmarzyk, P.T., Bray, G.A., Greenway, F.L., Johnson, W.D., Newton, Jr,  R.L., Ravussin,  E.,  Ryan, D.H., Smith, S.R., Bouchard, C., (2010). Racial differences in abdominal depot-specific adiposity in white and African American adults. American Journal of Clinical Nutrition, 91, pp.7-15.
Krotkiewski, M., Bjorntorp, P., Sjostrom, L., Smith, U., (1983). Impact of obesity on metabolism in men and women: importance of regional adipose tissue distribution. Journal of Clinical Investigation, 72, pp.1150-1162.

Kurian, A.K., Cardarelli, K.M., (2007). Racial and ethnic differences in CVD risk factors: A systematic review. Ethnicity and Disease, 17, pp.143-152.

Lean, M.E., Han, T.S., Morrison, C.E., (1995). Waist circumference as measure for indicating weight management. British Medical Journal, 311, pp. 158-161.

Lear, S.A., Humphries, K. H., Kohli, S., Chockalingam, A., Frohlich, J.J., Birmingham, C.L., (2007). Visceral adipose tissue accumulation differs according to ethnic background: results of the Multicultural Community Health Assessment Trial (M-CHAT). American Journal of Clinical Nutrition, 86, pp.353-359.

Mindell, J., and Hirani, V., (2009) Physical measurements. National Diet and Nutritional Survey [online] available at http://www.food.gov.uk/multimedia/pdfs/publication/ndnsreport0809.pdf [Accessed September 2010]

McKeigue, P. M., Shah, B., Marmot, M.G., (1991). Relation of central obesity and insulin resistance with high diabetes prevalence and cardiovascular risk in South Asians. The Lancet, 337, pp.382 ? 386.

Must, A., Spadano, J., Coakley, E.H., Field, A.E., Colditz, G., Dietz, W.H., (1999). The disease burden associated with overweight and obesity. Journal of the American Medical Association. 282, pp.1523-1529.

Ode, J.J., Pivarnik, J.M., Reeves, M.J., Knous, J.L., (2007). Body Mass Index as a Predictor of % Fat in College Athletes and Nonathletes. Medicine & Science in Sports & Exercise, 39, pp.403-409.

Odgen, C.L., Carroll, M.D., Curtin, L.R., McDowell, M.A., Tabak, C.J., Flegal, K.M., (2006). Prevalence of overweight and obesity in the United States, 1999-2004. Journal of the American Medical Association, 295, pp.1549-1555.
Prentice, A., Jebb, S., (2001). Beyond body mass index. Obesity Reviews, 2,?pp.141???147.

Prentice, A., (2006).  The emerging epidemic of obesity in developing countries. International Journal of Epidemiology, 35, pp.93-99.

Primatesta, P., (2004). Health Survey for England Health of Ethnic Minorities available [online] at www.doh.gov.uk .[Accessed January 2008]
Quetelet, L.A.J., (1842). "A treatise on man and the development of his faculties." In: Comparative statistics in the 19th century. Edinburgh: William and Robert Chambers.

Vague., J., (1947) La differenciation sexuelle, facteur determinant des formes de l?obesite La Presse Medical 53 pp 339. 

Wang, J., (2003). Waist circumference: a simple, inexpensive, and reliable tool that should be included as part of physical examinations in the doctor?s office. American Journal of Clinical Nutrition, 78, pp.902?903. 

Wild, S., Roglic, G., Green, A., Sicree, R., King, H., (2004). Global prevalence of diabetes estimates for the year 2000 and projections for 2030. Diabetes Care, 27, pp.1047?1053.

Willett, W.C. Hu, F.B. Colditz, G.A., Manson, J.A.E., (2005). Underweight, overweight, obesity and excess deaths. Journal of the American Medical Association, 294, pp.551

Wu, Y., (2003). Overweight and obesity in China. British Medical Journal, 333, pp.362-363.?
World Health Organization (1998). Obesity: Preventing and managing a global epidemic. Geneva: WHO
World Health Organization (2004). Appropriate body-mass index for Asian populations and its implications for policy and intervention strategies. The Lancet, 363, pp.157?163.

World Health Organization (2006). Overweight and obesity. Fact sheet No 311 [online] available at: http://www.who.int/mediacentre/factsheets/fs311/en/index.html  [Accessed January 2007].
World Health Organization (2009b). CVDs. Fact sheet No317 [online] available at: http://www.who.int/mediacentre/factsheets/fs317/en/index.html [Accessed on  September 2009].
World Health Organization (2009c). Diabetes. Fact sheet No312 [online] available at: http://www.who.int/mediacentre/factsheets/fs312/en/index.html [Accessed on November 2009]
Zhu, S., Wang, Z.M., Heshka, S., Heo, M., Faith, M.S., Heymsfield, S.B., (2002) Waist circumference and obesity-associated risk factors among non-Hispanic whites in the third National Health and Nutrition Examination Survey: 
clinical action thresholds. American Journal of Clinical Nutrition, 76, pp.743?749.
AN HISTORICAL REVIEW OF BODY COMPOSITION ASSUMPTIONS AND METHODS 


3.1 Introduction 

Methods used to estimate body fat rely on assumptions derived from the chemical analysis of cadavers. In cadaver analysis the body is apportioned into fat, muscle, skin, and bone. The chemical analysis describes the compartments? chemical composition and its properties. Models of body composition have been developed to reflect the relationship between the body?s components and its properties. For example, the widely used two-compartmental model of body composition, assumes that body weight is divided into two compartments, fat and FFM, both with constant characteristics.
The majority of the body composition assumptions were derived from the analysis of a limited number of cadavers. Between 1945 and 1968 the classic studies by Mitchell et al., (1945), Widdowson et al., (1951), Forbes et al., (1956) and Moore et al.,(1968), resulted in the analysis of eight cadavers. The cadavers were predominately male and of Caucasian origin, therefore it has been questioned whether the assumptions derived from these analysis can be generalised to other population groups.
The aim of this chapter is to critically review the development of body composition assumptions and methods from cadaver analysis. Ethnic differences and the implications of these differences in body composition will be discussed. In addition, the three methods to be used in studies in this research programme will be critiqued with an emphasis on the estimation of body composition in different ethnic groups.

3.2 Cadaver studies

Cadaver studies form the basis for all body composition assumptions and the development of methods. Cadavers were analysed either anatomically or chemically. The anatomical analysis yielded values for gross tissue weights, while chemical analysis established the elements forming the body?s components. The earliest cadaver studies were undertaken in Europe in the 19th century; cadavers of adults and children were dissected into the anatomical components (Moleschott 1859, Bischoff 1863, Volkmann 1874, Liebig 1897, Welker and Brandt 1903, Vierrordt 1906). It was not until 1945, that the chemical analysis of cadavers was undertaken. Between 1945 and 1968 the classic works of Mitchell et al., (1945), Widdowson et al., (1951) Forbes et al., (1953), (1956) and Moore et al., (1968) resulted in the chemical analysis of eight adult cadavers, the details of which are shown in table 3.1. Six male and two female cadavers were dissected and chemically analysed. They were of Caucasian origin, with the exception of one Black male, aged 25 to 60 years. The cause of death was varied as shown in table 3.1. Two died as a result of an accident and suicide as opposed to disease (Widdowson et al., 1951c and Forbes et al., 1953) and can be considered more representative of a general population. However, the two representative cadavers had low BMI?s of 15.8 and 18.8kg/m2 respectively,  with the first being underweight.
Of the eight cadavers, five were not oedematous and were used to derive %ages for fat and water content (Forbes et al., 1953, Forbes et al., 1956a, Forbes et al., 1956b, Widdowson et al., 1951b and Widdowson et al., 1951c).  Therefore the bulk of body composition assumptions are derived from the analysis of five cadavers (Widdowson et al., 1964).

Table 3. 1 Classic cadaver studies between 1945 and 1968
ReferenceSubject characteristics
Cause of deathAge (y)GenderEthnicityHeight (m)Weight (kg)Mitchell et al., (1945)35MCaucasian1.8370.6Acute heart attackWiddowson et al.,(1951a)25MCaucasian1.7971.8UremiaWiddowson et al.,(1951b)48MCaucasian-63.8Infective endocarditisWiddowson et al., (1951c)42FCaucasian1.6945.1Suicide drowningForbes et al., (1953)46MCaucasian1.6953.8Fractured skull Forbes et al., (1956a)60MCaucasian1.7273.5Heart attackForbes et al., (1956b)48MBlack1.6962.0Infective endocarditisMoore et al., (1968)67FCaucasian-43.4CarcinomaM=Male, F=Female

A large number of indirect or in vivo methods have been developed to estimate body composition on the basis of the classic cadaver analyses. In the hierarchy of methods (Heymsfield et al., (1996), the most accurate and those considered as reference measures include the estimation of body density (Db) by hydrodensitometry (HD) Benhke et al., (1972) or ADP (McCrory et al., (1995), estimation of (TBW) by isotope dilution (Pace and Rathburn 1945, Schloerb et al., 1950) and total body potassium (TBK) by isotope counting (Forbes and Hursh 1961). More recently imaging techniques such as dual energy x-ray absorpitometry (DEXA) (Mazess et al., 1990) have become equally important. These methods are used to validate the less accurate and doubly indirect methods such as BIA and anthropometry. 
To date no other whole body chemical analysis of cadavers has been conducted for the purposes of body composition research. The Brussels Cadaver Analysis Series was undertaken in order to increase the number of cadavers for the purposes of body composition, however no chemical analysis was undertaken. The details of these analyses are reported in Appendix 3.1.The following sections will discuss the characteristics of the body derived from chemical analysis of the classic eight cadavers (table 3.1).
3.3 Characteristics of body components

Traditionally, body weight was separated into two compartments, FM and FFM. These compartments were assumed to have constant characteristics, i.e. composition and density (Keys and Brozek 1953, Siri 1956, 1961). 

3.3.1 Fat mass 

The fat mass compartment is an anhydrous compartment composed mostly of lipids. Lipids are a group of chemical compounds that are insoluble in water and soluble in organic solvents such as diethyl ether. Fat is a type of lipid and in the adult human approximately 90% of total body lipid is fat (Wang et al., 1992). The FM compartment is assumed to have a density of 0.9007g/cm3 (Fidanza et al., 1953). Table 3.2 shows the composition of FM in four of the classic cadaver studies.

Table 3. 2 Composition of fat mass
ReferenceWater (%)Fat (lipid) (%)Protein (%)Ash (%)Mitchell et al., (1945)50.142.47.10.5Forbes et al., (1953)23.071.65.90.2Forbes et al., (1956a)16.878.36.80.9Forbes et al., (1956b)83.94.212.80.8
There are wide variations in the size of water  (range 16.8-83.9%) and lipid (4.2-78.3%) compartments however the protein and ash compartments are more comparable between the cadavers.
In a separate analysis Widdowson and Dickerson (1964) reported great variability in the size of the compartment between individuals. In the eight classic cadavers, FM ranged from 4.3 to 27%. This variability in FM is a consequence of emaciated and oedematous nature of some of the cadavers analysed.  Consequently, the lipid content has a high water content and low fat content and therefore are not representative of healthy living humans. Lipid, water and protein are also referred to as adipose tissue (Wang et al., 1992). The fatty acid composition of adipose tissue has been shown to vary by the kind of dietary fat consumed (Hegsted et al., 1962).
Fat mass is the component of primary interest in body composition research because of its relationship to the development of disease. However, because several factors influence its size, between individuals, it is not the basis of body composition measurement. Fat free mass is considered the most stable compartment and is discussed in section 3.3.2
3.3.2 Fat free mass

Fat free mass, refers to all non-fat tissues of the body, including bone and muscle. Its chemical constituents include nitrogen, sodium, potassium, chlorine, calcium, phosphorous, magnesium, iron, copper, zinc, boron, and cobalt. However, other trace elements such as beryllium have been identified by some authors (Widdowson and Dickerson 1964) Table 3.3 shows the properties of FFM.

Table 3. 3 Properties of fat free mass
ComponentDensity (g/ml)FFM (%)Water0.993773.8Protein1.3419.4Mineral3.0386.8          Osseous2.9825.6         Non-Osseous3.3171.2FFM1.10100Adapted from Roche et al., (1996)
The composition of FFM is assumed to be constant between groups. Furthermore, the densities of its components, water, protein and mineral are also assumed to be constant as shown in table 3.3. The constants are interdependent i.e. a constant density of FFM (DFFM) can only be assumed if hydration, mineralisation and protein are constant. 

3.3.2.1 Hydration of fat free mass

The constant hydration of FFM was established by Pace and Rathburn (1945) who chemically analysed 50 guinea pigs and found the hydration of FFM to be constant at 72.4 ( 2.1%. Other studies in mammals including humans have found remarkably consistent hydration across different species and support a hydration constant of approximately 73% with range of between 70 - 76% (Sheng and Huggins 1979, Wang et al., 1999). In addition, the classic cadavers shown in table 3.1 supported a hydration constant of 73.7(3.6% with a range between 68.4 - 80.8%. 
Hydration of FFM varies in vivo particularly with age and level of adiposity.  Older adults and young children have been reported to have a higher hydration fraction in FFM than younger adults (Schoeller 1989, Hewitt et al., 1993).  In children the hydration fraction is initially high compared with younger adults, however this has been shown to reduce with growth and stabilise to approximately 73% at maturity. The effect of ageing on FFM hydration is inconclusive (Bossingham et al., 2005).  
Currently available data on various ethnic groups do not indicate that the hydration of FFM differs with ethnicity (Wang et al., 1999, Wagner and Heyward 2000, Deurenberg et al., 2001).


3.3.2.2 Protein content of fat free mass

Protein content can be assessed by neutron activation analysis through the use of nitrogen isotopes (Cohn and Dombrowski 1971). However, its use is limited as there are very few neutron activation analysers. The ratio of protein to mineral is assumed to be constant (Siri 1956, 1961). However, ethnic differences in this ratio have been reported. This will be discussed in section 3.4.2.

3.3.2.3 Mineralisation of fat free mass

The mineral fraction of FFM of 6.8% assumed by cadaver analysis is often inferred from measures of bone mineral density (BMD) or bone mineral content (BMC) (Evans 2001). In vitro analysis of skeletons showed a higher bone mineral density and bone mineral content in male skeletons compared with female skeletons and in AAs skeletons compared with Caucasian skeletons (Merz et al., 1956, Seale 1959, Wagner and Heyward 2000). These findings have also been confirmed in vivo (Cohn et al., 1971, Cohn et al., 1977, Cotes and Adams 1993).  Dual energy x-ray absorptiometry is considered a gold standard in vivo measure of bone mineral content and bone mineral density (Mazess et al., 1990). However, DEXA measurements have not been verified by human cadaver analyses, although several studies have used animal models (Jebb et al., 1995). In addition DEXA measurements performed on the same subjects but using different manufacturers? instruments, showed substantial differences in the individual?s estimates for bone mineral content and density (Prior et al., 1997, Ellis 2001).
The variation in bone mineral content and density does not account for all variability in mineralisation of FFM (Evans 2001). In the classic study, Moulton (1923) summarised the chemical analysis of nine mammal species including rats, guinea pigs, rabbits, cats, dogs, pigs, cattle and humans. It was concluded that at birth mammalian skeletons showed a higher hydration of FFM and low protein and mineralisation. With growth, the hydration of FFM declines while concentration of protein and minerals increases. Age related changes in potassium and phosphorous content have been observed (He et al., 2003). 
The potassium content of the body among African-Americans was found to be higher than in Caucasians leading to biased values for muscle mass, body cell mass or FFM when the normal values for calculating body composition from 40K are applied (Wagner and Heyward 2000, Deurenberg and Deurenberg-Yap 2003). It has been noted that even within a homogenous population there was some variability in the chemical and physical constancy of FFM between individuals (Fuller et al., 1992). 

3.4 Body composition models

3.4.1 The two-compartment model of body composition

3.4.1.1 Assumptions

The two-compartment (2-C) model assumes body weight is divided into two distinct compartments, FM and FFM both with constant densities of 1.100g/cm3 and 0.9007g/cm3 respectively (Fidanza et al., 1953, Keys and Brozek 1953, Siri 1956, 1961). Therefore
Wt= FM + FFM
and
Wt/Db = FM/DFM +FFM/DFFM
Where Wt is body weight, Db is density, FM is fat mass, FFM is fat free mass, DFM  is density of fat mass and DFFM  is density of fat free mass. Solving for the two equations,
FM/ BW= DFM x (-DFFM Db)/ Db x (DFFM - DFM)
Inserting the values of DFM of 0.900g/cm3 and DFFM of 1.100g/cm3, the fraction of body weight as FM can be estimated based on the measurement of body density. This results in the commonly used prediction equation of Siri (1961) as follows 
FM/BW= 4.95/Db-4.50 and
FM= 4.95 x BW/Db ? 4.50 x BW
An equally popular 2-C densitometry equation is the equation of Brozek et al., (1953) which is as follows
FM=4.57/Db-4.142)*100
Although similar to the equation of Siri (1956, 1961), Brozek et al.,  (1953) assumed a DFM of 0.9007g/cm3 at a temperature of 37oC instead of 36oC. 
The alternative two compartmental model is the hydrometry model developed by Pace and Rathburn (1945)
%BF=[wt- (TBW/0.732)/wt]*100
Where wt=body weight, TBW=total body water
The model assumes a constant hydration of FFM of 0.732 

3.4.1.2 Limitations and sources of error in the two-compartment model

Siri (1961) noted that the natural variability in the components and densities of FFM results in systematic errors for the estimation of %BF using the 2-C model.  The main sources of the variability were attributed to the hydration fraction of FFM. Water is the largest component in FFM and also the most variable. Consequently, fluctuations in TBW result in changes in DFFM and therefore influence the final estimation of %BF. A theoretical error of 3-4% was calculated for the estimation of %BF from 2-C prediction equations Lohman (1981). 

3.4.1.3 Ethnic considerations

The assumption of a constant density of FFM has been challenged. Estimation of %BF using the 2-C model results in systematic errors in people whose density of FFM differs from 1.100g/cm3.  It has been shown that the density of FFM in men of African origin is greater than that assumed by Siri (1956, 1961) (Schutte et al., 1984, Wagner and Heyward 2001). In addition women of African origin have also been shown to have a greater DFFM compared with Caucasian women. Chapter five of this thesis compares Siri ?s (1961) classic 2-C model with ethnic specific 2-C models that assume a different higher DFFM. 

3.4.2 Three-compartment models

The three-compartment (3-C) model controls for the biological variability of hydration of FFM. It was noted that the largest source of error in the 2-C model was the variability in hydration of FFM. Therefore, the 3-C model improves on the 2-C model by adding a separate measure of TBW to the density estimate (Siri 1956, 1961).  The 3-C model assumes that the ratio of protein to mineral is constant and that this ratio is not altered by abnormal hydration. Most importantly it is assumed that the estimation of %BF is not strongly affected by fluctuation or uncertainty in the mineral to protein ratio
The resulting equation is
1/Db=FM/DFM+TBW/DTBW+(/D(

Where FM=fat mass, DFM= density of fat mass, TBW=total body water, DTBW=density of total body water, (= combined protein and mineral compartment of 0.35, D(= density of alpha of 1.565g/cm3. Substituting the values for densities of water from cadaver analysis (0.9933g/cm3) and alpha, the final equation is
%BF= 2.118/Db-0.78*(TBW/wt)-1.354 (Siri 1956, 1961)
Where Db =body density, TBW=total body water, Wt=body weight

Although reliant on fewer assumptions than the 2-C model, the introduction of another measurement technique may result in a cumulative error which could influence the final estimation of FFM (Withers et al., 1998). The three-compartment model is discussed further in chapter five of this thesis.   


3.4.3 The four-compartment model

The four-compartment (4-C) model divides the FFM into TBW, protein and bone mineral content.  In these models, the compartments are measured individually e.g. FM can be derived by underwater weighing, bone mineral content by DEXA. Four- compartment model equations are used widely to validate other methods; they are assumed to be a credible reference model as they combine the accurate estimation of the variable compartments of FFM, i.e. TBW, protein and mineral (Withers et al., 1998). 
However, for both the 3-C and 4-C models, the measurement errors from the different methods measuring the individual compartments are cumulative and impact directly on the final measurement of body fat. In addition these methods require cumbersome and expensive equipment, which limit their use in some settings (Heymsfield et al., 1990).


3.4.4 Multi- compartmental models

Wang et al., (1992) developed a five level model of body composition to categorise the body?s components. The five levels are atomic, molecular, cellular, tissue and whole body (Figure 3.1). 

N, P, Ca, K, Na, ClGlycogenExtracellular solidsAdipose tissue


Whole
Body
HydrogenProtein
Extracellular fluid
Skeletal Carbon

Lipid Visceral organs and residualCell massWaterMuscular tissueOxygenLevel five 
(Whole body)Level four 
(Tissue)Level three 
(Cellular)Level two 
(Molecular)Level one 
(Atomic)N=nitrogen, P= phosphorous, Ca= calcium, K= potassium, Na= sodium, Cl= chlorine
Figure 3. 1 The five level body composition model developed by Wang et al., (1992)


This model assumes a constant composition of the body weight and the mass of various components on different levels is relatively constant. An important implication of the constancy is that there are stable proportions of the different components on the same level. For example, on the molecular level the average ratio of TBW content to FFM is relatively constant in healthy subjects. Following the development of the five level model by Wang et al., (1992) several multi-compartment models have been developed. These models bring together components from different levels of the Wang et al., (1999) five level model. Examples of multi-compartment models are shown in table 3.4. The multi-compartment models are an improvement in terms of the widely used two-compartment model. However, because of the requirement for a number of methods to measure individual compartments of FFM, they are not widely used.  

Table 3. 4 Multi-compartment models of body composition
Atomic11-component modelBW=O + C + H + N + Ca + P + K + S + Na + Cl + MgMolecular6-component modelBW= F + A + Pro+ MS + Mo + G4-component modelBW = F + A + Pro+ M3 component modelBW= F + solids3 component model BW=F + A + residual2 component modelBW=F + FFMCellularBW=CM + ECF + ECSBW=F + BCM + ECF + ECSTissue BW=AT + SM + bone + other tissuesAdapted from Roche et al., 1996 O= oxygen, C= carbon, H= hydrogen, N= nitrogen, Ca= calcium, P= phosphorous, K= potassium, S= sulphur, Na= sodium, Cl= chlorine, Mg= magnesium, BW=body weight, F=Fat, A=Water, Pro=protein, Ms= soft tissue mineral, Mo=bone mineral, G=glycogen, M=mineral, FFM=fat free mass, CM=cell mass, ECF=extracellular fluid, ECS=extracellular solids, BCM= body cell mass, AT=Adipose tissue, SM=Skeletal mass.
The following sections will discuss ethnic differences in body composition and illustrate the implications of these differences through the review of ADP, BIA and anthropometry.

3.6 In vivo body composition methods

In vivo body composition methods, also known as indirect methods, estimate body composition based on the assumptions derived from the analysis of cadavers shown in table 3.1. Heymsfield et al., (1996) developed a system to categorise in vivo methods. They proposed that all in vivo methods could be summarised according to the equation 
C = f (Q)
Where C is an unknown component, Q is a measureable quantity and f is the mathematical function that links Q to C.  The measurable quantities used in this equation are properties and components. Component based methods measure quantities such as body weight, height, and impedance. Property based methods measure quantities such as TBW, total body nitrogen and total body carbon. Component and property based or combined methods measure one or more property and are based on the assumptions of 3-C or 4-C methods. Table 3.5 shows examples of component, property and combined component and property based methods.


Table 3. 5 In vivo body composition methods
In vivo methodsProperty based methodsComponent based methodsCombined methodsType IType IIAll Type IIType IIAnthropometry including SKF

BIADensitometry
ADP
DEXATBW
TBK3-compartmental models 
TBW and body density

4-compartmental models
TBK, TBW and BMDAdapted from Heymsfield et al., 1996 TBW=total body water, TBK=total body potassium, BMD= bone mineral density, ADP=air displacement plethysmography, DEXA= dual energy x-ray absorptiometry, SKF= skinfold thickness analysis, BIA= BIA

Heymsfield and colleagues (1996) subdivided the methods into type I and type II methods, depending on their approach to developing the mathematical function f. For type I methods a criterion or reference method e.g. densitometry or DEXA, is used to measure the component (C) of interest in a well-characterised group of subjects. The only requirement for type I methods is that a significant correlation exists between the component of interest (C) and the measurable quantity (Q). The resulting prediction equations are population specific and should be cross-validated before application (Heymsfield et al., 1996). Type II methods are formulated from what is often referred to as a model which has a theoretical or experimental basis, for example the assumption of constant density of FFM. These are also referred to as reference methods. Table 3.6 shows the categorisation of methods based on the system of Heymsfield et al., (1996).


3.6.1 Air displacement plethysmography 

Air displacement plethysmography refers to the estimation of body composition from the measurement of the volume of air (Gnaedinger et al., 1963). The technique of volume displacement or plethysmography has been used to measure body volume for centuries, however it is not until recently that a viable system for air displacement in the form of the BOD POD was developed (Life Measurement Inc, Concord, California). Fig 3.2 

Figure 3. 2 Cross sectional representation of the BOD POD

The BOD POD consists of a fibreglass shell containing two main chambers. A test chamber, approximately 450 litres, and a reference chamber, approximately 300 litres, separated by a moulded fibreglass seat, which forms a common wall between the chambers. A volume perturbing element in the form of a moving diaphragm is mounted between the two chambers. The movement of the diaphragm is precisely controlled by an electronic servo system. During operation, the oscillation of the diaphragm creates equal but opposite volume changes in the test and reference chambers, generating small pressure changes. The pressure change is roughly ( 0.5 mmHg which is comparable to the change in pressure while moving from the first floor to the second floor in an elevator.
The BOD POD estimates body composition by measuring the pressure-volume relationship between the two chambers. Under isothermal conditions, i.e. when temperature is constant, pressure and volume are inversely related (Boyles Law). However, in the BOD POD (the air acts adiabatically) the temperature is not constant. The pressure volume relationship is therefore corrected by a temperature constant (approximately 1.4 for air). The final estimate of body volume (Vb) is corrected for thoracic gas volume (TGV), which is the air in the lungs and gastrointestinal tract, and surface area artifact (SAA), which is the air near the skin. Body density is calculated using the equation of density shown in section 3.4, i.e. body weight divided by Vb. percentagebody fat is derived from body density using the 2-C model equations shown in table 3.6, which are integrated into the BOD POD?s software.


Table 3. 6 Formulae used to convert body density to percentagebody fat in the BOD POD system
NameEquationPopulationSiri (1961)%fat = (4.95/Db-4.50)*100General PopulationSchutte et al., (1984)%fat = (4.374/ Db -3.928)*100African American malesOrtiz et al., (1992)%fat = (4.83/ Db - 4.37)*100African American femalesBrozek et al., (1953)%fat = (4.57/ Db -4.142)*100Lean and obese individualsLohman et al., (1981)%fat = (C1/ Db - C2)*100Children <19 years Db= Body density and C1 and C2 are constants based on age and gender

The main source of error in the estimation of body volume and consequently body composition in ADP is the calculation of thoracic gas volume (TGV) (McCrory et al., 1998). Due to its isothermal nature, air in the lungs is 40% more compressible than the surrounding air (adiabatic) and therefore volume determined by Boyle's law is underestimated by 40%. Thus, a measurement of TGV by the ADP is incorporated into the testing procedure. 

3.6.1.1 Validity of air displacement plethysmography for estimating percentagebody fat
The main advantage of estimating body composition using ADP compared with hydrodensitometry is its ease of use. Hydrodensitometry is time consuming, labour intensive, requires heavy cumbersome equipment and for the subject to be completely submerged underwater while holding their breath. The procedure is uncomfortable and not accessible to a wide range of subject types including, the physically disabled, morbidly obese and children. Therefore ADP has a higher degree of subject compliance and acceptability (Dempster et al., 1995, McCrory et al., 1995,).
Eighteen studies have compared %BF estimated by the ADP with that estimated by hydrodensitometry (Table 3.7). Participants were adults with a BMI range from 17-47kg/m2, aged 18-86 years and predominately Caucasian with the exception of studies by Collins et al (1999), (2004), Millard-Stafford et al., (2001) and Wagner et al (2000) who had a large number of AAs in the groups. Most studies used the equation of Siri (1961) to convert body density to percentagebody fat. However, the equations of Schutte et al (1984) and Brozek et al (1953) were used in studies that included African-American participants (Biaggi et al., 1999, Levenhagen et al., 1999, Wagner et al., 2000, Utter et al., 2003, Collins et al., 2004).
Mean differences for %BF between ADP and HD ranged from -4.0 to 1.9%. Siri (1956) suggested that a %BF difference range of ?3.5% was acceptable for the agreement between two methods.  The upper limit (1.9%) for %BF fell within the acceptable range, however the lower limit (-4.0%) was outside of the limit. This suggests that there is less agreement between the methods at lower levels of %BF. Eleven of the 18 studies found no significant differences between the two methods; in six studies ADP underestimated %BF compared with hydrodensitometry and one found that ADP overestimated %BF. One study found a gender effect, where ADP underestimated body fat in females and overestimated fat in males, compared with hydrodensitometry (Biaggi et al., 1999, Fields et al., 2002).
Table 3. 7 Validity of air displacement plethysmography against hydrodensitometry in adults
Subject characteristicsRegression analysisReferenceN (M/F)AgeBMI Ethnic ModelMean difSlope r2SEESex effectRace effectReported findingsMcCrory et al., (1995)42M, 26F20-56NR47C,19H, 7A, 4AAS-0.3( 0.20.940.931.81NNRNo differences between methodsIwaoka et al., (1998)7M31-4422(4-S-4.0(3.10.780.82NRNA-No differences between methodsBiaggi et al., (1999)23M, 24F19-48NRNRSc + S-0.090.820.893.1Y1N/ANo differences between methodsCollins et al., (1999)69 M19.5(1NR32C,37AS-1.90.9060.892.2NAN%BFBP underestimated CW %BFHDLevenhagen et al., (1999)10M,10F31.1(1.825.2(0.9NRS+Sc-0.50.770.94NRYNNo differences between methods for whole population Nunez et al., (1999)228M, 44F20-8620-29.5NRS0.1NR0.90NRNN/ANo differences between methodsDewit et al., (2000)5M, 5F19-4121(1.6NRS-3.3NRNRNRNRN/ABVBP + %BFBP underestimated CW BVHD + %BFHDFields et al., (2000)267F18-5517-3467CNR0.90.960.942.3NAN/ABDBP underestimated CW BDHDWagner et al., (2000)230M19-4519-4030ASc + W1.90.930.84NRNABDBP underestimated CW BDHD 3Wells et al., (2000)2231(822(3-S-2.2(3.3NRNRNRNRN/ANo differences between methodsMillard-Stafford et al., (2001)40M, 10F25(0.8NR35C, 15AS-2.8(4.10.760.78NRNRN/A%BFBP underestimated CW %BFHDFields et al., (2001)242F19-54NR39C, 3AS0.140.900.942.3NANRNo differences between methodsYee et al., (2001)230M, 28F70-7927(4.28NRS0.8NR0.91NRNNANo difference between methodsUtter et al., (2003)66M20(2NRNRB +Sc0.33(2.340.8060.8032.121NANANo differences between methodsVescovi et al., (2002)5M, 10F25(9.2NRNRS-0.6NRNRNRNRNANo differences between methodsDemerath et al., (2002)41M, 46F18-6917.8-47109C, 10SEA, 3EA, 4AS1.6(3.60.960.863.7NNR%BFBP overestimated CF %BFHDCollins et al., (2004)64M23.9(5.3NR25C, 39S +ScN 4%BFBP  underestimated CW HW 4-C modelGinde et al., (2005)89M, 34F46.5(16.931.5(7.3NRS0.001(0.010.940.883.58NRNRNo difference between methodsM=Male, F=Female, NR= Not reported, N= No, Y=Yes, S= Siri Equation, Sc= Schutte equation, W=Wagner equation, B= Brozek equation, A= African American or Black, C= Caucasian or European American, SEA= South East Asian, EA= East Asians, AA=Asian American, H= Hispanic, CW= Compared with, BD= Body density, BP= BOD POD, HD= Hydrodensitometry, BV= Body volume, 1 %BFBP underestimated in females and overestimated in males CF %BFHD, 2 Compared body density estimates from each method 3 %BFBP overestimated CF %BFHD regardless of equation used, 4 Race did not influence estimation of %BF, ADP underestimated %BF in both groups compared with other methods
Theoretically, hydrodensitometry and ADP should give the same estimates of Db and percentagebody fat as they are based on the same principles. The differences can be attributed to the measurements of body weight and Vb in particular the correction for TGV and SAA. Hydrodensitometry accounts for 100% lung volume while  ADP accounts for only 40% of the lung volume. As a result, the overall estimation of Vb may not be comparable. Hydrodensitometry has been shown to give more accurate estimations of Db and consequently percentagebody fat (%BF) when compared with ADP (McCrory et al., 1998, Dewit et al., 2000, Fields et al., 2002). However, it has been suggested that methods based on the same assumptions should not be used to validate each other (Duerenberg et al., 1998). 
The main advantages DEXA has over HD is that it does not rely on the assumptions of densitometry, specifically that of constant density of FFM, it is a more ?subject friendly? technique as it does not require complete submersion in water (Mazess et al., 1990). Eleven studies compared %BF estimated by the ADP with DEXA (Table 3.8). 
Table 3. 8 Validity of air displacement plethysmography against dual energy x-ray absorptiometry in adults
Subject characteristicsRegression analysisBland-AltmanCommentN (M, F) AgeBMIEthnicModelMean DiffSloper2SEELimits of agreementSex effectRace effectSardinha et al., (1998)62M31-4618-34.562CS-2.6(2.7NRNRNR-2.6,7.8NANR%BFBP underestimated CF %BFDXACollins et al., (1999)20M20(1NRNRS-2.01.020.892.4NRNANR%BFBP underestimated CW %BFDXALevenhagen et al., (1999)2019-4720-36S-3.0(3.70.990.88NR-4.4, 10.4NNRGood correlation between methods-diffs?Nunez et al., (1999)28M, 44F20-8620(29.5S-0.850.910.943.53NRNNRNo differences found between methodsMiyatake et al., (1999)1628(721(3SNR0.910.883.5NRNRNRStatistical comparison of the slope was not reportedKoda et al., (2000)72140-79NRJapanesBrozek-1.3( 0.14 (F) and 1.22(0.13 (M)NR0.78-0.81NRNRYNRWagner et al., (2000)30M19-4518-40.5Sc + W-1.67NR0.862.84NRNANR%BFBP overestimated CF %BFDXAMillard-Stafford et al., (2001)40M, 10F25(0.8NRS-2.5(3.7NRNRNRNRNNR%BFBP underestimated CF %BFDXAFields et al., (2001)42F19-54NRS0.6(3.41.100.913.4-6.1, 7.2NANRWeyers et al., (2002)10M, 12F39-5125-37.8S-2.5NRNRNR-7.4, 3.2NNR%BFBP underestimated CW %BFDXA. ADP + DXA detect similar changes in %BF due to weight lossCollins et al., (2004)64 M23.9(5.3NR39A, 25CS+ScNM=Male, F=Female, NR= Not reported, N= No, Y=Yes, S= Siri Equation, Sc= Schutte equation, W=Wagner equation, B= Brozek equation, A= African American or Black, C= Caucasian or European American, SEA= South East Asian, EA= East Asians, AA=Asian American, H= Hispanic, CW= Compared with, BD= Body density, BP= BOD POD, DXA=Dual energy x-ray absorptiometry, The participants were adults BMI ranged from 17 to 40kg/m2. Most studies were conducted in young to middle aged-subjects, but two studies included adults over 55 years. Mean differences between %BF measured by ADP and DEXA varied widely. These differences were significant in three of the studies; four studies showed negative differences ranging from -2.0% to -3.0% and one study showed a positive (1.7% BF) difference. One additional study, with a substantial sample size of 721 subjects and an overall mean difference in %BF of -0.1% reported a significant gender difference being -1.3% for females and a positive mean difference of 1.2 % for males (Fields et al., 2002). 
The inconsistent trend in the estimation of percentagebody fat by DEXA and ADP may be related to the inherent limitations of DEXA as discussed in section 3.3. Moreover, DEXA is reliant on the assumption of constant hydration of FFM. This influences the ability of DEXA to predict soft tissue accurately e.g. body fat (Roubenoff et al., (1990).

3.6.1.2 Air displacement plethysmography in different ethnic groups

A limited number of studies have tested the validity of ADP in adults of different ethnic groups (McCrory et al., 1995, Nunez et al., 1999, Collins et al., 1999, Wagner and Heyward 2001, Millard-Stafford et al., 2001, Collins et al., 2004, Aleman-Mateo et al., 2007). The samples consisted mainly of African-Americans with the exception of one study in elderly Mexicans (Aleman-Mateo et al., 2007). 
The equation of Siri (1961) is the default equation used to estimate body fat using ADP. However, the equations of Schutte et al., (1984) and Ortiz et al., (1992) have been integrated into the ADP software for use in African-American men and women respectively. Schutte et al., (1984) found the density of FFM to be 1.113g/ml in a study of 15 black men. The greater density of FFM was attributed to the greater bone mineral and density in black versus white men. The equation developed by Schutte et al., (1984) corrects for % body fat by adding 3% fat to the density of FFM in lean black men. Ortiz et al., (1992) found African-American women had a higher density of FFM of 1.106g/ml in a study that compared 28 age-matched African-American and Caucasian women. Higher protein content and bone mineral content was found in African-American compared with Caucasian women. These equations have been accepted and are used widely to estimate body composition in men and women of African origin. The limitations of their use are discussed in Chapter five of this thesis. 
Ethnicity was not shown to contribute significantly to the difference in percentagebody fat between ADP and HD in two studies (Nunez et al., 1999 and McCrory et al., 1995).  While two others did not report the effect of ethnicity (Collins et al., 1999 and Millard-Stafford et al., 2001). Wagner and Heyward (2001) cross-validated the Schutte et al., (1984) equation, and two other widely used 2-C equations, Siri (1961) and Brozek et al., (1963), in a sample of 30 black men. Dual energy x-ray absorptiometry (DEXA) and a 4-C model (Friedl et al., 1992) were used as criterion measures of percentagebody fat. The correlations between the 4-C model and the 2-C equations were strong, and Bland-Altman analysis showed good agreement between the models. However the equation by Schutte et al., (1984) consistently overestimated %BF in 87% of the sample, while Siri (1961) and Brozek et al., (1963) underestimated percentagebody fat in 87% and 90% of the sample respectively. It was concluded that if the assumed density of FFM in the 2-C formulae of Brozek et al., (1963) and Siri (1961) is 1.100g/ml, which is the density of FFM derived in Caucasians, then these equations would underestimate percentagebody fat in black men. However, as the Schutte et al., (1984) formulae overestimated percentagebody fat; the assumed density of FFM for black men of 1.113g/ml must be incorrect. Wagner et al., (2001) estimated the density of FFM for their sample of black men to be 1.10570g/ml. It was proposed that because of the lack of heterogeneity in the original sample of Shutte and colleagues (1984) (young collegiate athletes) and the use of an inappropriate criterion measure (TBW estimation by isotope dilution) may have contributed to the inaccuracy of the resulting formula. However, this study lacked a Caucasian control group and therefore the effects seen could be the result of other factors, not necessarily ethnicity. 
Collins et al., (2004) in a study of the effect of ethnicity and musculoskeletal development on the accuracy of ADP found that ADP using the equation of Schutte et al., (1984) underestimated percentagebody fat in both blacks (-3.7%) and whites (-3.6%) when compared with a criterion 4-C model and the differences were independent of ethnicity. In this sample of black men density of FFM was estimated as 1.098 (0.002) g/ml which is significantly lower than that assumed by the Schutte et al., (1984). Collins et al., (2004) concluded that there is no need for the use of ethnic specific formula when estimating percentagebody fat from body density, as there is no biological basis for the use of the different formulae. The use of ethnic specific versus generalised equations will be discussed in Chapter five of this thesis.
In elderly Mexicans the density of FFM was similar to that assumed by Siri (1961) and therefore ADP was found to be a valid measure when compared with a 4-C criterion method. However, despite the findings of constant density, it was reported that this sample had a higher hydration of FFM than the assumed 0.732 (Pace and Rathburn 1945). Considering that the FFM constants are interdependent, it is unexpected that the density of FFM in equal to that assumed for a hydration of 0.732.
It remains inconclusive whether ethnicity significantly affects estimation of body composition by ADP in different ethnic groups. More studies in a wider range of groups are needed. Moreover, the equations of Schutte et al., (1984) and Ortiz et al., (1992) should not be applied indiscriminately to all populations of African origin when ADP estimates body fat.


3.6.2 Bioelectric Impedance Analysis

In bioelectric impedance analysis, a low level electric current, typically 50kHz, is passed through the body and the current impedance or resistance to flow is measured. Biological tissues such as FFM, extra and intracellular fluids are highly conductive, while FM and air filled spaces, such as the lungs, are not (Kushner and Schoeller 1986). 
Nyboer (1970) first related electrical impedance to biological function and studied the arterial pulse waveforms and pulsatile blood flow to organs using electrical impedance plethysmography. However, it was the pioneering work of Thamasset in 1962 and 1963, who first conducted studies to estimate TBW from electrical impedance using subcutaneously inserted needles. Hoffer et al., (1969) defined the relationship between TBW and total body fat demonstrating a high level of correlation (r=0.92).
The main assumptions underpinning the use of bioelectric impedance are that the body can be modelled as a cylindrical conductor with its length proportional to its height (ht) and inversely proportional to its cross-sectional area (Hoffer et al., 1969, Lukasaki et al., 1985, Kushner and Scholler 1986).
However, the body can be viewed as not one but five cylinders, comprising the arms, legs and torso and its conductivity is not constant (Kyle et al., 2004). Furthermore, arms and legs account for 8% and 34% of total body mass respectively, but account for 47% and 50% of the total resistance measured (Baumgartner 1996). Conversely, the trunk may account for 46% of body mass and but have little effect on measured resistance (Baumgartner et al., 1998, Jackson et al., 2002).
Bioelectric impedance analysis is a widely used to assess TBW and body fat in field and clinical settings. Table 3.9 summaries the advantages and disadvantages of BIA. It is clear that the advantages of bioelectric impedance are overwhelmingly practical. The disadvantages are all related to the underpinning assumptions.


Table 3. 9 Advantages and disadvantages of bioelectric impedance analysis
AdvantagesDisadvantagesSafe and non-invasive
No technical skill required
Subject co-operation is minimal
Measurements are reliable and rapid
Instrument is inexpensive and portable 
Errors due to variation in hydration of FFM
Disproportionate contribution of various body segments
Standard prediction equations may not be valid in all population groups
Based on information provided by Lukasaki et al., (1985), Kushner and Schoellor (1986), Ellis (2001), Kyle et al., (2004)


3.6.2.1 Validity of BIA of estimation of body fat in different ethnic groups

The estimation of body composition from BIA is dependent on predictive equations. However, the generalised equations were developed in Caucasians, and such their applicability in non-Caucasian groups is questionable (Kyle et al., 2004). It is assumed that ethnic specific equations will a better estimate of body composition than a generalised equation in population similar to the one in which the ethnic specific equation has been developed (Kyle et al., 2004).  However, an inconsistent trend has emerged when BIA equations are applied in different ethnic groups. 
Scholler and Luke (2000) compared TBW estimated by two ethnic specific and one generalised equation in 89 Jamaican and Nigerian men and women. The ethnic specific equations developed in AA populations overestimated TBW compared with deuterium oxide. Whereas the generalised equation developed in Caucasians gave an estimate of %BF/TBW that was not significantly different from deuterium oxide. Schoeller and Luke (2000) attributed the differences between the equations to differences in body fat and not ethnicity per se. However, differences in body fat amount and have been reported between ethnic groups, for example, populations of Asian origin have less FFM compared with their Caucasian counterparts (Deurenberg et al., 2002).
Diuom et al., (2005), tested the accuracy of of 23 prediction equations developed in samples of Caucasian, AA and African men and women. TBW was determined in 36 Senegalese women using all 23 equations. There was no consistency in the under and overestimation of TBW according to the population in which the equation was developed. It was concluded that differences in lengths of limbs contribute significantly to the variability in BIA measures between groups. 

3.6.3 Anthropometry

Anthropometric measurements of the human body include weight, height, skeletal breadths, circumferences and skinfold thickness. These measurements are used to describe body size and also to produce ratios and indices. Furthermore they can be incorporated into formulae validated by other methods to predict tissue masses, body proportions desirable weight ranges and somatotype (Lohman et al., 1988). The advantages and disadvantages of anthropometry are summarised in table 3.11.


Table 3. 10 Advantages and disadvantages of anthropometric measures 
AdvantagesDisadvantagesSafe and non-invasive
Subject co-operation is minimal
Measurements are reliable and rapid
Instrument is inexpensive and portable 

Errors due to inter-individual variability in fat distribution
High technical skill
Poor precision in obese subjects
Population specific predictive equationsUlijasek and Kerr (1999),  

The following sections will briefly discuss skinfold thickness analysis. 
3.6.4.1 Skinfold thickness measurement

The measurement of skinfold thicknesses is made by grasping the skin and adjacent subcutaneous tissue between the thumb and forefinger, shaking it gently to exclude underlying muscle and pulling it away from the body just far enough to allow the jaws of the caliper to impinge on the skin which surrounds the fat layer, as shown in figure 3.3.  


Figure 3. 3 Skinfold thinkness analysis
Source: Top End Sports.com

Skinfold thickness analysis is based on the assumption that the thickness of adipose tissue reflects a constant proportion of total body fat and the sites selected for the measurement represent the average thickness of the subcutaneous adipose tissue (Lohman 1981). In addition it is assumed that subcutaneous adipose tissue makes up approximately 50% of the body, However, the evidence in support of this statement is limited (Norgan 2005).
Skinfolds are used to assess fat distribution and to estimate %BF using prediction equations such as the commonly used Durnin and Womersely (1974) equations. However ethnic specific skinfold equations have been developed for homogenous populations in order to account for the differences in the distribution of adipose tissue. 
Significant differences exist in the distribution of subcutaneous adipose tissue which may affect the application of generalised skinfold equations (Robson et al., 1971, Cronk 1982). Zillikens and Conway (1990) investigated differences in body fat distribution between AA and Caucasian women and the effect on the prediction of %BF in AAs using a generalised skinfold equation. African American females have relatively more fat than Caucasian females at the subscapula site than the triceps or thigh. In addition, AA females had a more android body fat distribution than the Caucasian females. In the AAs generalised equations of Durnin and Wormersely (1974), Jackson and Pollock (1978) and Jackson et al., (1980) correlated with %BF estimated using deuterium oxide (D20) dilution. The Jackson equations underestimated %BF while only the Durnin and Womersely (1974) equation predicted %BF accurately. 
In contrast, Irwin et al., (1998) found that the Jackson et al., (1980) sum of seven skinfolds (?7SKF) equation predic (1980) sum of seven skinfolds (?7SKF) equation predicted %BF more accurately that Durnin and Womersely (1974) in AA women. African American women have a greater of upper body fat than Caucasian women, it was assumed that the high correlation with Jackson et al., (1980) (?7SKF) represented %BF better than other equations using only three of four sites (Irwin et al., 1998). The disparity between the results of Zillikens and Conway (1990) and Irwin et al., (1998) may be related to the different reference methods u20 dilution and hydrodensitometry respectively. Although D20 dilution space is known to be 4% greater than actual TBW, %BF estimated from a 2-C model has greater errors especially in AA women because of differences in density of FFM (Brandon 1998). More recently studies in African women (Diuom et al., 2005) and Singaporean Chinese, Indians and Malays (Duerenberg et al., 2003) showed that the relationship between skinfold analysis and %BF is significantly affected by racial differences in subcutaneous fat distribution.


3.7 Conclusion

This chapter critically reviewed the development of body composition assumptions and methods. The underlying concept of constant composition and density of FFM is the cornerstone of body composition research and the development of methods. However, this concept has been challenged and it has been shown that FFM composition is not constant. Despite this, a large number of body composition methods have been developed on the basis of these assumptions. It is now accepted that ethnic differences exist in body composition, ideally new assumptions would be developed however because of the limitations associated with cadaver analysis, this is not possible.  


3.8 References

Agyemang, C., Bhopal, R., Bruijnzeels, M., (2005). Negro, Black, Black African, African Caribbean, African American or what? Labelling African origin populations in the health arena in the 21st century. Journal of Epidemiology and Community Health, 59, pp.1014-1018.
Ball, S.D., (2005). Interdevice variability in percentagefat estimates using the BOD POD. European Journal of Clinical Nutrition, 59, pp.996-1001.
Baumgartner, R, N., Chumlea, W.C., Roche, A.F., (1989). Estimation of body composition from bioelectric impedance of body segments. American Journal of Clinical Nutrition, 50, pp.221-226.
Behnke, A.R., Feen, B.G., Welham, W.C., (1942). The specific gravity of healthy men. Journal of the American Medical Association, 118, pp.492-501.
Biaggi, R.R., Vollman, M.W., Nies, M.A., (1999). Comparison of air displacement plethysmography with hydrostatic weighing and BIA for the assessment of body composition in health adults. American Journal of Clinical Nutrition, 69, pp.898-903.
Bischoff, E., (1863). Einige gewichts und trockenbestimmungen der organe des menschilichen korpers, Zeit  Rationelle Medicin, 3, pp.75-85.
Boddy, K., King, P.C., Womersely, J., Durnin, J.V.G.A. (1973). Body potassium and fat free mass. Clinical Science, 44, pp.622-625.
Boddy, K., Holloway, I., Elliot, A., (1973). A simple facilitation for total body in vivo activation analysis. International Journal of Radiation Applications and Instrumentation, 24, pp.428-430.
Bossingham, M.J., Carnell, N.S., Campbell, W.W., (2005). Water balance, hydration status and fat free mass hydration in younger and older adults. American Journal of Clinical Nutrition, 81, pp.1342-1350.
Brandon, L., (1998). Comparison of existing skinfold equations for estimating body fat in African American and white women. American Journal of Clinical Nutrition, 67, pp.1155-1161.
Brozek, J., Grande, F., Anderson, J.T., Keys, A.,(1963). Densitometirc analysis of body composition: revision of some quantitative assumptions. Annals of New York Academy of Science, 110, pp.113-140.
Chen, Y. M., Ho, S.C., Lam, S.S.H., Chan, S.S.G, (2006). Validity of body mass index and WC in the classification of obesity as compared to % body fat in Chinese middle-aged women. International Journal of Obesity, (30), pp.918-925.
Clarys, J.P., Provyn, S., Marfell-Jones, M.J., (2005). Cadaver studies and their impact on the understanding of human adiposity. Ergonomics, 48, pp.1445-1461.
Clarys, J.P., Martin, A.D., Marfell-Jones, M.J., Janssens, V., Caboor, D., Drinkwater, D.T., (1999). Human body composition: A review of adult dissection data. American Journal of Human Biology, 11, pp.167-174.
Clarys, J.P., Martin, A.D., Drinkwater, D.T., (1985). Gross tissue weights in the human body by cadaver dissection. Human Biology, 56, pp.459-473.
Clarys, J.P., Marfell-Jones, M.J., (1986). Anthropometric prediction of component tissue masses in the minor limb segments of the human body. Human Biology, 58, pp.761-796.
Cohn, S.H., Abesamis, C., Zanzi, I., Aloia, J.F., Yasumura, S., Ellis, K.J., (1977). Body elemental composition: comparison between black and white adults. American Journal of Physiology: Endocrinology, Metabolism and Gastrointestinal Physiology, 232, pp.E419-E422.
Cohn, S.H., and Palmer, H.E., (1974). Recent advances in whole body counting: a review. Journal of Nuclear Biology and Medicine, 1, pp.155-165.
Cohn, S.H., and Dombrowski, C.S., (1971). Measurement of total body calcium, sodium, choline, nitrogen and phosphorous in man by in vivo neutron activation analysis. Journal of Nuclear Medicine, 12, pp.499-505.
Collins, M.A., Millard-Stafford, M.L., Sapring, P.B., (1999). Evaluation of the BOD POD for assessing body fat in collegiate football players. Medicine and Science in Sport and Exercise, 31, pp.1350-1356.
Collins, A.L., and McCarthy, H.D., (2003). Evaluation of factors determining the precision of body composition measurements by air displacement plethysmography. European Journal of Clinical Nutrition, 57, pp.770-776.
Collins, A.L., Saunders, S., McCarthy, H.D., Williams, J.E., Fuller, N.J., (2004). Within and between laboratory precision in the measurement of body volume using air displacement plethysmography and its effect on human body composition assessment. International Journal of Obesity, 28, pp.80-90.
Collins, M. A., Millard-Stafford, M.L., Evans, E.M., Snow, T. K., Cureton, K.J., Rosskopf, L.B., (2004). Effect of race and musculoskeletal development on the accuracy of air displacement plethysmography. Medicine and Science in Sport and Exercise, 36, pp.1070-1077.
Cooper, R.S., (2003). Race, genes and health-new wine in old bottles? International Journal of Epidemiology, 32, pp.23-25.
Cote, K., Adams, W.C., (1993). Effect of bone density on body composition estimates in young adult black and white women. Medicine and Science in Sport and Exercise, 25, pp.290-296.
Cronk, C.E., and Roche, A.F., (1982). Race-and sex-specific reference data for triceps and Subscapular skinfolds and weight/ stature2. American Journal of Clinical Nutrition, 35, pp.347-354.
Dempster, P., Aitkens, S., (1995). A new air displacement method for the determination of human body composition. Medicine and Science in Sports and Exercise, 27, pp.1692-1697.
Deurenberg, P., (1996). Limitations of the bioelectric impedance method for the assessment of body fat n severe obesity. American Journal of Clinical Nutrition, 64, pp.449S-452S.
Deurenberg, P., Deurenberg-Yap, M., Schouten, F.J.M., (2002). Validity of total and segmental impedance measurements for prediction of body composition across ethnic groups. European Journal of Clinical Nutrition, 56 pp.214-220.
Deurenberg, P., Deurenberg-Yap, M., & Guricci, S., (2002). Asians are different from Caucasians and from each other in their body mass index/body fat percentagerelationship. Obesity Reviews, 3, pp.141-146.
Deurenberg, P., Deurenberg-Yap, M., Schouten, F.J.M., (2002). Validity of total and segmental impedance for prediction of body composition across ethnic populations. European Journal of Clinical Nutrition, 56, pp.214-220.
Deurenberg, P., Yap, M., and W A van Staveren W.A., (1998). Body mass index and % body fat: a meta-analysis among ethnic groups. International Journal of Obesity, 22, pp.1164-1171.
Dewit, O., Fuller, N.J., Fewtrell, M.S., Elia, M., Wells, J.C.K. (2000). Whole body air displacement plethysmography compared with hydrodenstiometry for body composition analysis. Archives of Diseases in Childhood, 82, pp.159- 164.
Dioum, A., Gartner, A., Delpeuch, F., Maire, B., (2005) Body composition predicted from skin folds in African Women: A cross validation study using air displacement plethysmography and a black specific equation. British Journal of Nutrition 93 pp.973-979.
Dudeja, V., Misra, A., Pandey, G., Devina, G., Kumar, G., Vikram, N.K., (2001). BMI does not accurately predict overweight in Asian Indians in Northern India. British Journal of Nutrition, 86, pp.105-112.
Ellis, K., (2000). and Human body composition: In Vivo methods. Physiological Reviews, 80, pp.649-680.
Fidanza, F., Keys, A., Anderson, J.T., (1953). Density of body fat in man and other mammals.  Journal of Applied Physiology, 6, pp.252-256.
Fields, D.A., Goran, M.I., McCrory, M.A., (2002). Body composition assessment via air displacement plethysmography in adults and children: a review. American Journal of Clinical Nutrition, 75, pp.453-467.
Forbes, G.B., Hursh, J.B., (1961). Estimation of total body fat from potassium-40 content. Science, 16, pp.1917-1918.
Forbes, R.M., Cooper, A.R., Mitchell, H.H., (1953). The composition of the adult human body as determined by chemical analysis. Journal of Biological Chemistry. 203, pp.359-366.
Forbes, R.M., Mitchell, H.H., Cooper, A.R., (1956). Further studies on the gross composition of the adult human body. Journal of Biological Chemistry, 223, pp.969-975.
Friedl, K.E., DeLuca, J.P., Marchitelli, L.J., Vogel, J.A., (1992). Reliability of body fat estimates from a four compartmental model by using density, body water and bone mineral measurements. American Journal of Clinical Nutrition,  55, pp.764-770.
Gnaedinger, R.H., Reineke, E.P., Pearson, A.M., Van Huss, W.D., Wessel, J.A., Montoye, H.J., (1963). Determination of body density by air displacement. Helium dilution and underwater weighing. Annals of the New York Academy of Sciences, 110, pp.96-108.
Gallagher, D., Visser, M., Sepulveda, D., Pierson, R.N., Harris, T., Heymsfield, S.B., (1996). How useful is body mass index for comparison of body fatness across age, sex and ethnic groups. American Journal of Epidemiology, 143, pp.228-239. 
Gartner, A., Diuom, A., Delpeych, F., Maire, B., Schutz, Y., (2004). Hand-to-hand impedencemetry to predict body composition of African women as measured by air displacement plethysmography. European Journal of Clinical Nutrition, 58, pp.523-531.
Garrow, J.S., & Webster J., (1985). Quetelet's index (W/H2) as a measure of fatness. International Journal of Obesity, 9, pp.147-153.
Hegsted, D.M., Jack, C.W., Stare, F.J., (1962). The composition of human adipose tissue from several parts of the world. American Journal of Clinical Nutrition, 10, pp.11-18.
Heymsfield, S. B., Wang, Z., Visser, M., Gallagher, D., Pierson Jr, R. N. (1996). Techniques used in the measurement of body composition: An overview with emphasis on bioelectrical impedance analysis. American Journal of Clincal Nutrition, 64, 478S-484S
Heymsfield, S.B., Wang, Z.M., Withers, R., (1996). Multicomponent molecular-level models of body composition analysis. In Roche A, Heymsfield SB, Lohman T, eds. Human body composition. Champaign, IL: Human Kinetics Publishers 

Heymsfield, S.B., Gallagher, D., Mayer, L., Beetsch, J.and Pietrobelli, A., (2007). Scaling of human body composition to stature: new insights into body mass index. American Journal of Clinical Nutrition, 86, pp.82-91.
Hewit, M.J., Going, S.B., Williams, D.P., Lohman, T.G., (1993). Hydration of the fat free body mass in children and adults: implications for body composition assessment. American Journal of Physiology- Endocrinology and Metabolism, 28, pp.E88-E95.
Hoffman, D.J., Wang, Z., Gallagher, D., Heymsfield, SB (2005). Comparison of visceral adipose tissue mass in adult African Americans and whites. Obesity Research, 13, pp.66-74
Hoffer, E.C., Meador, C.K., Simpson, D.C., (1970). A relationship between whole body impedance and total body water volume.  Annals of the New York Academy of Science, 110, pp.452-461.
Iwaoako H., Yokoyama, T., Nakayama, T., (1998). Determination of % body fat by the newly developed sulfur hexafluoride dilution method and air displacement plethysmography. Journal of Nutrition Science and Vitaminology, 44, pp.561-568.
Jackson, A.S., Stanforth, P.R. Gagnon, J., Rankinen, T., Leon, A.S., Rao, D.C., Skinner, J.S., Bouchard, C., and Wilmore, J.H., (2002). The effect of sex, age and race on estimating % body fat from body mass index: the HERITAGE Family Study. International Journal of Obesity, 26, pp.786-796.
Jackson, A.S., and Pollock, M.L., (1978). Generalised equations for predicting body density of men. British Journal of Nutrition, 40, pp.497-504.
Jackson, A.S., Pollock, M.L., and Ward, A., (1980). Generalised equations for predicting body density of women. Medicine and Science in Sport and Exercise, 12, pp.131-137.
Janssens, V., Thys, P., Clarys, J.P., Kvis, H., Chowdhury, B., Zinzen, E., Cabri, J., (1994). Post-mortem limitations of body composition analysis by computed tomography. Ergonomics, 37, pp.207-216.
Keys, A., Brozek, J (1953). Body fat in adult men. Physiological Reviews, 33, pp.245-325.
Koda, M., Tsuzuku, S., Ando, F., Niino, N., Shimokata, H., (2000). Body composition by  air displacement plethysmography in middle-aged and elderly Japanese: comparison with dual energy x-ray absorpitometry. Annals of the New York Academy of Science, 904, pp.484-488.
Kushner, R.F., Schoeller, D.A., (1986). Estimation of total body water by BIA. American Journal of Clinical Nutrition, 44, pp.417-424.
Kyle, U.G., Bosaeus. I., De Lorenzo, A.D. Deurenberg, P., Elia, M., Gomez, J.M., Heitmann, B.L., Kent-Smith, L., Melchoir, J-C., Pirlich, M., Scharfetter, H., Schols, A.M.W.J., Pichard, C., (2004). BIA- part I: review of principles and methods. Clinical Nutrition, 23, pp.1226-1243.
Levenhagen, D.K., Borel, M.J., Welch, D.C., (1999). A comparison of air displacement plethysmography with three other techniques to determine body fat in healthy adults. Journal of Parenteral and Enteral Nutrition, 23, pp.293-299.
Liebig, G., (1874). Gewitchtsbestimmungen der organe des minschen korpers. Arch Anat Physiol Wissensch Medizin, pp.96-117.
Lohman, T.G., (1981). Skinfolds and body density and their relation to body fatness. Human Biology,  53, pp.181-225.
Lukasaki, H.C., Johnson, P.E., Bolonchuk, W.W., Lykken, G.I., (1985). Assessment of fat free mass using bioelectric impedance measurements of the human body. American Journal of Clinical Nutrition, 41, pp.810-817.
Martin, A.D., Drinkwater, D.T., (1991). Variability in the measures of body fat: Assumptions or technique. Sports Medicine, 11, pp.277-288.
McCrory, M.A., Gomez, T.D., Bernauer, E.M., Mole, P.A., (1995). Evaluation of a new air displacement plethysmography for measuring human body composition. Medicine and Science in Sports and Exercise, 27, pp.1686-1691.
McCrory, M.A., Mole, P.A., Gomez, T.D., Dewey, K.G., Bernauer, E.M., (1998). Body composition by air displacement plethysmography by using predicted and measured thoracic gas volumes. Journal of Applied Physiology, 84, pp.1475-1479.
Merz, A.L., Trotter, M., Peterson, R.R., (1956). Estimation of skeleton weight in the living. American Journal of Physical Anthropology, 14, pp.589-609.
Millard-Stafford, M.L., Collins, M.A., Evans, E.M., Snow, T.K., Cureton, K.J., Rosskopf, L.B., (2001). Use of air displacement plethysmography for estimating body fat in a four-compartment model. Medicine and Science in Sports and Exercise, 33, pp.1311-1317. 
Mitchell, H.H., Hamilton, T.S., Steggerda, F.R., Bean, H.W., (1945). The chemical composition of the adult human body and its bearing on the biochemistry of growth. Journal of Biological Chemistry, 158, pp.625-637.
Mitchell, A.D., (2007). Impact of research with cattle, pigs and sheep on nutrition concepts: body composition and growth. Journal of Nutrition, 137, pp. 711-714.
Miyatake, N., Nonaka, K., Fujii, M., (1999). A new air displacement plethysmograph for the determination of Japanese body composition. Diabetes, Obesity and Metabolism, 1, pp.347-351.
Moore, F.D., Lister, J., Boyden, C.M., Ball, M.R., Sullivan, N., Dagher, F.J., (1968). The skeleton as a feature of body composition. Human Biology, 40, pp.136-188.
Moore, F. D., (1946). Determination of total body water and solids with isotopes. Science, 104, pp.157-160.

Noreen, E. E., Lemon, P.W. R. (2006). Reliability of air displacement plethysmography in a large, heterogeneous sample. Medicine & Science in Sports & Exercise 38 , pp.1505-1509.
Nyboer, J., Kreuder, M.M, Hannapel, L., (1950). Electrical impedance plethysmography: A physical and physiologic approach to peripheral vascular study. Circulation, 2, pp.811-821.
Nunez, C., Kovera, A.J., Pietrobelli, A., (1999). Body composition in children and adults by air displacement plethysmography. European Journal of Clinical Nutrition, 53, pp.382-387.
Ortiz, O., Russell, M., Daley, T.L., Baumgartner, R.N., (1992). Differences in skeletal muscle and bone mineral mass between black and white females and their relevance to estimates of body composition. American Journal of Clinical Nutrition, 55, pp.8-13.
Pace, N., Rathburn, E.N., (1945). The body water and chemically combined nitrogen content in relation to fat content. Journal of Biological Chemistry, 158, pp.685-691.
Razak, F., Anand, S.S., Shannon, H., Vuksan, V., Davis, B., Jacobs, R., Teo, K. K., McQueen, M., Yusuf, S., (2007). Defining obesity cut points in a multiethnic population. Circulation, 115, pp.2111-2118. 
Robson, J.R.K., Bazin, M., Soderstrom, R., (1971). Ethnic differences in skinfold thickness. American Journal of Clinical Nutrition, 24, pp.864-868.
Roche, A.F. (1996) Densitometry. In Roche, A.F., Heymsfield, S.B., Lohman, T.G., (1996). Human Body Composition Champaign: Human Kinetics
Rush,E.C Goedecke, J.H.,Jennings, C., Micklesfield, L.,Dugas, L., Lambert, E.V., and Plank, L.D.(2007). Body mass index, fat and muscle differences in urban women of five ethnicities from two countries. International Journal of Obesity, 31, pp.1232-1239.
Sardinha, L.B., Lohman, T.G., Teixeira, P., Guedes, D.P., Going, S.B., (1998). Comparison of air displacement plethysmography with dual energy x-ray absorptiometry and three field methods for estimating body composition in middle-aged  men. American Journal of Clinical Nutrition, 68, pp.786-793. 
Seale, R.U., (1959). The weight of the dry fat free skeleton of American white and Negroes. American Journal of Physical Anthropology, 17, pp.37-48.
Schloerb, P.R., Friis-Hansen, B.J., Edelman, I.S., Solomon, A.K., Moore, F.D., (1950). The measurement of total body water in the human subject by deuterium oxide dilution ; with a consideration of the dynamics of deuterium distribution. Journal of Clinical Investigation, 29, pp.1296?1310.
Schoeller, D.A., van Santen, E., Peterson, D.W., Dietz, W., Jaspan, J., Klein, P.D., (1980). Total body water measurement in humans with 18O and 2H labeled water. American Journal of Clinical Nutrition, 33, pp.2686-2693.
Schoeller, D.A., (1989). Changes in total body water with age. American Journal of Clinical Nutrition, 50, pp.1176-1181.
Schoeller, D.A., and Luke, A., (2000). BIA prediction equations differ between African-Americans and Caucasians, but it is not clear why? Annals of the New York Academy of Sciences, 904, pp.225-226.
Schutte, J.E., Townsend, E.J., Hugg, J., Shoup, R.F., Malina, R.N., Blomquist, G.C.,  (1984). Density of lean mass is greater in blacks than in whites. Journal of Applied Physiology, 56, pp.1647-1649.
Sheng, H.P., Huggins, R.A., (1979). A review of body composition studies with emphasis on total body water and fat. American Journal of Clinical Nutrition, 32, pp.630-647.
Siri, W.E., (1956). The gross composition of the body. In: Advances in Bioelogical and Medical Physics edited by Tobias, C.A., and Lawrence, J.H., New York: Academic
Siri, W.E. (1961). Body composition from fluid spaces and density: analysis of methods. In: Techniques for measuring body composition, edited by J. Brozek, and A. Henschel. Washington, DC: Natl.Acad.Sciences/ Natl.Res. Council, pp.223-224.
Thomas, L.W., (1962). The chemical analysis of adipose tissue of man and mice. Experimental Physiology, 47, pp.179-188.
Ulijasek, S.J., Kerr, D.A., (1999). Anthropometric measurement error and the assessment of nutritional status. British Journal of Nutrition, 82, pp.165-177.
Wang, Z., Pierson Jnr, R.N., Heymsfield, S.B., (1992). The five-level model: a new approach to organizing body composition research. American Journal of Clinical Nutrition, 56, pp.19-28.
Wang, Z., Deurenberg, P., Wang, W., Pietrobelli, A., Baumgartner, R.N., Heysmfield, S., (1999). Hydration of fat free mass: a review and critique of a classic body composition constant. American Journal of Clinical Nutrition, 69, pp.833-841.
Wagner, D.R., Heyward, V.H., Gibson, A.L., (2001). Validation of air displacement plethysmography for assessing body composition. Medicine and Science in Sports and Exercise, 32, pp.1339-1344.
Wagner, D.R., Heyward V.H., (2000). Measures of body composition in blacks and whites: a comparative review. American Journal of Clinical Nutrition, 6, pp.1392-1402.
Wagner, D.R., Heyward, V.H., (2001). Validity of two-compartmental models for estimating body fat of black men. Journal of Applied Physiology, 90, pp.649-656.
Welker, H., and Brandt, A., (1903). Gewichteswerthe der korperorgane be idem menschen und den thieren. Archieve fur Anthropologie, 28, pp.1-89.
Withers, R.T., LaForgia, J., Pillans, R.K., (1998). Comparisons of two- three- and four-compartment models of body composition analysis in men and women. Journal of Applied Physiology, 85, pp.238-245. 
Widdowson, E.M., McCance, R.A., Srapy, C.M., (1951). The chemical composition of the human body. Clinical Science, 10, pp.113-125.
Wong, W.W., Cochran, W.J., Klish, W.J., O?Brian-Smith, E., Lee, L.S., Klein, P.D., (1988). In vivo isotope-fractionation factors and the measurement of deuterium- and oxygen-18-dilution spaces from plasma, urine, saliva, respiratory water vapour, and carbon dioxide. American Journal of Clinical Nutrition, 47, pp.1-6.
World Health Organization: (2006). Global database on Body Mass Index: An Interactive Tool for Monitoring Nutritional Transition                                         [online] available at http://www.who.int/bmi/index.jsp [Accessed January 2008]
Zillikens, M.C., and Conway, J.M., (1990). Anthropometry in blacks: application of generalised skinfold equations and differences in fat patterning between blacks and whites. American Journal of Clinical Nutrition, 52, pp.45-51
BODY MASS INDEX AND ADIPOSITY IN THREE ETHNIC GROUPS

4.1 Introduction

The relationship between BMI and adiposity is known to be age and gender specific (Deurenberg et al., 1991, Gallagher et al., 1996). However, the effect of ethnicity on this relationship is controversial. The use of BMI as a proxy for body fat was validated in a study that compared the accuracy of BMI in predicting body fat with an average estimated by three laboratory methods including densitometry, TBW estimation and total body potassium counting (TBK) (Garrow and Webster 1985). Each method and BMI were compared with an average of the three laboratory estimates. For women the standard deveiation (SD) was 3.4, 3.3, 3.5 and 4.2kg for fat estimated from densitometry, TBW, TBK and BMI respectively. For men the SD was 5.8, 3.8. 4.1 and 5.8 kg respectively. Densitometry, TBW and TBK contributed to the average against which the methods were compared, therefore their individual standard deviations were smaller than the SD for BMI, which did not contribute to the average. The SD for BMI was not significantly different from the other methods, therefore BMI was considered a valid proxy for adiposity. However, the generalisability these findings to other populations is questionable, given that the sample was predominately Caucasian with one Afro-Caribbean woman and one Asian man (pers comm.). 
It has been shown that at the same BMI, Caucasians differ in the BMI and %BF relationship compared with Asians (Wang et al., 1994), Polynesians (Swinburn et al., 1996), Indonesians (Gurrici et al., 1998), Asian Indians (Dudeja et al., 2001), Hispanic Americans (Fernandez et al., 2003), Taiwanese (Chang et al., 2003) and Koreans (Chung et al., 2005). However, the direction of these differences was inconsistent. The BMI and %BF relationship has also been found to differ between people of same racial background but of different ethnicity. At any given BMI, Nigerians had a lower %BF compared with Jamaicans and AAs of similar age and gender (Luke et al., 1997). Singaporean Chinese had the lowest %BF compared with Malays and Indians who had the highest %BF at comparable BMIs (Deurenberg et al., 2002, Deurenberg-Yap et al., 2002) and BMI did not represent the same %BF in American compared with European Caucasians; Europeans had a 3.8 % higher body fat at the same BMI (Deurenberg et al., 1998). 
In these studies, differences in the BMI and %BF relationship have been attributed to differences in physical activity levels between populations; the more active having greater muscle mass and less body fat their than non-active counterparts (Luke et al., 1997). However, this is not ethnic specific and may be applicable to any group of people. Secondly, differences in body build, i.e. taller people with longer limbs had a lower BMI than those with shorter legs. Gallagher et al., (1996) found that differences in lower limb length did not affect the relationship between BMI and %BF in AA and Caucasian men and women. Finally, there may be differences in muscularity, amount and distribution of body fat. For example, people of Asian origin have been reported to have less skeletal muscle mass and a greater amount of central fat compared with those of Caucasian origin.  Body mass index, is unable to detect differences in muscularity and adiposity (Deurenberg et al., 2000, Deurenberg-Yap et al., 2002, Rush et al., 2007). 
Not all studies have found differences in the BMI and %BF relationship in different ethnic groups. Gallagher et al., (1996) found that BMI reflected the same relative %BF in AA compared with Caucasian men and women. Only age and gender influenced the ability of BMI to predict %BF; BMI, age and gender explained 67% of the variance in %BF in AA and Caucasian men and women. No significant variance was explained by the addition of ethnicity to the model. Deurenberg et al., (1997) studied the BMI-%BF relationship in Singaporean Chinese and Dutch Caucasians. Body mass index represented similar mean (SD) %BF in Chinese (25.9 (8.4)%) and Caucasians (26.5 (9.6)%). In addition ethnicity did not add significantly to the variance in %BF. More recently, Jackson et al., (2002) investigated the effect of age, gender and ethnicity on the BMI and %BF relationship in 655 AA and Caucasian men and women. The effect of ethnicity was found to be small and not significant in males. In females BMI underestimated %BF of AA by 2.0%, and overestimated %BF in Caucasian females by 0.8%. As BMI increased the ethnic difference became progressively smaller. Therefore, it was concluded that ethnicity had a small and non-significant effect on the BMI-%BF relationship.
The inconsistency in the findings of the BMI and %BF relationship in different ethnic groups has implications for the identification of excessive adiposity and its co-morbidities. Body mass index may not be an appropriate proxy for adiposity in all ethnic groups. Therefore, it is important to explore further the relationship between BMI and %BF in different ethnic groups. Ethnic differences in body size may influence the relation between BMI and %BF and thus when comparing the BMI and %BF relationship, an attempt to control for the effects of these differences must be made. 
Thus the aim of this study was to compare the relationship between BMI and %BF estimated by ADP in Afro-Caribbean, Asian and Caucasian men and women. The participants were matched for age, gender and BMI to minimise the effects introduced by these variables.

4.2 Methods 

4.2.1 Subjects

Subjects were recruited from the staff and student population at a University in South East England, through advertisements on staff and student media, (students? newsletter, staff and student intranet), posters, canvassing and through personal contacts. Three hundred and twenty-two (322) subjects were recruited and tested in this study. One hundred and six were Afro-Caribbean, 41 were Asian, 165 were Caucasian and 10 were mixed race or of another ethnic group and have been excluded from these analyses. Subjects were individually matched for age (? 3years), gender and BMI (? 2 kg/m2).  After matching it was possible to include 26 Afro-Caribbean, 26 Asian and 26 Caucasian men and women in the study.

4.2.1.1 Ethical Approval
The Faculty Ethics Committee gave approval. All subjects gave written informed consent.  
4.2.1.2 Selection Criteria

Healthy adults aged 18-65 years were recruited to the study. The following people were excluded
Pregnant women and those who are less than nine months post partum or were breast-feeding.
People with self-reported conditions that affect fluid balance, body composition, disturbances in temperature and metabolic rate or bone mass.  
People with medical implants such as pacemakers etc.
People known to be claustrophobic.
4.2.1.3 Ethnic Origin

Ethnicity was self-reported, subjects self categorised according to the Office for National Statistics categories (2001) (appendix 4.1).


4.2.2 Outcome Measures

All measurements were conducted on one occasion in a dedicated cardiovascular and nutrition research laboratory within the university campus. Subjects were asked to report to the laboratory not less than one hour after eating, having emptied their bladder and refrained from exercise in the previous 24 hours. 
4.2.2.1 Anthropometry

All anthropometric measurements were taken according to the guidelines set by the International Society for the Advancement of Kinanthropometry (ISAK) manual (Marfell-Jones et al., (2006)
4.2.2.1.1 Height
Height was measured (to the nearest completed 10 mm) in the Frankfort plane using a freestanding stadiometer (Invicta Products, Leicester). 
4.2.2.1.2 Weight
Weight was measured (to the nearest completed 10 g) using the electronic scale (Tanita Corporation, Japan) provided with the BOD POD S/T system. Subjects were barefoot and wore tight fitting swimwear. 

4.2.2.1.3 Waist Circumference
Waist circumference was taken with the subject standing arms at the sides, feet together and abdomen relaxed, at the narrowest part of the torso above the umbilicus and below xiphoid process.
4.2.2.1.4 Hip circumference
Hip circumference was taken with subject standing and feet together; a horizontal measure was taken at the maximal circumference of the buttocks.


4.2.2.3 Air displacement plethysmography

Air displacement plethysmography was used to estimate %BF from the body density in the BOD POD body composition tracking system (Life Measure Inc). It consisted of a fibreglass shell containing two main chambers. A moulded fibreglass seat formed a common wall separating the front (test) and rear (reference) chambers. The subject used the front chamber during testing while the rear chamber was used as a reference volume. Subjects entered the BOD POD through a hinged door at the front of the test chamber. Two electromagnets and an air gasket sealed the door during measurement of volume. Body density was estimated by the measurement of the volume of air displaced by the subject while seated in the BOD POD. Volume was corrected for thoracic gas volume (TGV), the inside the lungs, and surface area artefact (SAA), the air near the skin. Volume was converted to body density using the equation 
Db= Body weight/ volume
Density was converted to %BF using the equation of Siri (1961), which is the BOD POD?s default equation


4.2.3 Statistical Analysis

All statistical analysis was carried out using SPSS version 15.0 (SPSS Inc., Chicago, IL, USA). Normality was assessed using the Kolmorgorov-Sminov test. All data met the assumptions for normality. One way analysis of variance (ANOVA) was used to show differences in age, weight, height, BMI, WCs and waist to hip ratio (expressed as mean (SD)) in the unmatched and matched groups. In the matched group, Pearson?s correlation coefficient was used to assess the relationship between BMI and %BF. Multiple linear regression analysis was used to investigate the possible influence of age, gender and ethnicity on the BMI and %BF relationship. The dummy variables for ethnicity were Ethnicity1 and Ethnicity2. For Caucasians Ethnicity1 and Ethnicity2 =0, for Afro-Caribbeans, Ethnicity1 =1 and Ethncity2 =0, and for Asian Ethnicity1 =0 and Ethnicity2=1. Dummy variables for gender were males=0 and females=1. 

The independent variables were evaluated for collinearity. High correlations between independent variables of 0.7 or more indicates collinearity and the model would be biased (Field 2005). In this case one of the variables would have been eliminated or a composite variable formed. To assess the contribution of each independent variable?s contribution of %BF, standardized beta coefficients were compared. The highest coefficient contributes the largest amount to the model.


4.3 Results

4.3.1 Unmatched group comparisons

Table 4.1 shows age, height, weight, BMI and %BF comparisons for the unmatched group. 
 Table 4. 1 Unmatched subject characteristics (mean (SD))
Afro- CaribbeanAsianCaucasianNumber (M, F)106 (26, 80)41 (21, 20)165 (66, 99)Age (years)30.1  (10.2)30.1 (10.4)33.8 (13.1)*Height (m)1.66 (0.08)1.66 (0.09)1.70 (0.09)*Weight (kg)73.8 (15.8)67.7 (13.5)71.0 (15.5)BMI (kg/m2)26.7 (5.1)**24.5 (3.9)24.5 (4.5)WC84.8 (13.3)84.2 (10.9)83.5 (13.6)WHR0.81 (0.09)0.85 (0.08)***0.82 (0.08)Body fat (%)32.8 (9.1)**30.1 (9.3) 28.0 (11.0)M=male, F=female, BMI=body mass index, WC=Waist Circumference, WHR= waist-to-hip ratio. *Caucasian group significantly older (p=0.02) and taller (p=0.001) than Afro-Caribbean and Asian group. **Afro-Caribbean group significantly higher BMI (p=0.001) and significantly higher %BF (p=0.001) than Caucasian and Asian group. *** Asian group significantly higher WHR (p=0.02) compared with Caucasian and Afro-Caribbean groups.

Caucasians were significantly older [F 2, 309= 3.84 p=0.02] and taller [F 2, 309=7.24 p=0.001] than Afro-Caribbeans and Asians. Afro-Caribbeans had a significantly higher BMI [F 2, 308 =7.54 p=0.001] and %BF [F 2, 308=6.98 p=0.001] than Caucasians and Asians. Asians had a significantly greater waist-to-hip ratio [F 2, 305=4.0 p=0.02] than Caucasians and Afro-Caribbeans.

Table 4.2 shows the distribution of overweight and obesity between the groups before matching using the WHO classifications weight.
Table 4. 2 Classification of weight using World Health Organization BMI categories (N, % of group in this category)
Afro-CaribbeanAsianCaucasianNormal weight, N, (%)46, (43.4%)25, (61%)104, (63%)Overweight, N, (%)39, (36.8%)14, (34.1%)45, (27.3%)Obese I, N, (%)18, (17.0%)2, (4.9%)11, (6.7%)Obese II, N, (%)3, (2.8%)-5, (3.0%)*Normal=(18.5-24.9kg/m2), Overweight=(25.0-29.9kg/m2), Obese I=(30.0-34.9kg/m2), Obese II=(35.0-39.9kg/m2) (Source: WHO 2000).

In all three groups, the majority were classified as normal weight. More Afro-Caribbean men and women were classified as overweight and obese (36.8% and 17.0% respectively) compared with Caucasian and Asian groups. 
4.3.2 Matched group comparisons

Table 4.3 shows comparisons of age, height, weight, BMI and %BF in the matched Afro-Caribbeans, Asians and Caucasians.

Table 4. 3 Matched subject characteristics (mean (SD))
Afro- CaribbeanAsianCaucasianNumber (M, F)(14, 12)(14, 12)(14, 12)Age (years)26.6 (7.5)26.9 (7.7)27.0 (7.2)Height (m)1.70 (0.08)1.67 (0.1)1.71 (0.08)Weight (kg)70.2 (12.0)68.0 (13.0)71.5 (11.2)BMI (kg/m2)24.2 (3.1)24.2 (3.1)24.4 (3.1)WC (cm)80.7 (9.8)83.8 (9.7)84.0 (9.4)WHR0.82 (0.1)0.85 (0.06)0.83 (0.07)Body fat (%)27.0 (9.0)29.5 (8.1)25.5 (10.1)M=Male, F= Female, BMI=body mass index, WC=Waist C, WHR=waist-to-hip ratio
There were no significant differences in any of the measures in Afro-Caribbeans, Asians and Caucasians. After matching, 16 of each group were classified as normal weight and 10 of each group as overweight using the WHO BMI categories.


4.3.2.1 Relationship between body mass index and % body fat

There was a weak positive non significant correlation between BMI and %BF in Afro-Caribbeans (r=0.22 n=26 p=0.28), and Asians (r=0.28 n=26 p=0.16). However, in Caucasians the relationship was stronger and significant (r=0.55 n=26 p =0.003). 

4.3.2.2 Age, gender, body mass index and ethnicity as predictors of percentagebody fat 

Table 4.4 shows the summary of the model used to predict %BF in Afro-Caribbeans, Asians and Caucasians. 
Table 4. 4 Summary of model to predict percentagebody fat 
Dependant variable- percentagebody fatIndependent variables/ PredictorsRR2SEAdjusted R2P valuesBMI, Age, Gender, Ethnicity1 and Ethnicity2 0.760.605.990.570.04**Model significantly predicts %BF, (p=0.04)
The model, which included BMI, age, gender and the dummy variables ethnicity1 and ethncity2 explained 60% of the variance in %BF. The adjusted R2 corrected for the optimistic overestimation of the true population value, thus the model explained 57% of the variance in %BF.


4.3.2.2.3 Evaluating the contribution of independent variables to the variance and prediction of percentagebody fat

Table 4.5 shows the individual contribution of the independent variables to the variance in %BF.
Table 4. 5 Evaluation of the contribution of independent variables to the variance and prediction of percentagebody fat
Dependant variable-percentagebody fat BSE  B
(P values(Constant)-24.546.5 0.0001BMI (kg/m2)1.750.310.58*0.0001Age0.070.120.050.59Gender12.471.700.68*0.0001Ethincity1
(C vs AC)1.81.660.090.28Ethnicity2
(C vs A)4.351.660.23**0.01AC=Afro-Caribbean, A=Asian, C=Caucasian *significantly influenced the BMI-%BF relationship (p=0.0001), **(p=0.01)
Body mass index, gender and ethnicity2 significantly predicted %BF. In addition the largest beta value is for gender (0.68), implying that gender had the largest influence on the variance in %BF, followed by BMI (0.58) and ethnicity2 (0.23).


To assess the contribution of ethnicity, the two dummy variables, ethncity1 and ethnicity2 were tested against the model. Table 4.6 shows the model summary of ethnicity1 and ethnicity2.
Table 4. 6 Summary of prediction of model of ethnicity variables
Dependant variable- percentagebody fatIndependent variables/ PredictorsRR2SEAdjusted R2P valuesEthnicity1 and Ethnicity2 0.180.039.120.0070.03**model significantly predicts %BF
The combined ethnicity variable significantly predicts %BF but accounts only for 0.7% of the variance in %BF.

4.4 Discussion

It has been suggested that the relationship between BMI and %BF differs with ethnicity (Duerenberg et al., 1998). However, it has also been reported that only age and gender and not ethnicity influence the relationship between BMI and %BF (Gallagher et al., 1996). Therefore, the aim of this study was to compare the relationship between BMI and %BF in Afro-Caribbean, Asian and Caucasian men and women matched for age, gender and BMI. Correlations showed that the relationship between BMI and %BF differed between Afro-Caribbeans, Asians and Caucasians. In addition multiple regression analysis showed that ethnicity had a significant effect on the relationship between BMI and %BF. However, the effect of ethnicity on the BMI-%BF relationship was small, accounting for only 0.7% of the variance in %BF.
The findings of the current study are consistent with that of Jackson et al., (2002) where BMI and %BF was compared between AA and Caucasian men and women. It was found that the effect of ethnicity was significant but small and found only in women. Other studies have found that BMI-%BF relationship differed with ethnicity. Wang et al., (1994) studied the correlations between BMI and %BF measured by dual photon absorptiometry (DPA) in Asians compared with Caucasians. Asians had a lower BMI but higher %BF compared with Caucasians. Gurrici et al., (1998) found that Indonesians having the same weight, height, age and gender had 4.8% more body fat compared to Dutch Caucasians. In addition, the Indonesians with the same % BF, age and gender had a 2.9 kg/m2 lower BMI compared with the Caucasians. The addition of group (i.e. Indonesian or Caucasian) to the regression model explained 0.02% of the variance in %BF. 


4.4.1 Effect of individual matching for age, gender and body mass index

Deurenberg et al., (1998) suggested that in order to compare the relationship between BMI and %BF accurately in different ethnic groups, the groups should be comparable in terms of age, gender and body build. Deurenberg et al., (1999) matched Singaporean and Beijing Chinese to Dutch Caucasians for age (?3years), gender and BMI (?2kg/m2) and found that the BMI and %BF relationship differed between the two groups. After matching, the two Chinese groups did not differ significantly, however, the Caucasian group remained significantly heavier and taller than the Chinese. In this case, matching did not make the groups comparable. 
In the current study, subjects were individually matched for age (?3years), gender and BMI (?2kg/m2). Before matching, the groups differed significantly in age, height, BMI, WHR and %BF. After matching for age, gender and BMI, there were no significant differences between the groups. In terms of these measures the groups could be considered comparable with the exception of ethnicity. Results of Pearsons? correlation showed that BMI-%BF relationship differed between the groups.  Therefore differences in the BMI and%BF relationship seen in this group of matched Afro-Caribbeans, Asians and Caucasians can be attributed to ethnicity.
Other studies have narrowed the BMI and age range of their study participants to achieve comparability. Gallagher et al., (1996), reported no difference in BMI-%BF relationship in a study of participants whose BMI did not exceed 35kg/m2. Duerenberg et al., (1997) found that individually matching Chinese to Dutch Caucasian was difficult because the Chinese were significantly smaller in body build than the Caucasians. Therefore, the Caucasian group was selected such that the values for age, weight, height and BMI did not exceed the minimum and maximum values for the Chinese. In both these studies, ethnicity did not contribute to the variance in %BF. In both cases gender and age were more significant predictors of %BF. 

4.4.2 Use of different methods for estimating % body fat in different ethnic groups

In these studies, different methods were used to estimate %BF, including underwater weighing, deuterium dilution and dual energy x-ray absorptiometry, skinfold analysis and BIA. The accuracy of these methods varies greatly in their ability to predict %BF (Heymsfield et al., 1996).  The assumptions that govern their use were derived from the analysis of a limited number of predominately male Caucasian cadavers (Mitchell et al., 1945, Widdowson et al., 1951 Forbes et al., 1953, 1956 and Moore et al., 1968). Therefore, they may be biased when applied to non-Caucasian groups.
Multi-compartmental body composition models, in particular, the four-compartmental model is based on fewer assumptions and as such is considered to be independent of age, gender and ethnicity (Withers et al., 1998). Of the studies reviewed here, only two used multi-compartmental models for the estimation of %BF. Density was estimated from hydrodensitometry and air displacement; TBW from tritium dilution and deuterium oxide; and mineral from dual photon absorptiometry and dual x-ray absoptiometry, using the equations of Heymsfield et al., (1992) and Baumgartner et al., (1991) respectively, (Gallagher et al., 1996, Deurenberg-Yap et al., 2000). However, two-compartment models are more widely used as the requirement for several reference methods, which maybe expensive and require technical expertise, limits the use of multi-compartment models.
In this study the two-compartment model of Siri (1961) was used to convert body density, measured by ADP, to %BF. In the matched participants, the results showed no significant differences in mean %BF between the groups. However, it is possible that the %BF measure in Afro-Caribbeans is biased. This is because the Siri (1961) equation, is based on the assumption of constant density of FFM in all population groups. However, it is known that DFFM differs between Afro-Caribbeans and Caucasians (Schutte et al., 1994, Ortiz et al., 1992, Wagner and Heyward 2001).

4.4.3 Implications for the use of BMI as a proxy for adiposity

The widespread use of BMI is related to its practical advantages over other methods. It is easily administered, uses inexpensive and portable equipment making it useful for large-scale epidemiological studies and clinical use. In 2004, the WHO published recommendations for the lowering of BMI cut-off points for overweight and obesity, from 25 to 23 kg/m2 and 30 to 27 kg/m2 respectively when applied to Asian populations. This was in response to reports of higher %BF at lower BMIs than Caucasians in this population group (Wang et al., 1994, Deurenberg et al., 2001, Deurenberg-Yap et al., 2002). Moreover, it has been suggested that cut-off points for overweight and obesity should be higher in AAs to reflect the lower %BF at the same BMI as American whites (Wagner and Heyward 2000). 
This study adds to the evidence that shows that BMI does not represent adiposity equally in people of different ethnic groups. Therefore, the universal cut-off points for overweight and obesity may not be appropriate in all population groups.


4.4.4 Study limitations

Three hundred and twenty-two (322) participants were recruited to this study from the staff and student population. This was a convenience sample and therefore the results of this study may not be generalisable to other groups. The groups differed significantly in age as the Caucasian population reflect the staff who were significantly older than the students. The Asian group had more males than females, this also reflected the composition of the Asian students on the university campus.
It was difficult to match the study participants as, firstly, the number of Asians recruited to the study was low. Secondly, the Afro-Caribbean and Asian group had the largest differences, Afro-Caribbeans had the highest BMI, while the Asian group had the lowest. The large number of Caucasian participants made it easier to match the groups. Only 26 of each group were matched.
percentagebody fat in this study was estimated by ADP and the two-compartment densitometry equation of Siri (1961).  This method is not free of assumptions, which have been found to be biased when applied to non-Caucasian groups. However, it was not possible to use a multi-compartment model, as air displacement was the only reference method available.


4.5 Conclusion

The effect of ethnicity on the BMI-%BF relationship is controversial. This study showed a small and significant effect of ethnicity. This finding supports that BMI does not represent adiposity equally in all ethnic groups. Future studies will explore the relationship between adiposity and risk factors for CVD in different ethnic groups.


4.6 References

Chang, C-J., Wu, C-H., Chang, C-S., Yao, W-J., Yang, Y-C., Wu, J-S., and Lu, F-H., (2003). Low body mass index but high % body fat in Taiwanese subjects: implications of obesity cutoffs. International Journal of Obesity, 27, pp.253 ? 259.

Chung, S., Song, M-Y., Shin, H-D., Kim, D-Y., He, Q., Heshka, S., Wang, J., Thornton,J., Laferr?re, F., Xavier Pi-Sunyer, F.,  Gallagher, D., (2005). Korean and Caucasian overweight premenopausal women have different relationship of body mass index to % body fat with age. Journal of Applied Physiology, 99, pp.103-107.

Deurenberg, P., (2001). Universal cut-off point BMI points for obesity are not appropriate. British Journal of Nutrition, 85, pp.135-136.
Deurenberg, P., Deurenberg-Yap, M., Wang, J., Lin, F.P., Schmidt, G., (1999). The impact of body build on the relationship between body mass index and % body fat. International Journal of Obesity, 23, pp.537-542.
Deurenberg, P., Duerenberg-Yap, M., & Guricci, S., (2002). Asians are different from Caucasians and from each other in their body mass index/body fat percentagerelationship. Obesity Reviews, 3, pp.141-146.
Deurenberg, P., Ge K., Hautvast, J.G.A.J., Wang, J., (1997). Body mass index as predictor for body fat: comparison between Chinese and Dutch adult subjects. Asia Pacific Journal of Clinical Nutrition, 6, pp.102-105.
Deurenberg, P., Yap, M., van Stavern, W.A., (1998). Body mass index and percentagebody fat: a meta-analysis among different ethnic groups. International Journal of Obesity, 22, pp.1164-1171. 
Deurenberg-Yap, M., Schmidt, G., van Staveren, W.A., Deurenberg, P., (2000). The paradox of low body mass index and high percentagebody fat among Chinese, Malays and Indians in Singapore. International Journal of Obesity, 24, pp.101-1017.
Deurenberg-Yap, M., Chew, S.K., Deurenberg, P., (2002). Elevated body fat percentageand cardiovascular risks at low body mass index levels among Singaporean Chinese, Malays and Indians. Obesity Reviews, 3, pp.209-215.
Dudeja, V., Misra, A., Pandey, G., Devina, G., Kumar, G., Vikram, N.K., (2001). BMI does not accurately predict overweight in Asian Indians in Northern India. British Journal of Nutrition, 86, pp.105-112.
Field, A., (2005). Discovering statistics using SPSS. 2nd edn. Sage Publications: London
Forbes, R.M., Cooper, A.R., Mitchell, H.H., (1953). The composition of the adult human body as determined by chemical analysis. Journal of Biological Chemistry, 203, pp.359-366.
Forbes, R.M., Mitchell, H.H., Cooper, A.R., (1956). Further studies on the gross composition of the adult human body. Journal of biological chemistry, 223, pp.969-975.
Gallgher, D., Visser, M., Sepulveda, D., Pierson, R.N., Harris, T., Heymsfield, S.B., (1996). How useful is body mass index for comparison of body fatness across age, sex and ethnic groups? American Journal of Epidemiology, 143, pp.228-239. 
Gallagher, D., Heymsfield, S.B., Heo, M., Jebb, S., Murgatroyd, P.R., Sakamato, Y., (2000). Health percentagebody fat ranges: an approach for developing guidelines based on body mass index. American Journal of Clinical Nutrition, 72, pp.694-710.
Garrow, J.S., Webster J., (1985). Quetelet's index (W/H2) as a measure of fatness. International Journal of Obesity, 9, pp.147-153.
Gurrici, S., Hartriyanti, Y., Hautvast, J.G.A.J., Deurenberg, P., (1998). Relationship between body fat and body mass index: differences between Indonesians and Dutch Caucasians. European Journal of Clinical Nutrition, 52, pp.779-783.

Heymsfield, S. B., Tighe, A., Wang, Z. M., (1992). Nutritional assessment by 
anthropometric and biochemical methods. In: Shils, M.E., Olson, J. A., Shike, M., eds. Modern nutrition in health and disease. 8th ed. Philadelphia: Lea and Febiger , pp.812?41.

Heymsfield, S.B., Wang, Z.M., Withers, R., (1996). Multicomponent molecular-level models of body composition analysis. In Roche A, Heymsfield SB, Lohman T, eds. Human body composition. Champaign, IL: Human Kinetics Publishers 

Jackson, A.S., Stanfroth, P.R., Gagnon, J., Rankinen, T., Leon, A.S., Rao, D.C., Skinner, J.S., Bouchard, C., Wilmore, J.H., (2002). The effect of sex, age and race on estimating percentagebody fat from body mass index: The Heritage Family Study. International Journal of Obesity, 26, pp.789-796.
Luke, A., Arvizu, R.D., Rotimi, C., Prewitt, E., Forrester, T., Rainford, W., Ogunbiyi, O.J., Schoeller, D.A. Mcgee, D., Cooper, R.S., (1997). Relation between body mass index and body fat in black population samples from Nigeria, Jamaica, and the United States. American Journal of Epidemiology, 145, pp.620-628.
Marfell-Jones, M., Olds, T., Stewart, A. and Carter, L., (2006). International standards for anthropometric assessment. ISAK: Potchefstroom
Mitchell, H.H., Hamilton, T.S., Steggerda, F.R., Bean, H.W., (1945). The chemical composition of the adult human body and its bearing on the biochemistry of growth. Journal of biological chemistry, 158, pp.625-637.
Moore, F.D., Lister, J., Boyden, C.M., Ball, M.R., Sullivan, N., Dagher, F.J., (1968). The skeleton as a feature of body composition. Human Biology, 40, pp.136-188.
Rush,E.C Goedecke, J.H.,Jennings, C.,?Micklesfield, L.,Dugas, L., Lambert, E.V., and Plank, L.D.(2007). Body mass index, fat and muscle differences in urban women of five ethnicities from two countries. International Journal of Obesity,  31, pp.1232-1239.
Swinburn, B.A., Craig, P.L., Daniel, R., Dent, D.P.D., Strauss, B.J.G., (1996). Body composition differences between Polynesians and Caucasians assessed by BIA. International Journal of Obesity, 20, pp.889-894.
Wagner, D.R., Heyward, V.H., (2000). Measures of body composition in blacks and whites: a comparative review. American Journal of Clinical Nutrition, 71, pp.1392-1402. 
Wang, J., Thornton, J.C., Russell, M., Burastero, S., Heymsfield, S., (1994). Asians have lower body mass index (BMI) but higher % body fat than do whites: comparisons of anthropometric measurements. American Journal of Clinical Nutrition,  60, pp.23-28.
Withers, R.T., JaForigia, J., Pilans, R.K., Shipp, N.J., Chatterton, B.E., Schultz, C.G., Leaney, F., (1998). Comparisons of two-, three- and four-compartmental models of body composition analysis in men and women. Journal of Applied Physiology, 1, pp.238-245.
Widdowson, E.M., McCance, R.A., Srapy, C.M., (1951). The chemical composition of the human body. Clinical Science, 10, pp.113-125.
World Health Organization (2000). Obesity: preventing and managing the global epidemic. Report of a WHO Consultation WHO Technical Report Series, 894, Geneva: World Health Organization. 
World Health Organization (2004). Appropriate body mass index for Asian populations and its implications for policy and intervention strategies. The Lancet, 363, pp.157-16
 A COMPARISON OF TWO COMPARTMENT DENSITOMETRY EQUATIONS FOR THE ESTIMATION OF percentageBODY FAT IN AFRO-CARIBBEANS

5.1 Introduction

The two-compartmental (2-C) densitometry equation of Siri (1961) (table 5.1) is widely used to predict percentagebody fat (%BF) from body density (Db) in all population groups. It assumes that body weight is divided into two-compartments, FM and FFM, with constant densities of 0.9007g/cm3 and 1.100g/cm3 respectively. The assumption of constant DFFM is the basis upon which densitometry was established (Keys and Brozek 1953, Brozek et al., 1963). From cadaver analysis, Keys and Brozek (1953) found a constant DFFM could only be assumed if the composition of  FFM was constant i.e. constant mineralisation, hydration and protein content with constant densities. 


Table 5. 1 Body composition prediction equations
ReferenceEquationDensity of fat free massPopulationSiri (1961)%BF= (4.95/ Db-4.50)*1001.100g/cm3GeneralisedBrozek et al.,  (1963)%BF= (4.570/Db-4.142)*1001.100g/cm3GeneralisedPace and Rathburn (1945)%BF=[(wt-(TBW/0.732)/wt]*100-GeneralisedSchutte et al., (1984)%BF = (4.374/Db-3.928)*1001.113g/cm3African American MaleOrtiz et al., (1992)%BF = (4.83/Db- 4.37)*1001.106g/cm3African American FemaleWagner and Heyward (2001)%BF= [4.858/Db-4.394]*1001.1570g/cm3African American Male%BF=% body fat, wt=weight in kilograms
However, these assumptions are based on a small number of predominately male Caucasian cadavers. Therefore, their applicability to non-Caucasian groups is questionable (Deurenberg and Deurenberg-Yap 2003).

Differences in the composition of FFM have been found between people of African origin and those of Caucasian origin. Cohn et al., (1977) compared levels of total body calcium, phosphorous, sodium, chlorine and potassium in  AA men and women aged 30-80years with an age and sex matched Caucasian population. Total body calcium (TBCa) was higher in the AA than the Caucasians, AA men and women had 21.9% and 16.7% higher TBCa than the Caucasian men and women respectively. In addition total body potassium (TBK) was 16.8% and 15.3% higher in AA men and women, reflecting a higher FFM. This study was among the first to show the difference in mineralisation of FFM in AAs compared with Caucasians.
Wagner and Heyward (2000) reviewed in vitro and in vivo studies comparing populations of African origin and those of Caucasian origin (referred to as Black and White in the review) relative to FFM composition and including differences in hydration, mineralisation and protein content. The studies consistently showed no significant differences in hydration of FFM between Black and White populations; Black men and women had a higher bone mineral density (BMD) and protein content than White men and women. Moreover, the DFFM was higher in African compared Caucasian origin populations. The difference in DFFM was attributed to the higher mineral content of FFM in Black people. Therefore, it is possible that the assumptions of the generalised equation of Siri (1961), which was developed in a young Caucasian male population, may not be applicable to populations of African origin.
Alternative 2-C densitometry equations have been developed to estimate %BF in people of African origin. Schutte et al., (1984) compared FFM in 15 AA males with a height and weight matched sample of 19 Caucasian males. Fat free mass and %BF were estimated by densitometry (underwater weighing) and compared with  TBW estimation (deuterium oxide dilution). In the Caucasian group, there were no significant differences between estimations of FFM and %BF calculated by densitometry, using the equation of Brozek et al., (1963) (table 5.1.). Similarly, the measurement of FFM and %BF by TBW using the equation of Pace and Rathburn (1945) (table 5.1) showed no differences. However, in AA males densitometry yielded significantly greater FFM and lower %BF than the TBW analysis. Moreover, the DFFM in AA was estimated to be 1.113g/cm3, which was greater than that estimated by Siri (1961). 

From this study, Schutte et al., (1984) developed an alternative 2-C equation for the estimation of %BF from Db in AA males. The Schutte et al., (1984) equation assumes that the DFFM in AA males is 1.113g/cm3 and is widely used to estimate %BF in males of African origin (table 5.1). However, the size and lack of heterogeneity in the Schutte et al., (1984) sample has implications for the generalisability of the equation to other men of African origin. The sample consisted of young collegiate men of a narrow age range (18-24 years) therefore, the equation may not be appropriate for use in older men. In addition, the two-compartment hydrometry model (Pace and Rathburn 1945), which assumes a constant hydration of FFM of 73.2% was used as a reference for the estimation of %BF and FFM. Three or four compartment models are assumption free, as they measure the individual compartments of water, mineral and/or protein. Therefore, they would have been more appropriate reference models (Withers et al., 1998). 
Wagner and Heyward (2001) cross-validated the Schutte et al., (1984) equation, in a sample of 30 AA males. The four compartmental (4-C) model of Friedl et al., (1992) was used as the reference. The relationship between the 4-C model and the Schutte et al., (1984) were strong (r=0.98). However, the Schutte et al., (1984) equation consistently overestimated %BF in 87 % of the sample. It was concluded that the DFFM for AA males of 1.113g/cm3 assumed by Schutte et al., (1984) must be incorrect. Wagner and Heyward (2001) estimated the DFFM for their sample of AA male to be 1.10570g/cm3 and developed an alternative 2-C equation (table 5.1).
The Wagner and Heyward (2001) sample was more heterogeneous and representative of AA males aged 19-45 years than the Schutte et al., (1984) sample. However, Wagner and Heyward (2001) noted that their sample had a disproportionately high number of physically active participants. Physical activity increases muscularity, which may influence the hydration, mineralisation and protein content and consequently the DFFM (Millard-Stafford et al., 2001).
Visser et al., (1997) found the DFFM to be 1.099g/cm3 in sedentary Black men aged 20-94years, which was attributed to a higher mineralisation of FFM in Black versus White men (7.0 vs. 6.6%) but no difference in hydration. Millard-Stafford et al., (2001), found that the DFFM in a sample of 64 sedentary AA males was 1.098 (0.002)g/cm3 which is significantly lower than that assumed by the Schutte et al., (1984)  (1.113g/cm3) and Wagner and Heyward (1.1057g/cm3) but closer to that assumed by Siri (1961) (1.100g/cm3). Moreover, significant differences were found in %BF but not Db between AA and Caucasian males. Therefore, Millard-Stafford et al., (2001) concluded that there was no need for the use of ethnic specific equations when estimating %BF from Db, as there was no biological basis for the use of the different equations, in support of Visser et al., (1997). 
The equations are only applicable to men. Ortiz et al., (1992) compared FFM composition in 28 AA women matched for age, height, weight and menstrual status to 28 Caucasian women using a the 4-C model. African American women had an eight % higher total body potassium (TBK), 13.8% greater bone mineral content (BMC), and a bone mineral density of 1.18 (0.14), which was significantly greater than Caucasians women (1.09 (0.09) n= 28 pairs p<0.001). The DFFM in AA women was calculated as 1.106g/cm3 which was higher than that calculated for the Caucasian women which was 1.0961g/cm3. Ortiz et al., (1992) developed an alternative 2-C equation, which is used to estimate %BF in women of African origin (table 5.1).
Thus from the studies above, it is unclear whether ethnic specific equations provide a more accurate measure of %BF compared with the generalised equation of Siri (1961) in populations of African origin. It is necessary to compare these 2-C model equations with a multi-compartment model, which is based on fewer assumptions. However, the use of multi-compartment models is limited because of the requirement for three or four reference methods for example, hydrodensitometry, isotope dilution or dual energy x-ray absorptiometry. These methods are expensive, require technical expertise and large and cumbersome equipment. 
It is possible to develop alternative multi-compartmental models, which incorporate easily accessible measures of body composition such as BIA (BIA). Evans et al., (2001) studied the accuracy of modified multi-compartmental models incorporating TBW estimated by BIA instead of isotope dilution. The aim was establish if the modified three and four compartment models were equally accurate as a conventional 4-C model, which consisted of TBW estimated by isotope dilution. In addition, Evans et al., (2001) aimed to establish weather the modified 3- and 4-C models were more accurate than the 2-C densitometry equation of Siri (1961). It was found that the modified 3 and 4-C models were less accurate that the reference 4-C, but did not give a significantly different estimate of %BF compared with the 2-C densitometry equation of Siri (1961).
Thus it is hypothesised that given that the modified 3-C model did not differ significantly in the estimate %BF by density alone, this model can be used as a reference against which the 2-C densitometry equations of Schutte et al,.(1984), Ortiz et al., (1992), Wagner and Heyward (2001) and Siri (1961) are compared. In addition, DFFM can be calculated from this modified multi-compartment model, allowing for a comparison between Afro-Caribbeans and Caucasians.
Therefore, the aims of this study were to compare %BF estimated by the ethnic specific equations of Schutte et al.,  (1984), Ortiz et al., (1992) and Wagner and Heyward (2001), with the generalised equation of Siri (1961) and a modified 3-C reference (Siri 1956) incorporating density estimated by ADP and TBW estimated by BIA in Afro-Caribbean men and women. A secondary aim was to compare DFFM and body density between this Afro-Caribbean group and a Caucasian group.

5.2 Methods

5.2.1 Subjects

Subjects were recruited from the staff and student population at a university in South East England, through advertisement on staff and student media, (students? newsletter, staff and student intranet), posters, canvassing and through personal contacts. Sixty-seven Afro-Caribbean men and women participated in the study. The Caucasian control group consisted of 122 men and women. Subject characteristics are described in table 5.2.

Ethical approval and inclusion criteria is as described n chapter four of this thesis
5.2.2 Outcome Measures

All measurements were made on a single occasion in a dedicated research laboratory within the University campus. Subjects were asked to report to the laboratory not less than one hour after eating, having emptied their bladder and refrained from exercise in the previous 24 hours. 

5.2.2.1 Anthropometry

All anthropometric measurements were taken according to the guidelines set by the International Society for the Advancement of Kinanthropometry (ISAK) manual (Marfell-Jones et al., 2006).

5.2.2.1.1 Height
Height was measured (to the nearest completed10mm) using a free standing Leicester Height Measure (Invicta Products, Leicester). 

5.2.2.1.2 Weight
Weight was measured (to the nearest completed10g) using the electronic scale (Tanita Corporation, Japan) provided with the BOD POD S/T system. Subjects were barefoot and wore tight fitting swimwear. 

5.2.2.2 Body density

Body density was derived by ADP in the BOD POD body composition tracking system (Life Measure Inc.). Body density was calculated by dividing body weight by body volume. The volume of air displaced by the subject while seated in the BOD POD was taken as the equivalent to the subject body volume. This volume was corrected for thoracic gas volume (TGV) i.e. the inside the lungs and surface area artefact (SAA) i.e. the air near the skin. A moulded fibreglass seat formed a common wall separating the front (test) and rear (reference) chambers. The subject used the front chamber during testing while the rear chamber was used as a reference volume. Subjects entered the BOD POD through a hinged door at the front of the test chamber. Two electromagnets and an air gasket sealed the door during measurement of volume. 

5.2.2.3 Bioelectric Impedance Analysis

The Tanita bioelectric impedance analyser (Tanita Corporation, Japan) consisted of a platform on which two metal plates where mounted. A column attached the platform to the control panel, which was elevated. Two metal plates were mounted on the platform in the shape of feet, left and right hand grips were attached to the control panel, these were the electrodes used in the impedance analysis. The control panel also housed a small printer and a digital display screen. 
Prior to testing all jewellery and spectacles were removed. The subjects? age and height were entered into the control panel.  Subjects stood bare foot on their analyser, with their feet directly on the metal plates. Their weight was taken. Subjects were then instructed to grasp the two leads and hold them firmly in their hands and let go only when instructed.
A low level electric current (50kHz) was passed through the body and the current impedance or resistance to flow was measured. 

5.2.2.4 Two?compartmental densitometry equations

The following equations were used for calculations of %BF.
5.2.2.4.1 Generalised:

%BF=(4.95/Db-4.50)*100 (Siri 1961)
5.2.2.4.2 Ethnic specific:

%BF=(4.374/Db-3.928)*100 (Schutte et al., 1984) 
%BF= (4.83/Db-4.37)* 100 (Ortiz et al., 1992) 
%BF= (4.858/Db-4.394)*100 (Wagner and Heyward 2001) 


5.2.2.5 The reference model

The 3-C model of Siri (1956)  was used as the reference against which the 2-C models were compared
%BF=211.76/Db-78*(TBW/weight)-135.1 (Siri 1956)

5.2.2.5.1 Calculation of density of fat free mass

Density of FFM was solved from the equation
DFFM = 1/ [(W/DW)+((/D()]
Where W is the water compartment of the body as estimated by BIA, DW is the density of water, which according the original cadaver analysis by Keys and Brozek (1953) is 0.9937g/cm3, ( is the combined protein and mineral compartment which according to Siri (1956) is 0.35 with a density of 1.565g/cm3.

5.2.3 Statistical analysis

All statistical analysis was carried out using SPSS version 15.0 (SPSS Inc., Chicago, IL, USA). Normality was assessed using the Kolmorgorov-Sminov test. All data met the assumptions for normality. One way repeated measures ANOVA was used to note any significant differences between the estimations of %BF using the equations of Siri (1961), Schutte et al., 1984 and Wagner and Heyward (2000) in the Afro-Caribbean men and Siri (1961) and Ortiz et al., (1992) in Afro-Caribbean women. Bland-Altman analysis was conducted to assess the agreement in the estimation of %BF between the 2-C equations and the reference 3-C equation. To show agreement between methods, the limits have been placed at -3.5 and 3.5%, which is an acceptable difference between methods as suggest by Siri (1956).
5.3 Results 

Table 5.2 shows subject characteristics for the Afro-Caribbean group and the Caucasian comparison group.

Table 5. 2 Subject characteristics (mean (SD))
Afro-CaribbeanCaucasianMFMFN17504874Age (years)29.4 (8.5)31.2 (10.0)34.0 (13.0)35.2 (13.8)Height (m)1.73 (0.06)1.64 (0.06)1.78 (0.06)1.65 (0.06)Weight (kg)84.6 (14.0)74.0 (16.2)81.7 (15.1)64.9 (12.9)BMI (kg/m2)28.2 (3.7)27.4 (5.7)25.7 (4.2)24.0 (4.8)BMI=body mass index
The Caucasian comparison group consisted of 48 males and 74 females whose minimum and maximum values for age and BMI did not exceed those for the Afro-Caribbean group. 


5.3.1 Comparison of two-compartment equations with reference in Afro-Caribbean men and women

Table 5.3 shows percentagebody fat estimated by the densitometry equations compared with the 3-C reference.
Table 5. 3 percentagebody fat (mean (SD))
EquationMen (17)Women (50)Siri (1961)24.5 (6.2)*37.5 (7.4)*Schutte et al., (1984)26.5 (5.4)*-Ortiz et al.,  (1992)-38.7 (7.2)*Wagner & Heyward (2001)26.3 (6.0)*-3-C reference31.4 (6.5)40.4 (4.3)* significantly  different from the reference p<0.001, 

In the men %BF estimated by the three equations were significantly different from the 3-C reference ((%BF=211.76/Db-78*(TBW/weight)-135.1 (Siri 1956) F [1.026; 14.35]=14.629 p=0.002.  Compared with the 3-C reference, all 2-C equations yield significantly lower %BF estimates, with the biggest difference being between the 3-C reference and Siri (1961). There were no significant differences between by Schutte et al., (1984) and Wagner and Heyward (2001).
In the women %BF estimated by Siri (1961) and Ortiz et al., (1992) were significantly different from the reference F [1.005; 47.230]=13.98 p=0.0001. The 3-C reference gave significantly higher estimates of %BF (40.4 (4.3) n= 50, p<0.001). However, there were no significant differences between %BF estimated by Siri (1961) or by Ortiz et al., (1992). 


Table 5. 4 Agreement between 3-C reference and 2-C equations
Gender (n)Men  (17)Women (50)EquationMean diff (limits)Mean diff (limits)Siri (1961)6.61 (-4.7; 17.9)2.68 (-5.9; 11.2)Schutte et al., (1984)4.55 (-6.2; 15.3)-Wagner and Heyward (2001)4.72 (-6.5; 15.9)-Ortiz et al., (1992)-1.47 (-6.9; 9.8)Mean diff= Mean difference, limits=limits of agreement

Table 5.4 shows the agreement between %BF estimated by the reference and by the four 2-C equations. In men, the mean differences are large and are not within the acceptable range of ?3.5% for the difference in %BF estimation methods. In women, the mean differences are small and acceptable, however the limits of agreement are very wide.
Figure 5.1 A-E show the Bland-Altman analysis assessing the agreement of the four 2-C equations to the 3-C reference. 
AFRO-CARIBBEAN MALES

                                                                    Figure  5.1 A                                            Figure  5.1 B                                                    Figure 5.1 C
AFRO-CARIBBEAN FEMALE

                                                                                                        Figure 5.1 D                                                         Figure 5.1 E
Figure 5. 1 A-E  Bland-Altman analysis of the individual residual %BF scores for 2-C equations. Residual %BF= %BF from 3-C reference minus %BF from 2-C equation. Mean %BF= (%BF from 3-C reference +%BF from 2-C equation divided by 2)It can be seen in figures A, B and C (Afro-Caribbean men) that the majority of the points fell outside of the acceptable limits of ?3.5%. This suggests that there was poor agreement between the 2-C equations and the 3-C reference. In the Afro-Caribbean females, figures D and E show better agreement between the 2-C equations and the 3-C reference as a larger number of points fell with in the acceptable limits of ?3.5%.

 5.3.2 Comparison of density of FFM in Afro-Caribbean and Caucasian group

Table 5.5 shows the comparison of Db and DFFM between the Afro-Caribbean and Caucasian group.

Table 5. 5 Comparison of body density and density of FFM, and between Afro-Caribbeans and Caucasians (mean (SD))
Afro-CaribbeanCaucasianMFMFn17504874Mean Db (g/cm3)1.043 (0.01)1.016 (0.02)1.052* (0.02)1.025* (0.02)(Range)(1.025-1.072)(0.987-1.055)(1.009-1.088)(0.985-1.069)W (TBW/wt)0.585 (0.08)0.469 (0.056)0.601* (0.699)0.527* (0.065)(Range)(0.52-0.85)(0.35-0.60)0.471-0.696)(0.37-0.76)( (0.35/1.565)0.220.220.220.22Mean DFFM (g/cm3)1.242 (0.11)1.446 (0.12)1.212* (0.84)1.336* (0.12)(Range)(0.928-1.339)(1.210-1.745)(1.083-1.433)(1.015-1.673)Db=Body density, W= water fraction of body weight, TBW=total body water, wt=Weight (kg), (=combined protein: mineral ratio, DFFM=density of fat free mass *Significantly different p<0.001


Body density (Db) was significantly (p<0.001) higher in Caucasians compared with Afro-Caribbeans. The DFFM was significantly (p<0.001) lower in Caucasians compared with Afro-Caribbeans. Total body water by BIA was significantly higher in Caucasians compared with Afro-Caribbeans.
However, it must also be noted that the estimates of DFFM in both Afro-Caribbean and Caucasian groups was significantly higher than that assumed by the 2-C equations of Siri (1961), Schutte et al., (1984), Ortiz et al.,  (1984) and Wagner and Heyward (2001). In addition the fraction of water estimated by BIA was significantly lower in both groups that the established 73.2% (Keys and Brozek 1974).

5.4 Discussion

The purpose of this study was to compare %BF estimated by three ethnic specific densitometry equations with the generalised equation of Siri (1961) and a 3-C reference. The main findings showed small differences between %BF estimated by the ethnic specific equations and the generalised equation of Siri (1961). However, these differences were only significant in men. In comparison to the reference, all the densitometry equations underestimated %BF, with Siri (1961) to a greater extent than the ethnic specific equations. In addition the DFFM and Db were significantly different between the Afro-Caribbean and Caucasian groups.


5.4.1 Ethnic specific versus generalised equations in Afro-Caribbean men and women

The generalised equation of Siri (1961) yielded a lower %BF value compared with the ethnic specific equations in Afro-Caribbean males and females. This finding is consistent with Wagner and Heyward (2001) who found that the generalised equations of Siri (1961) and Brozek et al., (1953) underestimated  %BF when compared with a 4-C reference in 30 AA males. Wagner and Heyward (2001) calculated a DFFM of 1.1570 g/cm3 for their sample of AA males which is much higher than that assumed by Siri (1961). The Siri (1961) equation assumes a constant composition and DFFM  in all population groups of 1.100g/cm3.  However, the composition and density of FFM have been shown to differ between people of African origin and Caucasians. Ortiz et al., (1992) found that AA women had higher mineral and protein compartments and therefore a higher DFFM compared with Caucasian women. 

Siri (1956) noted that the protein: mineral ratio in FFM was constant and a variation would not affect estimation of adiposity from densitometry.  Schutte et al., (1984) first developed an alternative 2-C equation for use in AA men, based on the assumption of these differences in FFM composition particularly the higher protein and mineral components. However, Schutte and colleagues did not measure mineralisation of FFM in that study, instead hydration was measured and found to be similar between AA and Caucasian men matched for weight and height despite a higher DFFM. Therefore the difference was attributed to a greater mineral and protein content.  Wagner and Heyward (2001), first cross-validated the Schutte et al., (1984) equation. They measured total bone mineral and protein in their sample of AA males. However, they did not have a Caucasian control group against which to show the elevated mineral compartment in the AA men. Ortiz et al., (1992) did show a difference in mineralisation of FFM between AA women compared with Caucasian women. However, Ortiz and colleagues noted that the magnitude of the differences was small and further investigations were needed to confirm their findings.  Visser et al., (1997) and Millard-Stafford et al., (2001) also found small and non-significant differences in the mineralisation of FFM and therefore concluded that the difference in FFM was small and did not affect the estimation of %BF and therefore there was no biological basis for the development of ethnic specific equations. It was outside the scope of this study to measure differences in the mineralisation of FFM, however, the findings of this study appear to support those of Visser et al.,  (1997) and Millard-Stafford et al., (2001).
5.4.2 Modified three compartment reference model

The substitution of TBW estimated by BIA instead of the isotope dilution technique into a multi-compartment model is unconventional. However dilution techniques such as deuterium oxide (D20) are expensive, therefore their use is limited. It has been shown that BIA is a valid indirect measure of TBW when compared with D20 (Kushner et al., 1986) including when used in people of African origin (Diuom et al., 2005). However, Evans et al., (2001) did not recommend the substitution of TBW estimated by BIA into multi-compartmental models. This following a study that showed that modified 3- and 4-C model incorporating TBW estimated by BIA was less accurate than a traditional 4-C model and no more accurate than estimating density alone.

In this study %BF estimated by the reference was much higher compared with the 2-C densitometry equations. On closer inspection of the model components, the mean hydration fraction, was less than that estimated by Brozek et al., (1963) of 0.732. Siri (1961) estimated a biological variability of 2% in the hydration fraction of FFM and noted that values from 0.694-0.784 were common. Mean values in this study were 0.585, 0.469, 0.60, and 0.526 in Afro-Caribbean males and females and Caucasian males and females respectively. These values are clearly outside of the range identified by Siri (1961) for hydration of FFM and thus the validity of the reference in this study is questionable.

Moreover, the Afro-Caribbean males and females had a higher mean DFFM with a 0.03g/cm3 and 0.11g/cm3 difference respectively, compared with Caucasian males and females.  Schutte et al., (1984) calculated that in order to raise the DFFM from 1.100g/cm3 to 1.113g/cm3, a 0.03g/cm3 difference, a 36% greater bone mineral and protein content would be required. In this study assessment of mineralisation was not possible. However, the DFFM reported by Schutte et al., (1984) is the highest record for people of African origin. In this study the values were unacceptably high, which was related to the low hydration fraction reported.

5.4.5 Study limitations

This first limitation of this study was the small number of Afro-Caribbean males that participated in the study; although the number of Afro-Caribbean men was larger than that in Schutte et al., (1984). Secondly, the sample is a convenience sample recruited from a staff and student population at a University. Therefore the findings of this study may not be generalisable to other population groups. Thirdly, the use of BIA as part of a multi-compartment model limited the ability to give an accurate comparison of density of FFM between the groups. The estimations of hydration using the BIA were below the values estimated by Keys and Brozek (1974) of 73.2%. Therefore, the results of this study should be interpreted with caution.
5.5 Conclusion

It has been shown that the use of BIA in place a dilution technique for the estimation of TBW in a multi-compartmental model as a reference may not be appropriate. The variation in TBW estimates may lead to spurious results. In addition, this study cannot recommend the use of ethnic specific over generalised densitometry equations for the estimation of %BF in Afro-Caribbean males and females.


5.6 References

Brozek, J., Grande, F., Anderson, J.T., Keys, A., (1963). Densitometric analysis of body composition: revision of some quantitative assumptions. Annals of New York Academy of Science, 110, pp.113-140.
Cohn, S.H., Abesamis, C., Zanzi, I., Aloia, J.F., Yasumura, S., Ellis, K.J., (1977). Body elemental composition: comparison between black and white adults. American Journal of Physiology: Endocrinology, Metabolism and Gastrointestinal Physiology, 232, pp.E419-E422.
Collins, M. A., Millard-Stafford, M.L., Evans, E.M., Snow, T. K., Cureton, K.J., Rosskopf, L.B., (2004). Effect of race and musculoskeletal development on the accuracy of air displacement plethysmography. Medicine and Science in Sport and Exercise, 36(6), pp.1070-1077.
Deurenberg, P., Deurenberg-Yap, M., (2003). Validity of body composition methods across ethnic population group. Acta Diabetologica, 40 (Suppl 1), pp. s246-s249.
Evans, E.M., Arngrimsson, S.A., Cureton, K.J., (2001). Body composition estimates from multi-component models using BIA to determine body water. Medicine and Science in Sport and Exercise, 33(5), pp.839-845. 
Friedl, K.E., DeLuca, J.P., Marchitelli, L.J., Vogel, J.A., (1992). Reliability of body fat estimates from a four compartmental model by using density, body water and bone mineral measurements. American Journal of Clinical Nutrition,  55, pp.764-770.
Keys, A., Brozek, J (1953). Body fat in adult men. Physiological Reviews, 33, pp.245-325.
Millard-Stafford, M.L., Collins, M.A., Modlesky, C.M., Snow, T.K., Rosskopf. L.B.,  (2001). Effect of race and resistance training status on the density of fat free mass and % fat estimates. Journal of Applied Physiology, 91(3), pp. 1259-1268.
Ortiz, O., Russell, M., Daley, T.L., Baumgartner, R.N., (1992). Differences in skeletal muscle and bone mineral mass between black and white females and their relevance to estimates of body composition. American Journal of Clinical Nutrition, 55(1), pp.8-13.
Pace, N., Rathburn, E.N., (1945). The body water and chemically combined nitrogen content in relation to fat content. Journal of Biological Chemistry, 158, pp.685-691. 
Schutte, J.E., Townsend, E.J., Hugg, J., Shoup, R.F., Malina, R.N., Blomquist, G.C.,  (1984). Density of lean mass is greater in blacks than in whites. Journal of Applied Physiology, 56(6), pp.1647-1649.
Siri, W.E., (1956). The gross composition of the body. In: Advances in Biological and Medical Physics, edited by Tobias, C.A., and Lawrence, J.H., New York: Academic
Siri, W. E., (1961). Body composition from fluid spaces and density: analysis of methods. In: Brozek J, Hensche A (eds). Techniques for measuring body composition. National Academy of Sciences: Washington DC pp.223 ? 244.

Visser, M., Gallagher, D., Deurenberg, P., Wang, J., Pierson Jnr, R.N., Heymsfield, S.B., (1997). Density of fat free mass: relationship with race, age, and level of body fatness. American Journal of Physiology, 35, pp.E781-E787.
Wagner, D.R., Heyward V.H., (2000). Measures of body composition in blacks and whites: a comparative review. American Journal of Clinical Nutrition, 6, pp.1392-1402.
Wagner, D.R., Heyward, V.H., (2001). Validity of two-compartmental models for estimating body fat of black men. Journal of Applied Physiology, 90, pp. 649-656.
Withers, R.T., JaForigia, J., Pilans, R.K., Shipp, N.J., Chatterton, B.E., Schultz, C.G., Leaney, F., (1998). Comparisons of two-, three- and four-compartmental models of body composition analysis in men and women. Journal of Applied Physiology, 1, pp.238-245
A STUDY OF CVD RISK FACTORS IN AFRO-CARIBBEANS AND CAUCASIANS

6.1 Introduction

The prevalence of cardiovascular disease (CVD) differs between ethnic groups (Cappucio et al., 2004, Scarborough et al., 2010). In people of African origin, or Afro-Caribbeans in the UK, the prevalence of stroke is high, whereas coronary heart disease (CHD) is low compared with the Caucasian population (Agyemang and Bhopal 2003, Tillin et al., 2005, Agyemang et al., 2009). This disease pattern has been seen in other populations of African origin, including AAs and some urban Africans (Cappucio et al., 2004).
The difference in CVD prevalence may be related to variations in risk factors for CVD between Afro-Caribbeans and Caucasians (Agyemang et al., 2009). Established risk factors for CVD are shown in table 6.1


Table 6. 1 Established risk factors for CVD 
Non-modifiableModifiableRisk markers*AgeCigarette smokingLow socio-economic statusGender
HypertensionElevated prothrombolitic factors e.g. fibrinogenFamily historyHypercholesterolemiaMarkers of infection or inflammationDiabetesElevated homocysteineOverweight and obesity**Elevated lipoprotein (a)Physical inactivity**Psychological disorders (e.g. depression)Diet**These show an association with CVD but for whom a causal link has not been established **Known as predisposing risk factors, i.e. these work in part by increasing the effect of other risk factors. Source: 

The risk factors are separated into non-modifiable risk factors including age, gender and family history, and modifiable risk factors including smoking, hypertension, hypercholesterolemia, diabetes, obesity, physical inactivity, and unhealthy diet. Much of the burden of CVD mortality and morbidity is linked to the modifiable risk factors (Kurian and Carderelli 2007). Moreover, it has been suggested that CVD mortality in populations of African origin is predominately attributable to hypertension and lifestyle factors including overweight and obesity, diet and physical inactivity (Gaillard 2010).
Analysis of the Health Survey for England (HSE) (Primatesta et al., 2004) data showed that compared with the Caucasian population Afro-Caribbeans smoked less, consumed less alcohol, ate more fruit and vegetables and reported higher levels of physical activity than Caucasians. In addition, Afro-Caribbeans had a more favourable lipid profile, with lower triglycerides and increased high-density lipoprotein cholesterol (HDL-C). This favourable lipid profile is thought to offer protection against CHD in people of African origin (Lip and Boos 2005). However, Afro-Caribbeans, in particular black African women, were more overweight and obese had higher prevalence of hypertension and T2DM than Caucasians (Primatesta et al., 2004). A similar trend has been reported in the USA where AAs had higher rates of obesity, poor nutrition and physical inactivity compared with the general population and consequently higher rates of CVD mortality than Caucasian Americans (Lovejoy and Masa 2010). 
Composite measures of CVD risk, such as the Framingham risk score and the metabolic syndrome, have been developed to predict CVDs through the clustering of risk factors for CHD which include, cholesterol, high blood pressure, smoking and diabetes (Alberti et al., 2006, D?Agostino et al., 2008).  However, given that the risk factors included in these measures were identified from studies in predominately Caucasian populations, their ability to identify risk in non-Caucasian groups is questionable (Sumner 2009, Osei 2010).

Therefore the aim of this study was to compare the prevalence of CVD risk factors in Afro-Caribbean and Caucasian men and women matched for age, gender and BMI. In addition, the individual contributions of the risk factors that are used in the Framingham risk and metabolic syndrome were compared and the strongest risk factor identified in Afro-Caribbeans compared with Caucasians. 

6.2 Methods

6.2.1 Subjects

Subjects were recruited from the staff and student population at a university in South East England, through advertisement on staff and student media, students? newsletter, staff and student intranet, posters, canvassing and through personal contacts. Ninety-three subjects were recruited and tested in this study. Forty-two (42) were Caucasian, 39 were Afro-Caribbean and 12 were mixed race or other ethnic group and have been excluded from this analysis. Subjects were individually matched for age, gender and BMI. After matching in was possible to include 23 Afro-Caribbean and 23 Caucasian men and women in the study
6.2.1.1 Ethical approval and selection criteria

The details of ethical approval and selection criteria are described in chapter four of this thesis.
6.2.1.2 Ethnicity

Ethnicity was self-reported, subjects self categorised according to the Office for National Statistics categories (2001) (appendix 4.1)

6.2.2 Outcome measures

The details for collection of anthropometry, calculation of BMI and percentagebody fat are described in chapter four
6.2.2.1. The Framingham risk score

Ten-year risk of CVD was calculated using guidelines from the Framingham cardiovascular risk study (D?Agostino et al., 2008). A total cardiovascular risk score was allocated for a combination of scores derived from tables given in Appendix 6.1.
6.2.2.1.1 Smoking status
Smoking status was self-reported. Subjects were classified as smokers, past smokers or non-smokers. 
6.2.2.1.2 Diabetes status
Diabetes status was self-reported. 
6.2.2.1.3 Blood pressure
Blood pressure was measured using an automated sphygmomanometer (Omron Digital Blood Pressure Monitor HEM-907, Omron Healthcare, Europe) on the left arm according to the guidelines set by the British Hypertension Society (O?Brien et al., 2003). The subject was seated for at least five minutes prior to measurement, relaxed and not moving or speaking. The arm was supported at the level of the heart with no tight clothing constricting the arm.  The cuff was placed with the indicator mark on the cuff over the brachial artery. The bladder encircled at least 80% of the arm (but not more than 100%). Two readings of blood pressure were taken and the average of the two measurements was taken.
6.2.2.1.4 Blood lipids and glucose
Finger prick samples of capillary blood (35 ?l) were analysed to measure non-fasting total cholesterol, total triglycerides, and HDL and glucose levels using an automated cholesterol meter (Cholestech LDX, Cholestech Corporation California). 


6.2.3 Statistical analysis

All statistical analysis was carried out using SPSS version 15.0 (SPSS Inc., Chicago, IL, USA). Normality was assessed using the Kolmorgorov-Sminov test. Age, total cholesterol, LDL cholesterol, triglycerides, blood glucose, BMI and waist-to-hip ratio were not normally distributed. These variables were log transformed however, LDL cholesterol and WHR remained not normally distributed therefore, the Mann-Whitney test was used to show differences between the groups for LDL cholesterol and WHR. For the remaining variables, independent samples t tests were used to compare Afro-Caribbeans and Caucasians. Multiple regression analysis was used to identify the strongest predictors of CVD risk in each population group. The dependent variable was the Framingham risk score. The individual contribution of the elements of the Framingham risk score, HDL-C, total cholesterol, systolic BP were assessed by comparing standardised beta coefficients. Age was excluded from the analysis as participants were matched for age.

6.3 Results

Subject characteristics for Afro-Caribbean and Caucasian subjects are shown in table 6.2.

Table 6. 2 Subject characteristics (mean (SD))
Afro-Caribbean (23)Caucasian (23)Age28 (10.9)28.5 (11)Height1.65 (0.10)1.66 (0.09)Weight66.3 (14.3)67.3 (13.6)BMI (kg/m2)24.2 (4.2)24.3 (4.2)%BF (BOD POD)28.6 (8.4)28.8 (11.5)%BF=percentagebody fat, BMI = body mass index
There were no significant differences in age, height, weight, BMI and percentagebody fat between the groups. Mean BMI for the participants fell within the normal range, i.e. (BMI range 18.5-24.9kg/m2).
Table 6.3 shows the fat distribution measured by WC and WHR and skinfold measurements. 

Table 6. 3 Fat distribution measured by WC and waist-to-hip ratio 
Afro-CaribbeanCaucasianMale (6)Female (17)Male (6)Female (17)WC (cm)83.5 (6.82)76.2 (13.43)84.27 (8.48)77.73 (13.3)WHR0.84 (0.05)0.77 (0.11)0.87 (0.86)0.79 (0.09)WC=Waist Circumference, WHR=waist-to-hip ratio
There were no significant differences in fat distribution between Afro-Caribbean and Caucasian men and women as measured by any of the fat distribution measures. In addition, mean WC for both groups was below the cut-off points indicative of increased risk for CVD i.e. WC > 80 cm in women and 90cm in men.
Table 6.4 shows the lipid profile, which includes total cholesterol, HDL cholesterol, LDL cholesterol and triglycerides. Also shown is blood glucose and blood pressure (systolic and diastolic). 
Table 6. 4 Blood glucose, lipid profile and blood pressure
Afro-Caribbean (23)Caucasian (23)Total cholesterol4.28 (1.02)4.35 (1.08)HDL1.42 (0.44)1.36 (0.37)LDL2.09 (0.95)2.34 (.0.83)Triglycerides1.80 (1.59)1.67 (1.46)Blood glucose4.97 (1.07)5.18 (1.05)Systolic BP118.3 (13.0)112.9 (15.8)Diastolic BP69.0 (12.4)70.0 (9.0)BP=blood pressure
There were no significant differences in blood glucose, lipids and diastolic blood pressure between Afro-Caribbean and Caucasian men and women. Moreover, none of the measures were outside of normal ranges in both groups.

6.3.1 Ten-year cardiovascular risk

Ten year cardiovascular risk was estimated using the Framingham risk equations. Eighty %s (n=15) of Afro-Caribbean and matched Caucasian women (n=15) had a less than 1% risk of CVDs in the next ten years. The remaining 20% (n=2 in each group) had a 1% risk of CVD in the next ten years.
Fifty % (n=3) of Afro-Caribbean males had a 1.1% risk of a CVD event, (n=2) had a 1.4% risk and (n=1) had 2.8% risk. Twenty % (n=1) of the Caucasian males had a less than 1% risk of CVD over the next ten years, (n=2) had a 1.1% risk and (n=2) had a 2.3% risk of CVD in the next ten years. Overall the risk of CVD in the next ten years was very low in both Afro-Caribbeans and Caucasians.


6.3.2 Predictors of CVD risk in Afro-Caribbeans compared with Caucasians

Table 6.5 shows predictors of CVD, which are components of the Framingham risk score and the metabolic syndrome. 

Table 6. 5 Predictors of Framingham risk score in Afro-Caribbeans compared with Caucasians
Dependent Variable: Framingham Risk ScoreAfro-Caribbean  (n=23)Caucasian (n=23)BSE B(pBSE B(p(Constant)-8.802.62 0.004*-2.132.36 0.4Total cholesterol0.900.2410.770.002*-0.190.180.200.3HDL-C-1.540.59-0.560.02*-1.050.51-0.440.05*Glucose0.560.210.050.3-0.80.16-0.10.6Systolic BP0.060.020.580.005*0.020.020.310.2HDL-C=high density lipoprotein cholesterol, BP=blood pressure *significantly predicts the Framingham risk score in Afro-Caribbeans

The independent variables included total cholesterol, HDL-C, blood glucose, and systolic BP. Amongst Afro-Caribbeans, total cholesterol, HDL-C and systolic blood pressure significantly predicted the Framingham risk score.  Total cholesterol had the largest beta coefficient implying that it contributed the most to the variance. Amongst Caucasians, HDL-C was only significant predictor of the Framingham risk score. 


6.4 Discussion

The aim of this study was to identify risk factors for CVD in Afro-Caribbean and Caucasian men and women. The participants were matched for age, gender and BMI, to control for the differences introduced by these variables. In terms of these measures, the groups could be considered comparable. This study showed that there were differences in the predictors of CVD in Afro-Caribbeans compared with Caucasians.  Total cholesterol, HDL-C and systolic blood pressure significantly predicted the Framingham risk score in Afro-Caribbeans but only HDL-C predicted the risk score in Caucasians. 
The finding of differences in the contribution of CVD risk factors between Afro-Caribbeans and Caucasians is expected as it has been shown that these groups are at different risk of CVDs (Gaillard et al., 2010). However, it was unexpected that total cholesterol and HDL-C predict CVD risk in Afro-Caribbeans. This finding is not consistent with other reports of CVD risk factors in Afro-Caribbean men and women, who have been reported to have a favourable lipid profile, that is characterised by low levels of total cholesterol and increased HDL-C, which offers protection against coronary heart disease (Lip and Boos 2005).
Systolic blood pressure significantly predicted CVD risk in Afro-Caribbeans but not Caucasians. Gaillard et al., (2010) in a review of studies comparing risks for CVD in African with Caucasian origin populations, reported that a greater proportion of CVD mortality in populations of African origin was attributable to hypertension and lifestyle factors including overweight and obesity, diet and physical inactivity (Gaillard 2010). 


6.4.1 Study limitations

The use of a dependent variable such as the Framingham risk score, when non-Caucasian populations are examined may be inappropriate. Composite measures of CVD risk, such as the Framingham risk score and the metabolic syndrome, have been developed to predict CVDs through the clustering of risk factors for CHD which include, cholesterol, high blood pressure, smoking and diabetes (D?Agostino et al., 2008). The Framingham risk score is designed to predict the percentagerisk of a cardiovascular event, by estimating the risk of CHD in ten years. A 10-year risk of CHD (15% is equivalent to a 10-year CVD risk (20% (D?Agostino et al., 2008). However, given that the incidence of CHD is low in people of African origin, compared with Caucasians, this risk score would underestimate the overall risk for CVD when applied to Afro-Caribbeans. Consequently, the excess mortality from CVD may be inappropriately detected in this group using this measure, largely because of the young age and group homogeneity.

Therefore, it can be assumed in the case of both the Framingham risk score and the metabolic syndrome, that the elements of these composite risk scores may contribute differently to excess CVD mortality in African origin compared with Caucasian origin populations. 

The findings of this study must be interpreted with caution because of small subject numbers. Multiple regression analysis requires that for each predictor variable entered in to the model, five times as many subjects should be available. Therefore, the number of subjects would be 25 in each group.
 

6.5 Conclusion

The aim of this study was to compare differences in the risk factors for CVD in Afro-Caribbean and Caucasians in the UK. Results showed that total cholesterol, HDL-C and systolic blood pressure predicts the Framingham risk score in Afro-Caribbeans but not Caucasians. It is recommended that this study be repeated in older populations with and without established risk factors for CVD. In addition, risk scores predicting the risk of overall CVD in populations of African origin, could be based on increased blood pressure in order not to underestimate overall CVD mortality.


6.6 References

Ageyemang, C., Bhopal, R., (2003). Is the blood pressure of people from African origin adults in the UK higher or lower than that in European origin white people? A review of cross-sectional data. Journal of Human Hypertension, 17, pp.523-534. 

Agyemang, C., Addo, J., Bhopal, R., Atkins, A.D.G., Stronks, K., (2009). CVD, diabetes and established risk factors among populations of sub-Saharan descent in Europe: a literature review. Globalization and Health 5, pp.

Agyemang, C., Bhopal, R., Bruijnzeels, M., (2005). Negro, Black, Black African, African Caribbean, African American or what? Labelling African origin populations in the health arena in the 21st century. Journal of Epidemiology and Community Health, 59, pp.1014-1018.

Alberti, K. G. M. M., Zimmet, P., Shaw, J., (2006). Metabolic syndrome?a new worldwide definition. A Consensus Statement from the International 
Diabetes Federation The Lancet, 366, pp.1059-1062.
Cappucio, F.P., Barbato, A., Kerry, S.M., (2004). Hypertension, diabetes and cardiovascular risk in ethnic minorities in the UK. The British Journal of Diabetes and Vascular Disease, 15, pp.286-293.
Carroll, J.F., Chiapa, A.L., Rodriquez, M., Phelps, D.R., Cardarelli, K.M.,  
Vishwanatha, J.K., Bae, S., Cardarelli, R., (2008). Visceral Fat, Waist circumference, and BMI: Impact of Race/ethnicity. Obesity, 16, pp.600?607.

Conway, J.M., Yanovski, S.Z., Avila, N.A., Hubbard, V.S., (1995). Visceral adipose tissue differences in black and white women. American Journal of Clinical Nutrition, 61, pp.765-771.

D?Agostino, R.B., Ramachandran, Sr., Vasan, S., Pencina, M.J., Wolf, P.A., Cobain, M., Massaro, J.M., Kannel, W.B., (2008). General Cardiovascular Risk Profile for Use in Primary Care: The Framingham Heart Study. Circulation, 117, pp.743-753.

Deurenberg, P., Yap, M., van Staveren W.A., (1998). Body mass index and % body fat: a meta-analysis among ethnic groups. International Journal of Obesity, 22, pp.1164-1171.
Gaillard, T., (2010). Insulin Resistance and CVD Risk in Black People of the African Diaspora. Current Cardiovascular Risk Reports, 4, pp. 186-194
Kurian, A.K., Cardarelli, K.M., (2007). Racial and ethnic differences in CVD risk factors: A systematic review. Ethnicity and Disease, 17, pp.143-152.

Lip, G.Y.H., Boos, C.J., (2005). Ethnic differences in arterial responses, inflammation markers and metabolic profiles: Possible insights into ethnic differences in CVD and stroke. Journal of the American Heart Association 25, pp.2240-2242.

Lovejoy, J.?C.,?and?Sasagawa, M., (2010). Obesity, Nutrition, and Physical Activity in Blacks and Whites: Implications for CVD. Current Cardiovascular Risk Reports, 4, pp.202-208.
Marfell-Jones, M., Olds, T., Stewart, A. and Carter, L., (2006). International standards for anthropometric assessment. ISAK: Potchefstroom
O'Brien, E., Asmar, R., Beilin, L., Imai, Y., Mallion, J.M., Mancia, G., Mengden, T., Myers, M., Padfield, P., Palatini, P., Parati, G., Pickering, T., Redon, J., Staessen, J., Stergiou, G., Verdecchia, P., (2003). European Society of Hypertension recommendations for conventional, ambulatory and home blood pressure measurement. Journal of Hypertension, 21, pp.821-848.
Osei, K., (2010). Metabolic Syndrome in Blacks: Are the Criteria Right? Current Diabetes Reports, 10, pp.199-208.?
Perry, A.C., Applegate, E.B., Jackson, M.L., Deprima, S., Goldberg, R.B., Ross, R., Kempner, L., Feldman, B.B., (2000). Racial differences in visceral adipose tissue but not anthropometric markers of health related variables. Journal of Applied Physiology, 89, pp.636-643.

Primatesta, P., (2004). Health Survey for England Health of Ethnic Minorities available [online] at www.doh.gov.uk .[Accessed January 2008].
Scarborough, P., Bhatnagar, P., Kaur, A., Smolina, K., Wickramsinghe, K., Rayner, M., (2010). Ethnic differences in CVD 2010 edn. British Heart Foundation Statistics Database. British Heart Foundation Health Promotion Research Group

Sumner, A.E., (2009). Ethnic differences in triglyceride levels and HDL lead to under-diagnosis of the metabolic syndrome in Black children and adults. Journal of Pediatrics, 155, pp.7-11.

Tillin, T., Forouhi, N., Johnston, D.G., McKeigue, P.M., Chaturvedi, N., Godsland, I.F., (2005). Metabolic syndrome and coronary heart disease in South Asians, African-Caribbeans and white Europeans: a UK population-based cross-sectional study. Diabetologia, 48, pp.649?656.
THE NUTRITION TRANSITION: A CASE STUDY OF ZIMBABWE

7.1 Introduction

It is anticipated that the global prevalence of overweight and obesity will increase from 1.6 billion overweight adults, including 400million obese in 2005 to 2.3 billion overweight including 700million obese by 2015 (Popkin and Gordon-Larsen 2004, Prentice 2006, WHO 2009). The majority of the increase will be accounted for almost entirely in LMCs (WHO 2009).
The increasing prevalence of overweight and obesity will be accompanied by a rise in the prevalence of non-communicable disease, particularly, CVDs (CVD) (including diabetes, and stroke) and certain cancers. In 2004, an estimated 17.1 million people died from CVD representing 31.5% and 26.8% in men and women respectively. Eighty-two % (82%) of CVD deaths occurred in LMCs with equal numbers of men and women affected. It is projected that by 2030, almost 23.6 million people will die from CVD, mainly from heart disease and stroke (WHO 2009). The largest percentageincrease will occur in these developing regions. The increase is as result of the rapidly occurring nutrition transition in these regions (Popkin 2006).

The nutrition transition (NT) describes the shifts in diet and physical activity patterns that result in lifestyle changes that promote the increased prevalence of overweight and obesity and their comorbidities (Popkin 2006). Two other transitions occur alongside the NT. The demographic transition describes changes in population age structure. Advances in medicine and public health have increased life expectancy and improved infant mortality. Thus the proportion of people above 60 years of age is projected to rise from about 7% in 2000 to 20% in 2050 and the proportion of people below the age of 15 years is projected to decline from 33% to 21% in developing countries (Schmidhuber and Shetty 2005). These changes in the population age structure are likely to impact on income growth and food consumption. A larger share of the overall population will be economically active with fewer dependents (Schmidhuber and Shetty 2005, Popkin 2009).  This potentially increases the amount of food available to each person and thus may increase the numbers of people becoming overweight and obese. The epidemiological transition, which was first described by Omran (1971) describes the changes in the types of prevalent disease. It describes the shift from a pattern of high prevalence of infectious disease associated with malnutrition, periodic famine, and poor environmental sanitation, to one of high prevalence of chronic and degenerative disease associated with over-nutrition and urban industrial lifestyles.

The nutrition, demographic and epidemiological transitions are associated with economic advancement, globalisation and for most LMCs, urbanisation (Popkin 2006).  Urbanisation is associated with the changes in diet and lifestyle, which promote overweight, obesity and chronic diseases (Popkin 2009). This chapter will describe the nutrition transition in LMCs and in particular the changes that have occurred in Zimbabwe. For the purposes of this chapter the terms LMCs and developing regions will be used interchangeably.


7.2 Description of the nutrition transition

The nutrition transition refers to changes in the composition of diet, usually accompanied by changes in physical activity (Popkin et al., 2004, 2006). Table 7.1 shows the stages and patterns of the nutrition transition.
Table 7. 1 Stages of the nutrition transition
DriversUrbanisation- economic growth, technological changes in work, leisure and food processing, and increased mass media 
StagesPattern One
Paleothic man/ hunter-gatherersPattern Two
Settlements begin/ monoculture period/ famine emergesPattern Three
Industrialisation/ receding faminePattern Four
Non-communicable diseasePattern Five
Desired societal/ behavioural changeDiet Wild plants and animals, waterCereals, waterStarchy, low variety, low fat, high fibre, waterIncreased fat, sugar and processed foods. Caloric beveragesReduced fat, increased fruit, vegetables, carbohydrate and fibre. Reduced caloric beverage Physical activityLabour intensiveLabour intensiveLabour intensive, work, job/homeShift in technology of work and leisure. Reduced physical activityReplace sedentarism with purposeful activity/ planned exerciseBody compositionLean and robustStature declinesStuntingObesity emergesReduced body fatnessPrevalent diseaseHigh infectious diseaseNutritional deficiencies emergeMother and child health deficiencies, weaning diseaseNutrition-related non-communicable disease e.g. type II diabetes mellitusReduction in nutrition related non-communicable diseaseDemographic changesLow fertility, low life expectancyHigh fertility, high mother and child mortality, low life expectancySlow mortality declineAccelerated life expectancy. Increased disabilityExtended life and healthy agingSource:  Popkin (2006)

The stages of the nutrition transition shown in table 7.1 describe the specific time. The patterns describe the historical developments within the stages and include changes in body composition, prevalent disease and demographic changes in addition to diet and physical activity changes (Popkin 2006).
Stage one describes the hunter-gather era. The diets were varied consisting of wild plants and animals, with water as the main beverage. Physical activity was high and related to the labour intensive activity of gathering food. As such humans in the hunter-gather era were lean and robust. However, modern medicine was yet to be established, therefore infectious diseases were most prevalent and fertility was low.
Stage two describes the beginning of settlements and monoculture. The diets became less varied and were mostly cereal based with water as the main beverage. The settlements were vulnerable to droughts, flood and other disasters and as such famine emerges. Physical activity was high which was related to the labour intensive production of food. All planting, harvesting and processing was done manually. As such, people in the settlement era were lean and robust. However, because of the low food variety and susceptibility to famine, stature declined, and nutritional deficiencies were dominant. 
Stage three describes a period of reduced famine and the beginning of industrialisation. The diets were starchy and high in fibre with little variety, and water as the main beverage. Physical activity was high related to the increase in manufacturing, mining and agricultural activities associated with industrialisation. These occupations were labour intensive and as such human body composition was lean and robust. However, because of the low variety in the diets, and in particular micronutrient deficiencies, stunting was prevalent.
Stage four describes the period of the emergence of non-communicable diseases. The diets were high in fats, salts, and sugars and low in fibre. Fast, convenience and highly processed foods were increasingly available and caloric beverages were widely consumed. Physical activity was reduced because of technological advances and mechanisation of work, travel and leisure activity. Moreover there was a significant shift in occupations, with sedentary occupations becoming more dominant. Consequently, the increased availability of energy and low expenditure, give rise to obesity and other chronic diseases.
Stage five describes a period of behavioural change. Conscience efforts are made to reduce fat, salt and sugar consumption and increase consumption of fruits and vegetables for fibre. In addition sedentarism is replaced with purposeful planned physical activity. Body fatness is reduced and nutrition related chronic diseases are reduced. 
The patterns of the NT as shown table 7.1, although represented sequentially may not occur in this order. In most countries it is common for two or more patterns to occur at the same time within the same region or country (Popkin 2004). In high-income countries, e.g. the UK, patterns four and five are most dominant. People of lower socio-economic status are disproportionately affected by overweight and obesity and its co-morbidities while those of higher social status are gaining the benefits of behaviour change, increased life expectancy and disability free old age (Butland et al., 2007). In developing countries, how far along a person is on the nutrition transition is determined by how far removed they are from their traditional diet and lifestyle (Popkin 2006). Thus because of the uneven development that is common in many LMCs, people in all five patterns can be found. This presents the governments of these countries with significant public health challenges (WHO/FAO 2002). 

7.3 Drivers of the nutrition transition in developing countries

The main driver of the nutrition transition in developing regions is urbanisation. In 2008, UN-Habitat reported that 50% of the world?s population lived in urban areas. It was estimated that this would grow to 70% by the year 2050. This growth accounted for almost entirely in LMCs (UN-Habitat 2008). Urbanisation in developing regions has been characterised by a growth in cities mainly through rural to urban migration but also the urbanisation of rural areas (Popkin 2000). Urbanisation is associated with economic and technological advances that affect food production and processing, changes to work, transportation and leisure activity and increases the exposure to mass media (Schmidhuber and Shetty 2005). 
Urbanisation is part of the broader process of, globalisation. Globalisation describes a process by which regional economies, societies, and cultures have become integrated through a global network of communication, transportation, and trade. The mechanisms within globalization which drive the nutrition transition include
Increases in food trade and global sourcing, 
Foreign direct investment, 
Global food advertising and marketing,
 The development of supermarkets and fast food chains,
Emergence of global agribusiness and transnational food companies, 
Development of global rules and institutions that govern the production, trade, distribution and marketing of food,
Cultural change and influence and,
Urbanisation 
(Hawkes 2006).

The following sections describe how these mechanisms through urbanisation promote the NT in developing regions.

7.3.1 Urbanisation and diet

Compared with the rural or traditional diet in developing countries, the energy in urban diets is mainly from fats and sweeteners (Popkin and Nielsen 2003). People living in urban areas eat greater amounts of animal products and a more diversified diet than their rural counterparts. The quality of urban diets may be compromised by considerably higher intakes of refined carbohydrates, processed foods, and saturated and total fat and lower intakes of fibre (Malik 2006). For the urban poor, the shift towards fast and convenience foods may also imply a shift away from fresh fruits and vegetables, pulses, potatoes and other roots and tubers (Popkin 2006).

A key element in the change in diets with urbanisation is the higher female participation in the work force. This represents a shift away from traditional time-intensive food preparation towards increased consumption of precooked convenience and fast foods and snacks (Popkin 2003). The distribution of processed, convenience foods and unhealthy snacks is often associated with the rise of supermarkets in urban areas in developing regions. Supermarkets provide a boost in the consumption of perishable foods such as meat and milk, therefore increasing the amount of animal source products consumed. In addition, they provide platforms for fast food chains (Hawkes 2006). 

Fast foods may also present a cheaper option than traditional foods. Mechanisation, technological and scientific advances in agriculture have increased the availability of food in developing countries. Most importantly, the price of food is much lower than what is was at the same level of development in the industrialised nations. Recently, Drewnoski and Popkin (2009) analysed global economic and food availability data for 1962?1994. A major shift in the structure of the global diet revealed an uncoupling of the relationship between incomes and fat intakes. Global availability of cheap vegetable oils and fats has resulted in greatly increased fat consumption among low-income nations. Consequently, the nutrition transition now occurs at lower levels of the gross national product than previously (Drewnoski and Popkin 2009).

7.3.2 Urbanisation and physical activity
 
Technological advances and mechanisation associated with urbanisation promote physical inactivity (Popkin 1999). This is mainly through a reduction in labour intensive work from the mechanisation of farming, motorisation of transport and changes in leisure activity. The urban economies in developing countries are shifting increasingly away from labour intensive agricultural and mining work, to more sedentary based occupations such as sales and clerical work. These occupations require less physical exertion and allow more leisure. Indeed, leisure activity has also been transformed, particularly by changes in food preparation, production, and processing and by the penetration of the mass media into the developing world (Popkin 2009).  Mass media represents sedentary leisure through television watching, listening to the  radio or reading newspapers.

7.4 Consequences of nutrition transition 

The main consequence of the nutrition transition in developing countries is the increase in overweight and obesity prevalence (Prentice 2006). However, limited amounts of prevalence data exist in many developing regions. This is due to irregular population surveys in the regions such as Sub-Saharan Africa. However, Mendez et al., (2005) investigated the nutritional status data of urban and rural women in 36 LMCs, 19 of which are in Sub-Saharan Africa. Surveys from the US Agency for International Development (USAID) were used with the exception of the data from China and Mexico, whose surveys were conducted by their respective governments. The prevalence of overweight in urban women ranged from 10.3% - 69.9% with a median of 32.4%. Overweight prevalence was >20% in 33 countries and >50% in 10 countries. Rural overweight ranged from 3.6%-65.9% (median 19.4%) and in 18 countries rural overweight prevalence was >20%.  A marked increase in overweight was associated with a gross national income (GNI) of above US$3000 and a level of urbanisation above 32%. Overall, it was found that a greater number of women were overweight than were under weight in both urban and rural settings in these middle- and low-income countries (Mendez et al., 2005). However, these studies were limited to women between the ages of 15-49 years as data in women are more readily available than that in men. 


Abubakari et al., (2007) reviewed 36 studies examining obesity prevalence and trends in the West African countries of Benin, Burkina Faso, Cape Verde, Cote d?Ivoire, Gambia, Ghana, Guinea, Guinea-Bissau, Liberia, Mali, Mauritania, Niger, Nigeria, Senegal, Sierra Leone and Togo. The studies included both men and women, and therefore give a more complete picture of obesity prevalence in West Africa. The mean BMI range was from 20.1-27.0kg/m2; the prevalence of obesity was estimated at 10%. Women were more likely to be obese than men, and urban residents more obese than rural residents. Time trend analysis indicated that obesity in urban West Africa has more than doubled in the last 15 years; women accounted almost entirely for the increase (Abubakari et al., 2007).

7.4.1 Double burden of malnutrition and disease

In many developing countries, it is common to find both forms of malnutrition in the same household (overweight mother and stunted child) or indeed in the same person (overweight and stunted child) (Popkin et al., 1996, WHO/FAO 2002). This is directly attributable to the consumption of high calorific but micronutrient deficient foods (Popkin 1996). These low quality diets associated with urbanisation, result in lowered immunity to infectious diseases (Hawkes 2006). 
Poor diet quality resulting in malnutrition may act as an independent risk factor for a future increase in CVDs. The thrifty genotype hypothesis (Neel 1962) and the thrifty phenotype hypothesis (Hales and Barker 1992, 2001) are two key hypotheses proposed to explain the effect of foetal and infant nutrition on the development of future CVD.
Neel in 1962 put forward the ?thrifty genotype? hypothesis, which states that some individuals are genetically predisposed to a thrifty genotype, making them exceptionally efficient in the intake/utilisation of food. The thrifty genotype is thought to have offered a survival advantage to individuals in the hunter-gatherer and early agricultural societies (stages one and two of the nutrition transition) who were subject to periods of nutritional hardship, by favouring fat deposition during periods of abundance. However, in modern times of constant abundance, this genotype is disadvantageous. Instead, the thrifty genotype is associated with increased overweight, obesity and CVD, in particular type II diabetes mellitus (Prentice 2005).
Central to the thrifty gene hypothesis is the assumption that the mechanism behind the increase incidence of diabetes is the alteration in the insulin-signalling pathway, which allowed for rapid fat deposition during times of abundance. During times of famine, reduced fecundity was noted in both men and women and thus those able to conceive passed on this thrifty genotype to their offspring (Neel 1962).
This concept of altered insulin signaling is the basis for the thrifty phenotype hypothesis proposed by Hales and Barker (1992). It was proposed that poor foetal and infant nutrition cause poor development of pancreatic beta cell mass and function leading to insulin resistance. Thus, the foetus is ?programmed? for survival if the nutritional environment remains the same. However, the nutritional environment rarely stays the same, therefore predisposing individuals to obesity, type II diabetes and CVD (Prentice 2008).
Thus, the burden of CVDs are likely to increase in many developing regions who are undergoing the nutrition transition whilst a large burden of undernutrition remains.


7.5 Zimbabwe

Zimbabwe is a low-income developing country in Southern Africa. It is bordered by Zambia, Mozambique, South Africa and Botswana as shown in figure 7.1. 


Figure 7. 1 Map of Zimbabwe

Zimbabwe is a landlocked country of 390 000km2. It is part of a great plateau, which constitutes the major feature of the geology of southern Africa. Almost the entire surface area of Zimbabwe is more than 300m above sea level, with nearly 80 % of the land lying more than 900m above sea level and about 5 % lying more than 1,500m above sea level. 

7.5.1 The population

In 2008, the Zimbabwean population was estimated at 13,228,000. Figure 7.2 shows the population distribution pyramid

Figure 7. 2 Zimbabwean population pyramid (2005-2006)

In 2005, the majority of the Zimbabwean population was below the age range of 15-19 years. This distribution is typical of many low-income developing countries, where the population is young, and households have a large numbers of dependants. The Zimbabwean population is growing at an annual rate of approximately 0.8% compared with an annual growth rate of 2.2% for the rest of Africa (WHO 2008). One the characteristics of the demographic transition in developing countries is the slow down in population growth (Popkin 2009). Table 7.2 shows the population size and annual growth rate in Zimbabwe from 1901 to 2002.
Table 7. 2 Annual population growth rates between 1901-2002
YearPopulation (000)Annual growth rate1901713-19119072.4192111472.4193114642.5194120063.2195128923.5196139693.5196951343.3198276083.01992104123.12002116321.1(Source: Central Statistics Office (2002) and Zimbabwe Demographic and Health Survey 2005-6)
The population growth rate peaked in 1961 and there has been a progressive decline since. If the population growth rate continues to decline, the population pyramid will shift with larger numbers of older people and fewer younger people, which is similar to the population pyramids of developed countries. 
However, crude death rates have increased and life expectancy has decreased in Zimbabwe over the last 20 years. Table 7.3 shows a comparison of the birth, death and life expectancy in 1992 and 2002 census years.


Table 7. 3 Birth, death rates and life expectancy in Zimbabwe 
1992 census2002 censusCrude birth rate, births per 1000 of the population34.530.3Crude death rate, births per 1000 of the population9.517.2Life expectancy61.045.0(Source: Zimbabwe Demographic and Health Survey 2005-6)
The crude death rate has increased from 9.5 to 17.2 and life expectancy has declined from 61 years to 45 years from 1992 to 2002 respectively. In 2008 life expectancy for Zimbabwean adults was 44 years in males and 43 years in females (WHO 2008). The increased death rate and low life expectancy reflects the effects of the Human Immunodeficiency Virus and the Acquired Immunodeficiency Syndrome (HIV/AIDS) epidemic in Zimbabwe. The majority of deaths in adults between the ages of 15-49 years can be attributed to HIV/AIDS. AIDS, unlike other diseases, affects people in the most productive age groups i.e. between the ages of 15-49 years. This has large effects on the population structure and on the economy (Bollinger et al., 1999).

7.5.2 The economy

Mining and agriculture are the backbone of the Zimbabwean economy. In particular, agriculture provides approximately 70% of the employment within the country. Moreover, the large-scale commercial farming sector is the major source of food production and security. The main agricultural export products are tobacco, maize, cotton, sugar, and groundnuts. As such, food insecurity was low (Nkungula and Harris 2005). 

The epidemic of HIV/AIDS has had a large effect on the agricultural sector of Zimbabwe and thus the country?s food security. This is mainly through the loss of labour and the shrinking of the number of economically active individuals (Bollinger et al., 1999). The country?s ability to provide food for the populous has been compounded by adoption of inappropriate government policies.
In the decade following independence in 1980, Zimbabwe?s economic growth was strong, and living standards improved significantly. In the late 1990s, Zimbabwe?s economic growth began to slow down. In 1990 Economic Structural Adjustment Programme (ESAP) was introduced into the Zimbabwean economy. The ESAP was associated with the removal of controls on a wide range of commodities, including consumer prices of staple foods and liberalization of markets.  In the short term this caused sharp increases in prices of food, falling real incomes especially among the poor for whom the staple diet compromised up to 50% of total intake. It was expected that in the future, growth in farm productivity and nonfarm employment would reduce the food price (Jayne and Rubey 1992).  However, this was prior to the effect of the HIV/AIDS epidemic, which affected farm productivity. In addition, Zimbabwe experienced droughts in 1991 and 1992 which, significantly affected the availability of food, and the country and in particular rural areas where food aid was required (World Bank 2010).

On the 14th of November 1997 (now called Black Friday) the Zimbabwean dollar lost 71.5 per cent of its value against the US Dollar and the stock market crashed, wiping off 46 per cent from the value of shares as external investors lost confidence in the currency. Since 1999, Zimbabwe?s economic conditions continued to deteriorate, reaching a critical level as hyperinflation reached 231 million %/year in July 2008 (Zikhali 2008). The collapse in the economy and hyperinflation resulted in wide spread food shortages, mass emigration of skilled professionals, which affected education and health sectors most severely.  In 2008 the Zimbabwean dollar was abandoned in favour of a multi-currency system, including the US dollars and the South African rand. This has brought some stability to the economy but is yet to revive the crucial agricultural sector. Currently, a large number of food products are imported into the country. This has brought greater variety to the Zimbabwean diet, most importantly, increased the availability of processed, convenience and fast foods. This is most readily seen in urban areas, although these foods are penetrating some rural areas.


7.5.3 Urbanisation

In 2001, 36% of Zimbabwe population was urbanised and growing at an annual rate of four %. Slum dwellers accounted for three % of the urban population (UN-Habitat 2008). Urbanisation in Zimbabwe is facilitated by rural to urban migration and to a lesser extent urbanisation of rural areas. The following sections describe changes in diet and physical activity patterns in Zimbabwe associated with urbanisation.


7.5.4 Evidence of dietary transition in Zimbabwe

The staple diet in Zimbabwe consists of cereals, namely maize meal, vegetables mainly green leaves cooked with tomato and peanut butter, and meat when available. During times of scarcity, households may reduce the number of meals and the quality of the meal to only sadza and relish or rape (UNICEF Summary, 1999).

Food balance sheets from Food and Agricultural Organisation have been used to analyse the changes in diet and energy intake in Zimbabwe. Data were available for 1961-2007. Table 7.4 shows the different sources of dietary energy in Zimbabwe between 1961 and 2007. The figures represent energy in kcal/capita/day.Table 7. 4 Food supply in Zimbabwe between 1980 and 2007 (kcal/capita/day)
YearCerealsStarchy rootsSugar and sweetenersPulsesOil cropsVegetable oilsVegetablesFruitMeatAnimal FatsMilk excluding butter196115093998447837121879189719651461351443913154121756158619701498331693415785121562398019751445282012812710312156149951980115128171261031421116634587198512813222248721499154944681990124431216367.51559165751571995118741211315516871442543120001103452563265201714624650200512054927154912261012824145200712805026248892101017854645Cereals ?wheat, rice, maize, soghum, millet.  Starchy roots- cassava, potatoes, sweet potatoes.  Sugar and sweeteners-sugar, non-centrifugal, sugar(raw), other sweetners. Pulses-beans, peas, other. Oil crops-soya beans, groundnuts, sunflower seed, rape and mustard seed, cottonseed, olives, other. Vegetable oils- oils from oil crops. Vegetables- tomatoes, onions, other.  Fruit-oranges, apples, pineapples, grapes and other.  Meat-bovine meat, mutton and goat, pork, poultry and other meat. Animal fat- butter, ghee, cream, raw animal fats, fish oil, fish liver oil
Source: Food and Agricultural Organisation (FAO) (2010)Figure 7.3 shows the individual groups plotted against the year

Figure 7. 3 Food groups plotted against year (Source: FAO 2010)

The consumption of cereals fluctuated between 1961 and 2007, although progressively decreasing. The consumption of sugar and sweetners, and vegetable oils has increased. The increase of these two is a feature of the nutrition transition in many developing countries ( Popkin 2006). Energy from all the products appeared to remain at a steady state, with the exception of oil crops which have declined.
Figure 7.4 shows total energy intake in kcal/capita/day for the years 1961-2007 in Zimbabwe.

Figure 7. 4 Total energy availability (kcal/capita/day) from 1961-2007 in Zimbabwe
(Source FAO 2010)

It can be seen that energy fluctuated greatly between the years 1961 to 2007. From 1961, energy intake increased and peaked in 1970 before declining and increasing to its highest point in 1985; energy intake was lowest point in 1995. This low point represents the years when Zimbabwe experienced droughts in two consecutive farming seasons, which affected the production of cereal crops from which the population derives most of its energy. After 1995 energy intake rose.
The mean total energy intake represented in figure 7.4, although increasing, is below the recommended values for men and women for 2500 and 2000 kcals per day (FAO/WHO/UNU 1985). However, these data represent combined data for rural and urban populations. Therefore, this not a reliable estimate of the changing energy intake in the Zimbabwean population. It is known that the diets of rural and urban populations differ in urban Africa. For example, in South Africa, urban populations consume greater amounts of dietary fat, salt and sugar compared with the rural population, consequently, the energy intake is much higher (Bourne et al., 2002).

7.5.5 Evidence of changes in physical activity

A reduction in physical activity, in most developing countries is associated with changes in leisure, transportation and work. Table 7.5 shows levels of physical inactivity associated with work, leisure and transportation in the Zimbabwean population.

Table 7. 5 Levels of physical inactivity in leisure, transport and work in Zimbabwean adults, in rural and urban areas
Leisure time inactiveTransport inactiveWork inactiveMaleFemaleMaleFemaleMaleFemale% (inactive)9597.413.522.27281.8 Source: Zimbabwe Non-Communicable Disease Risk Factors, preliminary report (2005)

The majority of the population sampled in this survey were physically inactive in leisure and work, and were least physically inactive in transportation. However, these data represent combined rural and urban physical inactivity.  Therefore, differences in physical inactivity associated with urbanisation are not apparent.
Economic advancement in many developing countries has accelerated the service sector, and capital-intensive processes have come to dominate industrial production, (Popkin 2000). These changes are most apparent in urban areas. Table 7.6 shows the occupation in Zimbabwean adults in rural and urban areas.


Table 7. 6 Difference in occupation between rural and urban men and women (%)
ResidenceGenderProfessional/
Technical/ ManagerialClericalSales and servicesSkilled manualUnskilled manualDomesticAgricultureFemaleUrban11.18.645.612.70.513.76.2Rural4.41.219.57.91.89.155.5MaleUrban14.05.025.833.59.66.23.0Rural3.00.68.112.87.09.156.6Source: Zimbabwe Demographic and Health Survey (2005/06)

Among urban men, the most common occupations are skilled manual labour (34% and sales and services (26%). Urban women are most often employed in sales and services (46%). In rural areas, more than half of women (56%) and men (57%) are employed in agriculture. There is a clear divide between occupation in rural and urban populations, with agriculture the main occupation in rural men and women.  Agriculture in rural Zimbabwe is mostly subsistence farming where mechanisation is limited and is labour intensive. 

Leisure activity has been changed, particularly by changes in food preparation, production, and processing and by the revolutionary penetration of the mass media into the developing world (Popkin 2000). Exposure to mass media including television watching, access to newspapers and radio can act as a proxy measure of not only information penetration but also use of leisure time activity. Table 7.7 shows the exposure to mass media in rural compared with urban populations


Table 7. 7 Exposure to mass media in rural and urban Zimbabwe
ResidenceGenderReads a newspaper at least once a weekWatched television at least once a weekListens to radio at least once a weekAll three at least once a weekNo media at least once a weekFemaleUrban48.977.977.437.88.8Rural8.69.228.62.366.4MaleUrban71.981.083.856.24.5Rural19.218.751.97.641.6Source: Zimbabwe Demographic and Health Survey (2005/06)

Twenty-five % (25 %) of women and 40 % of men read newspapers at least once a week, 36 % of women and 44 % of men watch television at least once a week, and 48 % of women and 64 % of men listen to the radio at least once a week. Generally, urban residents and men are more likely to be exposed to all forms of mass media than rural residents and women. Sixty-six % (66 %) of rural women, 9 % of urban women, 42 % of rural men, and 5 % of urban men reported having no exposure to any form of mass media.

The data show that there have been changes in the diet and physical activity in Zimbabwe. The changes in diet have been affected by several factors as described above, thus there is an inconsistent trend in energy intake. Physical inactivity appears to have increased mostly in work and leisure sectors. In addition there is a clear urban divide in mass media. However, the data are confounded by the joining together of rural and urban populations.


7.5.6 Double burden of disease in Zimbabwe

The burden of disease in Zimbabwe over the past twenty years has been dominated by the epidemic of HIV/AIDS, which is the leading health concern in the country.  Other infectious diseases, perinatal and nutritional disorders are of major importance and non-communicable diseases are becoming increasingly important (Zimbabwe Non-Communicable Disease Risk Factors, preliminary report (ZiNCoDs) 2005). Figures 7.5 and 7.6 show the causes of death in adult Zimbabwean men and women in 2004.

Figure 7. 5 Causes of death in Zimbabwean males (2004) 


Figure 7. 6Causes of death in Zimbabwean females (2004), 
Source: Mathers et al., (2004)

In both males and females, infectious and parasitic diseases account for 68.2 and 70.2 % of deaths respectively.  The large burden of infectious disease represents the effect of HIV/AIDS on the population. In 2003, the prevalence of HIV/AIDS was estimated at 24.6 % in adults aged 15-49years. Recent figures show a reduction with an estimated prevalence of 15.3% (14.6% - 16.1%) in adults? aged 15 to 49 years (UNAIDS 2008). HIV/AIDS remains most prevalent in adults between the ages of 15-49 years of age.  
Non-communicable diseases accounted for 15.8 % and 16.8 % for deaths in men and women respectively. Of the non-communicable diseases, circulatory disease, or CVDs accounted for the largest number of deaths in both men and women. The availability of data on non-communicable disease is limited. However, in a survey of admissions to the medical wards at United Bulawayo Hospitals in the city of Bulawayo, five of the top ten diseases were non-communicable diseases with three of these being CVDs (Mudiayi et al, 1997). In a review of available surveillance data for Harare it was shown that persons aged 45-64 years experience a relatively high mortality from hypertensive sequelae, but there was a low mortality from ischaemic heart disease (Razum, 1997). These available data suggest an increasing prevalence of CVD in Zimbabwe.
7.5.6.1 Selected risk factors for CVD in Zimbabwe

Risk factors for CVD associated with the nutrition transition include overweight and obesity, increased blood pressure, increased blood cholesterol and increased prevalence of diabetes. 
7.5.6.1.1 Overweight and obesity in Zimbabwe
Table 7.8 shows the prevalence of overweight and obesity in the two survey years, of the Zimbabwe demographic and health survey.

Table 7. 8 Overweight and obesity prevalence in Zimbabwean adults (%)
199920052005BMI (kg/m2)Rural and urban Rural and Urban RuralUrban Underweight (<18.5)5.66.84.26.8Normal (18.5-24.9)59.965.871.257.7Overweight (25-29.9)2717.8	13.724.0Obese (<30)7.57.24.26.8Source: Zimbabwe Demographic and Health Survey 1999 and 2005 

Between 1999 and 2005, underweight increased from 5.6 to 6.8 % in rural and urban populations combined. Normal weight increased from 59.9 to 65.8 %, Overweight declined from 27 % to 17.8 % and obesity declined from 7.5 % to 7.2 %. In 2005, the data available shows differences prevalence between in urban and rural populations.  Apart from normal weight, underweight, overweight and obesity were higher in the urban compared with the rural population. This data is limited to females between the ages of 15-49 years, and as such does not give a representative picture of the prevalence of overweight and obesity for the population as a whole. However, it is useful data as it gives a picture of female nutrition in the country. Female nutrition can affect the future of nutrition related non-communicable diseases (Barker 2004).
7.5.6.1.2 Hypertension

Table 7.9 shows mean blood pressure (BP) and the prevalence of hypertension in Zimbabwean men and women.

Table 7. 9 Blood pressure (BP) and prevalence of hypertension in adults aged 25-100years, combined rural and urban data
MaleFemaleMean Systolic BP (mmHg)131.1133.2Mean Diastolic BP (mmHg)81.783.9Raised BP* (%)2329Raised BP** (%)8.513.3*BP >140/90mmHg, **BP>160/95mmHg or on anti-hypertensive medication 
Source: Zimbabwe Non-Communicable Disease Risk Factors, preliminary report (2005)

The mean systolic and diastolic blood pressures for both men and women are not indicative of raised blood pressure. However, 23% of men and 29% of women had raised BP as defined by BP>140/90 and 8.5% and 13.3% of women had raised BP defined as BP>160/95 or taking antihypertensive medication. Clinical studies suggest that there is a high prevalence of hypertension and its sequalae in Zimbabwe (Matenga et al, 1986).  In 100 consecutive cases of stroke studied, 53% had hypertension, 50% of whom had defaulted treatment while the other 50% were newly diagnosed. In a study of hypertension awareness in communities with different levels of socio-economic development only 26% of hypertensives were aware of their elevated blood pressure status (Matenga et al, 1997). In another study, 66% of patients in a geriatric unit had a diagnosis of hypertension (Wilson and Nhiwatiwa, 1992). 


7.5.6.1.3 Diabetes prevalence in Zimbabwe

Table 7.10 shows the prevalence of diabetes in Zimbabwe, defined as fasting blood glucose of (7.8mmol/l


Table 7. 10 Diabetes prevalence in Zimbabwean men and women (2005) (%)
MaleFemaleFasting blood glucose (7.8 mmol/l and/or                              oral glucose (11.1 mmol/l 2.21.3Fasting glucose blood sample, blood value (mmol/l) (not specified, whole blood or plasma)9.810.2Mean blood glucose mmol/l5.35.3Source: Zimbabwe Non-Communicable Disease Risk Factors (ZiNCoDs) preliminary report (2005)

The prevalence of diabetes by these definitions is low in Zimbabwean men and women. However, a difference in methodology may confound these findings. Using a fasting blood glucose definition of diabetes mellitus with a cut-off level of > 7mmol/l gave a higher prevalence of diabetes compared to the use of the oral glucose tolerance test. However, the mean glucose fell within the normal ranges.


7.5.6.1.4 Cholesterol and levels of raised total cholesterol in Zimbabwe

Table 7.11 shows mean total cholesterol and the prevalence of raised cholesterol.

Table 7. 11 Cholesterol level and prevalence of raised cholesterol, 25-100years
MaleFemaleMean total cholesterol (mmol/l)4.34.3Raised cholesterol ? 5.2 mmol/l (%)20.421.4Raised cholesterol ? 6.5 mmol/l (%)3.14.8Source: National Survey Zimbabwe Non-Communicable Disease Risk Factors (ZiNCoDs) Preliminary Report

Twenty % (20.4 %) of men and 21.4 % of women had raised cholesterol d20.421.4Raised cholesterol ? 6.5 mmol/l (%)3.14.8Source: National Survey Zimbabwe Non-Communicable Disease Risk Factors (ZiNCoDs) Preliminary Report

Twenty % (20.4 %) of men and 21.4 % of women had raised cholesterol defined as total cholesterol ? 5.2 mmol/l, and 3.1 % of men and 4.8 % of women had a raised cholesterol (? 6.5 mmol/l). The general impression is that lipid disorders are uncommon among black Zimbabweans (Castle, 1982). This is a not unique to black Zimbabweans but is consistently reported in populations of African origin compared with Caucasian populations (Gaillard et al., 2009). 
The data above show that overweight and obesity, hypertension and raised cholesterol (? 5.2 mmol/l,) had similar prevalence (within a range of 20-29 % prevalence). However, th


7.6 Conclusion 

The prevalence of overweight and obesity is increasing globally. In developing countries, the nutrition transition has been linked to the increasing prevalence of overweight, obesity and its comorbidities. Using Zimbabwe as an example, it has been shown that demographic, epidemiologic, dietary and physical activity shifts are occurring. Thus, the country needs to prepare to deal with the co-existence of non-communicable and the communicable diseases. 


7.7 References

Abubakari, A.R., Lauder, W., Agyemang, C., Jones, M., Kirk, A., Bhopal R. S., (2008). Prevalence and time trends in obesity among adult West African populations: a meta-analysis. Obesity Reviews, 9,?pp.297???311.

Barker, D.J.P., (2004). The developmental origins of adult disease. Journal of the American College of Nutrition, 23, pp.588S-595S
Bollinger, L., Stover, J., Kerkhoven, R., Mutangadura, G., Mukurazita, D., (1999). The economic impact of AIDS in Zimbabwe.  The Futures Group International in collaboration with: Research Triangle Institute (RTI) The Centre for Development and Population Activities (CEDPA) [online] available at: http://psufasa.org/map_info/full%20documents/country%20documents/Zimbabwe/eco/The%20Economic%20Impact%20of%20AIDS%20in%20Zimbabwe.pdf [Accessed on July 2010]

Bourne, L.Y., Lambert, E.V., and Steyn, K., (2002) Where does the black population of South Africa stand on the nutrition transition? Public Health Nutrition: 5, pp.157?162.

Butland, B., Jebb, S., Kopelman, P., McPherson, K., Thomas, S., Mardell, J., Parry, V., (2007) Foresight- Tackling obesities: future choices project report.

Castle, W. M., (1982). Coronary heart disease risk factors in black and white men in Zimbabwe and the effect of living standards.  South African Medical Journal 61, pp.926-929. 

Chopra, M., (2004). From apartheid to globalization: Health and social change in South Africa. Hygiea Internationalis, 4, pp.153-174.
Doak, C., Adair, L., Bentley, M., Fengying, Z., Popkin, B., (2002). The under weight/over weight household: an exploration of household socio-demographic and dietary factors in China. Public Health Nutrition, 5, pp.215?221.
Drewnowski, A., Popkin, B.M., (2009). The Nutrition Transition: New Trends in the Global Diet. Nutrition Reviews, 55,?pp.31???43.

Ferro-Luzzi, A., Sette, S., Franklin, M., James, W.P.T., (1992). A simplified approach of assessing adult chronic energy deficiency. European Journal of Clinical Nutrition, 46, pp.173-186.
Food and Agriculture Organization /World Health Organization /United Nations University (1985). Energy and protein requirements. Geneva: World Health Organization 
Food and Agriculture Organization (2010). Food balance sheets, [online] available at http://faostat.fao.org/site/368/DesktopDefault.aspx?PageID=368#ancor [Accessed August 2010]
Gaillard, T., Schuster, D., Osei, K., (2009). Metabolic syndrome in black people of African diaspora: The paradox of current classification, definition and criteria. Ethnicity and Disease, 19, pp.S2-1 ? S2-7.
Hales, C.N., Barker, D.J.P., (1992). Type 2 (non-insulin-dependent) diabetes mellitus: the thrifty phenotype hypothesis. Diabetologia, 35, pp.595?601.

Hales, C.N., Barker, D.J.P., (2001). The thrifty phenotype hypothesis. 
British Medical Bulletin, 60, pp.5?20.

Hawkes, C., (2006). Uneven dietary development: linking the policies and processes of globalization with the nutrition transition, obesity and diet-related 
chronic diseases. Globalization and Health, 2, pp4.
Jayne, T.S. and Rubey, L., (1993). Maize milling, market reform and food security: The case of Zimbabwe. World development, 21,  pp.975-988,
Matenga, J., Kitair, I., Levy, L., (1986). Stroke among black people in Harare, Zimbabwe: Results of computed tomography associated risk factors. British Medical Journal, 229, pp.1649-1651.

Matenga, J., (1997). ?Stroke Incidence Rates among Black Residents of 
Harare?A Prospective Community-Based Study.? South African Medical Journal, 87, pp.606?609. 

Mathers, C. D., Bernard, C. Iburg, K. M., Inoue, M., Ma Fat, D., Shibuya, K., Stein, C., Tomijima, C., Xu, H., (2004). Global Burden of Disease: data sources, methods and results. [online] available at: http://www.who.int/healthinfo/bod/en/index.html  [Accessed July 2010].
Mendez, M. and Popkin, B.M., (2004). Globalization, urbanization and nutritional change in the developing world. In FAO. Globalization of food systems in developing countries: impact on food security and nutrition, pp.55-80. Food and Nutrition Paper No. 83. Rome, FAO
Mudiayi, T.K., Onyanga-Omara, A., and Gelman, M.L., (1997).  Trends of morbidity in general medicine at united Bulawayo Hospitals, Bulawayo, Zimbabwe.  Central African Journal of Medicine, 43, pp.213-214. 
Neel, J.V., (1962). Diabetes mellitus: A ?thrifty? genotype rendered detrimental by ?progress? .The American Journal of Human Genetics, 14, pp.353-362.

Neel, J.V., (1989). Update to 'The Study of Natural Selection in Primitive and Civilized Human Populations'. Human Biology, 61, pp.811?882.

Nkungula, A., and Harris, E. W., (2005). Foods frequently eaten by high school students in a high density area in Zimbabwe. Ecology of Food and Nutrition, 44, pp.135 ? 145.

Popkin, B.M., (1999). ?Urbanization, Lifestyle Changes and the Nutrition Transition,?. World Development, 27, pp.1905?1916.

Popkin, B.M., (2000). Achieving urban food and nutrition security in the developing world: Urbanisation and the nutrition transition. Focus, 7, (10).

Popkin, B.M., (2003). The nutrition transition in the developing world. Developmental Policy Review. 21, pp.581-597.

Popkin, B.M., (2006). Global nutrition dynamics: the world is shifting rapidly toward a diet linked with noncommunicable diseases. American Journal of Clinical Nutrition, 84, pp.289 ?298. 

Popkin, B. M., (2009). The Nutrition Transition in Low-Income Countries: An Emerging Crisis. Nutrition Reviews, 52,?pp.285???298.

Popkin, B.M., Nielson, S.J., (2003). The sweetening of the world?s diet. Obesity Research, 11, pp.1325-1332.
Popkin, B.M., and Gordon-Larsen, P., (2004). The nutrition transition: worldwide obesity dynamics and their determinants. International Journal of Obesity, 28, pp.S2-S9.
Prentice, A.M., (2005). Starvation in humans: Evolutionary background and contemporary implications. Mechanisms of Ageing and Development,
126, pp.76-98.

Prentice, A. M., (2006). The emerging epidemic of obesity in developing countries. Intenational Journal of Epidemiology, 35, pp.93-99.
Prentice, A.M., Hennig, B.J., Fulford, A.J., (2008). Evolutionary origins of the obesity epidemic: natural selection of thrifty genes or genetic drift following predation release? International Journal of Obesity, 32, pp.1607-1610.

Razum, O., (1997). Monitoring CVDs in Zimbabwe: a review of needs and options.  Central African Journal of Medicine, 42, pp.120-124. 
Schmidhuber, J., and Shetty, P., (2005). The nutrition transition to 2030. Why developing countries are likely to bear the major burden. Food Economics - Acta Agriculturae Scandinavica,  2, pp.150 ? 166.
UNAIDS (2008). Epidemiological fact sheet on HIV and AIDS. [online] available at: http://www.unaids.org/en/CountryResponses/Countries/zimbabwe.asp [Accessed July 2010].
Useh, U., Mbajiorgu, F.E., Feresu, S.A., Madzivire, D., (2000) Preliminary study of body mass index and its relationship to percent body fat in Harare, Zimbabwe. Central African Journal of Medicine. 46, pp.101-105.
Wilson, A.O., and Nhiwatiwa, R., (1992). Does Zimbabwe need geriatric services? Central African Journal Medicine 38, pp14-16.
World Bank (2010). Country profiles : Zimbabwe. [online] available at http://web.worldbank.org/WBSITE/EXTERNAL/COUNTRIES/AFRICAEXT/ZIMBABWEEXTN/0,,menuPK:375742~pagePK:141159~piPK:141110~theSitePK:375736,00.html [Accessed August 2010].
World Health Organization and Food and Agricultural Organization (2002) Diet, nutrition and the prevention of chronic diseases; Report of a joint WHO/FAO expert commission WHO Technical Report Series 916. Geneva: WHO
World Health Organization (2006). Obesity fact sheet.  [online] Available at http://www.who.int/mediacentre/factsheets/fs311/en/index.html [Accessed March 2007].
World Health Organisation (2009). CVD fact sheet.  [online] Available at http://www.who.int/mediacentre/factsheets/fs317/en/index.html [Accessed January 2010]
Zikhali, P., (2008). Fast track land reform and agricultural productivity in Zimbabwe. Environment for Development, Discussion papers series pp.8-30.

Zimbabwe demographic and health survey (2005/06). [online] available at www.measuredhs.com/pubs/pdf/FR186/FR186.pdf  [Accessed November 2008].

Zimbabwe non-communicable disease risk factors, preliminary report (2005).Minstry of Health and Child Welfare, University of Zimbabwe. [online] available at http://www.who.int/chp/steps/STEPS_Zimbabwe_Data.pdf [Accessed August 2010]
COMPARISON OF ANTHROPOMETRIC AND DEMOGRAPHIC PROFILES OF A GROUP OF URBAN AND RURAL ZIMBABWEANS

8.1 Introduction

A limited amount of data is available on the changes associated with the nutrition transition in many developing countries including Zimbabwe, so anthropometric measures can provide valuable information on overweight and obesity prevalence. In Southern African countries, overweight and obesity have been associated with being female and living in urban areas (Walker et al., 2001). Voster et al, (2005) compared levels of overweight and obesity in five groups of men and women including rural residents, farm workers, informal settlement dwellers, urban Africans living in townships and urban African professionals in South Africa who participated in the Transition and Health during Urbanisation of South Africans (THUSA) study. The five groups represented different levels of urbanisation; the rural and farm dwellers being least urbanised, the informal settlement dwellers representing populations who are in transition between rural and urban; and the township and urban professionals Africans were the most urbanised. It was shown that urbanisation had a significant impact on overweight in men; BMI increased from 20.7 kg/m2 in the rural men, 20.6 kg/m2 in farm dwellers and 20.3kg/m2 in informal settlement dwellers, to 21.3 kg/m2 in African townships and 23.1 kg/m2 in urban professionals. Among women, rural women had a mean BMI of 25.6 kg/m2, farm dwellers 26.3 kg/m2, informal settlement dwellers 26.7 kg/m2, in African townships 28.0 kg/m2 and 28.1 kg/m2 in urban professionals. This study showed a clear progression in obesity with urbanisation and that females were at greater risk.

In Zimbabwe, a limited amount of data is available on the prevalence of overweight and obesity. Early studies from Zimbabwe, which included both men and women, reported contrasting results on the prevalence of overweight and obesity. Mathe et al., (1985) investigated the nutritional status of men and women in an urban black township of Zimbabwe (Luveve, Bulawayo). The heads of 70 households were interviewed and the mid upper arm circumferences of their families were measured. On the basis of arm circumference measurements the population was not classified as underweight and obesity was observed in 50% of the adults and many preschool children. However, these findings were limited to a small sample of urban residents. In contrast, Zinyowera et al., (1994) sampled a large population of 775 men and 735 women in a nationally representative sample of rural and urban populations. It was reported that the prevalence of obesity was low (0%) in this population, however on average women were overweight and had larger mean WCs in both rural and urban areas. The authors concluded that obesity was not a problem in Zimbabwe and that overweight was limited to urban women. These study findings were reinforced by Ushe, et al., (2000) who documented the BMI and anthropometric characteristics of 140 healthy adults living in urban Harare. The mean BMI value indicated that males were of normal weight (BMI = 24.9 kg/m2) while females were overweight (BMI = 27.8 kg/m2) and no obesity was reported. In addition, women were found to have higher mean percentagebody fat, and skinfold thicknesses than their male counterparts. More recently, in the last Zimbabwe Demographic and Health Survery (ZDHS 2005/06) the prevalence of overweight was 13.7% in rural and 24 % in urban women; and the prevalence of obesity was 4.2 % in rural and 6.8 % in urban women. The findings from these studies indicates that overweight is a problem in women, particularly those living in urban areas; however, the findings on obesity are inconsistent.

Differences in study findings may be related to differences in socio-economic and demographic variables between urban and rural populations. Unlike in the developed countries where the relationship between overweight and obesity and socio-economic status (SES) is clearer, in developing countries this relationship has not been explored fully (Sobal et al 1989, Fezeu et al 2005).  Therefore, the aim of this study was to describe and compare the anthropometric profiles of an urban and rural population in Zimbabwe in relation to overweight and obesity. In addition, socio-economic and demographic variables were compared between urban and rural residents.
8.2 Methods

8.2.1 Subjects

Two groups of participants of African origin were recruited from an urban and rural area in Zimbabwe. A cross-sectional study design was employed to ensure equal numbers of males and females in each area. Convenience sampling was used to recruit participants in rural and urban Zimbabwe. The rural population consisted of 55 men and 108 women from Nkayi district in Matabeleland North. The urban population consisted of eight men and 17 women from an urban low-density suburb in Harare, and 28 male and 16 female students from the University of Zimbabwe.

8.2.2 Ethical approval

Ethical approval was granted by the Medical Research Council of Zimbabwe (MRCZ). Participants gave written informed consent.

8.2.3 Study sites

Study sites were identified through the assistance of the Central Statistics Office in Harare, Zimbabwe. The following areas were allocated
Urban
 Harare- Mavuku, Mafakose, Hatcliffe (high density townships in Harare)
 Urban (low density neighbourhood in Harare)
Rural
Nkayi Matabeleland North- Gwelutshena, Sebhumane, Tsheli, Mathendele and Nkayi centre 
A pilot of the study was conducted in a sample of students at the University of Zimbabwe. 

8.2.4 Recruitment strategy

On receiving approval from relevant community leaders, participants were recruited through canvassing and posters in the urban area. In the rural area, participants were informed of a date, time and location of the study site and invited to participate. In urban areas, house-to-house testing was conducted.

8.2.5 Outcome measures

All questionnaires used in the fieldwork are attached in appendix 8.1 as a field workbook. The demographics and socio-economic variables described in this chapter apply to chapters nine and ten, which described and compared, diet and physical activity, and blood pressure and lipid profile. These sections are in italics to emphasise this.

8.2.5.1 Demographics and socio-economic variables

A pre-coded questionnaire was used to collect demographic and socio-economic variables which included age, sex, smoking status, marital status, education, and occupation, number of dependants, household income and medical history. 
The outcomes for smoking status and medical history will be reported in chapter ten of this thesis.


8.2.5.2 Anthropometry

Anthropometric measurements were taken following guidelines set by the International Society for the Advancement of Kinanthropometry (ISAK) set out in the manual International Standards for Anthropometric Assessment (Marfell-Jones et al 2006). 
8.2.5.2.1 Height
Height was measured (to the nearest completed10 mm) in the Frankfort plane using a free standing Leicester Height Measure (Invicta Products, Leicester). 
8.2.5.2.2 Weight
Weight was measured (to the nearest completed 10 g) using the electronic scale. Subjects were barefoot and wore light clothing. 
8.2.5.2.3 Waist Circumference
Waist circumference was taken with the subject standing arms at the sides, feet together and abdomen relaxed, at the narrowest part of the torso above the umbilicus and below xiphoid process.
8.2.5.2.4 Hip circumference
Hip circumference was taken with subject standing and feet together; a horizontal measure was taken at the maximal circumference of the buttocks.
8.2.5.2.5 Skinfolds
Biceps, triceps, subscapula and suprailiac skinfolds were taken according to the ISAK guidelines.
8.2.5.2.6 Percent body fat
Percentage body fat was estimated body density derived from the sum of seven skinfold thicknesses including triceps, subscapula, suprailiac, abdominal, chest, thigh and midaxillar.


Body density and percentagebody fat were calculated using the following equations
Jackson and Pollock (1978)
Body density in men = 1.112-0.00043499 ((7) + 0.00000055 ((7)2 ? 0.00028826 (age) 
Jackson et al., (1980)
Body density in women = 1.097 ? 0.000046971 ((7) + 0.00000056 ((7)2 -0.00012828 (age)
Where (7 is the sum of seven skinfold thicknesses including triceps, chest, midaxillar, subscapula, suprailiac, abdominal and thigh.
percentagebody fat was calculated using the equation of Siri (1961)
%BF = (4.95/body density ? 4.5)*100

8.2.6 Statistical analysis

Data were tested for normal distribution.  Kolomogov-Sminorv test showed that in women, weight, height, WC, skinfolds and %BF were not normally distributed. Among men, waist and hip circumferences and skinfolds were not normally distributed.  Differences between measures were assessed using ANOVA and the Krusal-Wallis test for the variables that were not normally distributed. All statistical analysis was performed using SPSS.


8.3 Results

The results are presented for rural men and women from Nkayi compared with urban men and women from Urban and university students.

8.3.1 Demographics and socio-economic variables
Table 8.1 shows social demographic variables including gender, marital status, education and employment status.


Table 8. 1 Demographic and socioeconomic characteristics
RuralUrbanUniversityGender (n)
M (55)F (108)M (8)F(17)M (28)F (16)Mean age (SD) (years)40.2( 13.2)41.6 (12.2)39.6 (16.7)42.4 (13.5)21.1 (1.1)*20.8 (1.0)*Marital status (%)Single17.96.45035.3100100Married78.670.65035.3-Widowed1.816.4-23.5--Divorced /separated1.86.4-5.9--Education (%)No formal education14.311.9----Primary48.253.2-5.9---Secondary32.133.92517.6--University/ college5.40.97576.5100100Employment status (%)Employed16.61.862.541.2--Unemployed23.240.4-29.4--Student1.81.812.511.8100100Self-employed58.956.025.011.8--Other---5.9--Number of dependants (mean)6.96.81200*University students were significantly younger than urban and rural men and women F (2, 30)=80.6 p=0.0001

8.3.1.1 Mean age

There were no significant differences in age between rural and Urban men and women. University students were significantly younger than (p=0.0001) rural and Urban men and women.


8.3.1.2 Marital status

The majority of rural men and women were married, with more men (78.6%) than women (70.6%) being married. More rural women (16.4%) than rural men (1.8%) were widowed. There were more men (50%) than women (35.3%) who were single in Urban, more men (50%) than women (35.3%) were married, 23.5% of women were widowed and 5.9% were divorced or separated. All university students were unmarried.

8.3.1.3 Education

In this population those with no formal education are restricted to rural areas where 14.3% of men and 11.9% of women had no formal education. The majority of rural residents had primary education as their highest qualification, 48.2% in men and 53.2% in women.  A further 32.1% of men and 33.9% of women in the rural area had secondary education. A limited number had university or college education, 5.4% in men and 0.9% in women. Urban residents had higher levels of further education than rural residents, where 75% of men and 76.5% of women had university or college education. By virtue of being at university, all university students had high levels of education. 

8.3.1.4 Employment status

In the rural group 16.6% of men and 9.4% of women were employed, 23.2% of men and 40.4% of women were unemployed; 1.8% of men and women were students and 58.9% of men and 56.0% of women were self employed. A greater number of urban residents were employed compared to the rural population. More Urban men (62.5%) and women (41.2%) were employed compared to the rural population. No urban men were unemployed and 29.4% of women were unemployed; 12.5% and 11.8% of men and women were students; 25% of urban men and 11.8% of women were self employed.
In the rural group, the majority reported being self-employed with the most common occupation being farming in both men and women. Most of the urban residents were either employed or self-employed. Among men, occupations included, managers, company directors, university lecturers and businessmen. Among women occupations included administrators and lecturers. 

8.3.1.5 Number of dependents

Rural men and women had an average of six dependants while urban men and women had on average two dependants. University students had no dependants.

8.3.2 Anthropometric measures of adiposity 

Table 8.2 shows the results for weight, height, BMI and percentagebody fat  (%BF) derived from the sum of skinfold thickness in rural and urban men and women.
Table 8. 2 Comparison of body weight, height, BMI and body fat in rural and urban residents
RuralUrbanUniversityM (55)F (108)M (8)F (17)M (28)F (16)Weight (kg)*62.8(10.1)61.7 (12.3)78.4 (10.4)69.1 (18.6)62.5 (7.0)55.9 (7.9)Height 
(m)1.72 (0.06)1.61 (0.06)1.75 (0.54)1.60 (0.05)1.74 (0.06)1.60 (0.06)BMI (kg/m2)*21.1 (2.83)23.8 (4.3)25.5 (2.8)27.3 (7.7)20.6 (2.0)21.7 (2.5)%BF*8.4 (3.7)24.0 (9.1)20.2 (7.0)34.7 (11.2)5.6 (1.1)17.8 (3.1)*Signifantly higher (p<0.05) in Urban than rural and university group

Urban men and women were significantly heavier, had higher BMI, and higher %BF (p<0.05) compared with rural and university men and women. There were no significant differences in height between all the population groups. Post hoc analysis showed that there were no significant differences in weight and BMI between rural and university men. Using the WHO cut-off points for overweight (BMI 25-29.9kg.m2) and obesity (BMI>30kg/m2), the mean BMI of Urban men and women indicated that they were overweight, while mean BMI in rural men and women and university men indicated normal weight. Amongst Urban men 37% were overweight and 12.5% were obese, 35.3% of Urban women were overweight, 23.4% were obese. In the rural population 12.5% of men were overweight and none were obese, 33% of rural females were overweight and 6.4% were obese. Amongst the university students 82.1% of men were normal weight and 93.8% of women, there was no overweight and obesity.

Table 8.3 shows differences in body fat distribution measured by circumferences and skinfold thickness analysis. 

Table 8. 3 Anthropometric measures of fat distribution
RuralUrbanUniversityM (56)F (109)M (8)F (17)M (28)F(16)WC (cm)*76.6 (7.2)77.1 (10.1)90.3 (7.84)81.6 (16.1)72.0 (0.06)68.6 (4.1)WHR*0.84 (0.05)0.80 (0.06)0.87 (0.05)0.77 (0.08)0.81 (0.06)0.72 (0.05)Biceps3.8 (1.4)8.2 (4.6)7.5 (3.1)9.7 (6.1)3.7 (1.0)9.2 (7.6)Tricep*6.9 (3.2)19.9 (9.4)16.6 (5.3)24.9 (11.9)6.4 (1.4)13.6 (4.3)Subscap*8.7 (2.7)16.4 (9.3)22.8 (12.1)24.8 (15.6)9.3 (1.4)11.0 (2.6)Suprailiac*6.5 (2.7)14.2 (10.221.0 (8.9)22.6 (13.2)6.4 (2.2)10.1 (3.2)Subscap=Subscapular skinfold, WC=Waist circumference, WHR waist-to-hip ratio *significant differences between men in each group and between women in each group, p<0.05

Urban women, had significantly larger (p<0.05) WC, WHR, triceps, subscapular and suprailiac skinfold thicknesses compared with rural and university women. However, there were no significant differences in biceps skinfold between the women. Urban men had significantly larger WC, WHR and all skinfold thicknesses compared with rural and university men. Post-hoc analysis showed that rural and university men were not significantly different in measures of fat distribution.


8.4 Discussion

The aim of this study was to describe the anthropometric profile of urban and rural Zimbabweans. The populations included an urban population consisting of a group of university students and a group from a low-density neighbourhood of Urban in Harare. The rural population were men and women from Nkayi District in Matabeleland North. Demographic and socio-economic variables showed that compared with the rural population, the Urban and university students, had higher levels of education, employment and fewer dependents. Anthropometric measures showed that the Urban group were heavier, had a higher BMI, WC, WHR and larger skinfold thicknesses compared with the university students and rural residents. However, university and rural men were not significantly different in weight, BMI, and measures of fat distribution.

8.4.1 Demographic and socio-economic differences

The median number of years of educational attainment is six for both males and females and is higher in urban areas compared with rural and among the population in the highest wealth quintile.  According to the census data reported in the ZDHS (2005/06), 91% of males and 88% of females attended school. Of which, 42% of males and 43% for females attended primary school and a further 37% of males and 36% attended secondary school. The urban population (Urban and University) in this study cannot be considered representative of the general urban Zimbabwean population as the levels of education attainment were much higher than those reported in the census. In this group, 75% and 76.5% of males and females respectively from Urban had university or college education. In contrast education attainment in the rural population is similar to that estimated in the census. 
The high education attainment of the Urban population influenced the levels of employment. A larger number of Urban men and women were employed compared to rural populations. In the ZDHS (2006) variations by place of residence showed that a higher percentageof women and men in urban areas (40% and 65 %, respectively) were employed compared with their rural counterparts (35% and 61%, respectively). 
Nationally, agriculture employs the largest percentage of Zimbabweans: 34% of both women and men. After agriculture, sales and services (31 % of women) and skilled manual labour (22% of men) have the second highest percentageof all employed women and men, respectively.  However, in this study none of the urban residents were employed in agriculture, or skilled manual labour. As such this population is not representative of urban men and women in Zimbabwe in terms of employment. The majority of the rural population were subsistence farmers, which is in line with the general population of rural Zimbabweans where more than half of women (56%) and men (57%) are employed in agriculture (ZDHS 2006). 
In sub-Saharan Africa socio-economic markers such as education and employment status are positively correlated with overweight and obesity (Zirabi et a., 2009). Therefore it is expected that those with higher levels of education and employment, typically in urban areas, will have higher levels of overweight and obesity. In South Africa, Voster et al., (2005) showed that urban professionals, who represented the most educated and employed were at greater risk of overweight and obesity than those with lower levels of education, the majority of whom lived in rural and farm areas. This trend is similar to that found in this study and may explain why urban Urban residents were more overweight and obese than similar aged rural residents.
The number of dependents differed significantly between rural and urban populations. Urban populations had fewer dependants, than rural populations. This trend is consistent with other developing countries under-going the demographic transition. As the population becomes more urbanised the number of dependents reduces (Popkin 2000).

8.4.2 Anthropometric differences

It was expected that the urban population, and in particular the Urban group would have higher measures for weight, BMI and skinfold thicknesses compared with rural and university groups. This finding is consistent with other reports within Zimbabwe. Zinyowera et al., (1994) reported that although the prevalence of obesity was low, a greater number of urban men and women were overweight compared with the rural populations.  In addition, the ZDHS (20005/06) showed that a greater number of urban than rural women were overweight and obese compared with rural women. This trend was also reported in South Africa, Botswana and Namibia (Walker 2001, Puoane et al., 2002, Voster et al., 2005). 

Interestingly, rural and university men had few differences in body composition. This was particularly true for fat distribution where the rural and university men were not significantly different in any of the measures. However, this finding is consistent with other reports of differences in body fat in males. Zinyowera et al., (1994) reported that the prevalence of obesity measured by BMI was 0% in males in rural and urban Zimbabwe, while overweight was only apparent in women. In South Africa, Voster et al., (2005) documented a progressive increase in anthropometric measures of adiposity in five groups from the THUSA study.  The differences in mean BMI between the least urbanised and the most urbanised were small less than 3kg/m2 for men, and all men were normal weight. Therefore, overall men had less body fat between the groups and therefore the differences between groups, particularly if they are within the same BMI category is small.

In Southern Africa, overweight and obesity are associated with being female and living in urban areas (Walker 2001).  Therefore although men are also under-going nutrition transition, they are at a lower risk than women of overweight and obesity. Amongst females  overweight and obesity maybe present in all sections of society, including rural, and semi-urban and urban but the same may not true for males, with very few rural males being overweight and obese.


8.4.3 Study limitations

It was intended that a similar number of urban and rural participants would be recruited to this study. However, in 2008 while undertaking this study, a cholera outbreak occurred in the urban areas of Zimbabwe, with its epicentre in Harare. Therefore, recruitment of participants in urban areas was stopped. This significantly influenced the number of urban Zimbabweans recruited. 
The small number of urban participants recruited cannot be considered representative of the majority of urban Zimbabweans. This group represents an affluent and highly educated population. As such these findings cannot be generalised to other urban Zimbabweans. However, this study does show differences in urban and rural population that have been reported in other African countries.
8.5 Conclusion

The aim of this study was to report on the demographic and anthropometric profile of urban and rural Zimbabweans in relation to overweight and obesity. Urban residents from Urban were heavier, had a higher BMI, %BF and larger skinfolds compared to university students and rural residents. Rural men had similar anthropometric measures to university men.  However it is important to recognise that these results may not be generalisable in Zimbabwe. This study adds to other studies that show differences in the prevalence of overweight and obesity in rural compared with urban populations.


8.6 References

Fezeu, L., Minkoulou, E., Balkau, B., Kengne, A.P., Awah, P., Urwin, N., Alberti, G.K.M.M., Mbanya, J.C., (2005). Association between socio-economic status and adiposity in urban Cameroon. International Journal of Epidemiology, 35, pp.105-111.
Mathe, S., Matovu, H.L., Mossop, R.T., (1985). Nutritional status of an urban community in Zimbabwe. Central African Journal of Medicine, 31 pp.59-62.
Popkin, B.M., (2000). Achieving urban food and nutrition security in the developing world: Urbanisation and the nutrition transition. Focus, 7, (10).

Puoane, T., Steyn, K., Bradshaw, D., Laubscher, R.,?Fourie, J.,?Lambert, V., and Mbananga, N., (2002). Obesity in South Africa: The South African Demographic and Health Survey. Obesity Research, 10, pp.1038-1048.
Sobal, J., Stunkard, A.J., (1989). Socio-economic status and obesity: A review of the literature, Psychology Bulletin, 19, pp.260-275.
Sobngwi, E., Mbanya, J-C. N.,?Unwin, N. C.,?Kengne, A. P.,? Fezeu, L.,? Minkoulou, E. M.,? Aspray, T. J.,? Alberti, K. G. M. M.,? (2002). Physical activity and its relationship with obesity, hypertension and diabetes in urban and rural Cameroon. International Journal of Obesity, 26 pp.1009-1016.

Useh, U., Mbajiorgu, F.E., Feresu, S.A., Madzivire, D., (2000) Preliminary study of body mass index and its relationship to percent body fat in Harare, Zimbabwe. Central African Journal of Medicine. 46, pp.101-105.
Vorster, H.H., Venter, C.S., Wissing, M.P., and Margetts, B.M., (2005). The nutrition and health transition in the North West Province of South Africa: a review of the THUSA (Transition and Health during Urbanisation of South Africans) study. Public Health Nutrition, 8, pp.480?490.
Walker, A.R.P., Adam, F., Walker, B.F., (2001). World pandemic of obesity: the situation in Southern African populations. Public Health, 115, pp.368?372.
Watts, T.E., Siziya, S., (1997). Education, occupation and health status of people of age five years or more living in a high-density urban area in Zimbabwe. Central African Journal Medicine. 43, pp.260-264.
World Health Organization (1995). Physical status: The use and interpretation of anthropometry. Geneva: WHO
World Health Organization (1998). Obesity: Preventing and managing a global epidemic. Geneva: WHO
Zinyowera, T., and Msamati, B.C., (1994). Obesity: is it a problem in black Africans in Zimbabwe? Central African Journal of Medicine, 40, pp.33-38.
Ziraba A, K,,  Fotso, J.C.,  Ochako, R., (2009). Overweight and obesity in urban Africa: A problem of the rich or the poor? BMC Public Health 9, pp.465-474

Zimbabwe demographic and health survey (2005/06). [online] available at www.measuredhs.com/pubs/pdf/FR186/FR186.pdf  [Accessed November 2008].


 ENERGY BALANCE IN A SAMPLE OF RURAL AND URBAN ZIMBABWEANS
9.1 Introduction

The habitual energy intake in many developing countries including Zimbabwe is generally below the international recommended levels of the FAO/ WHO/UNU 1985 committee. Food insecurity is a problem for many people in both urban and rural Zimbabwe, and this might explain the low energy intake in these populations. The co-existence of low energy intake (under-nutrition) and high intake (over-nutrition) is a feature of the nutrition transition in developing countries (Popkin 2000). 
In Zimbabwe, there is a paucity of studies recording energy intake in urban or rural populations; the available data reports on protein energy malnutrition in child-bearing women and children (Allain et al., 1997). However, there is more data available in South African populations. Bourne et al., (2002) reviewed studies of energy from macronutrient intake (carbohydrate, protein and fat) in urban and rural South Africans between 1940 and 1992. Energy from protein intake remained consistent between the years while energy from carbohydrate intake decreased and energy from fat intake increased in both rural and urban populations.  These findings are consistent with the pattern reported for developing countries undergoing nutrition transition (Popkin 2009). 
MacIntyre et al., (2002), showed differences in energy intake and sources of energy with level of urbanisation by evaluating dietary intake of 1751 South African adults.  The sample was stratified into five groups including rural residents, farm workers, informal settlement dwellers, urban Africans living in townships and urban African professionals who participated in the Transition and Health during Urbanisation of South Africans (THUSA) study. The five groups represented differing levels of urbanisation, with the rural and farming populations being the least urbanised, the informal settlement dwellers representing populations in between rural and urban, and the urban township dwellers and urban professionals, the most urbanised. Mean energy and protein intakes for all strata were adequate when compared with recommendations from the FAO/WHO/UNU (1985). Mean energy from carbohydrate was highest in the rural and farm groups, 67% and 68% respectively, and decreased progressively with the lowest being in the urban professionals group at 56%. Conversely, energy from fat was highest in the urban professionals, at 31% and lowest in the rural group at 23%. Energy from protein remained consistent between the groups ranging from 11-13%.
The FAO food balance sheets record information on availability patterns in 157 countries. From these balance sheets, trends in availability can be observed, however the food balance sheets do not measure energy intake or expenditure. Analysis of food balance sheets for Zimbabwe between 1961 to 2007 shows a similar pattern in the sources of energy intake in Zimbabwean populations to that reported in South Africa. In 1961, intake of cereals (carbohydrates) was 1509 kcal/capita/day, this decreased to 1280 kcal/capita/day in 2007, whereas, intake of fat (from animal sources) in 1961 was 18 kcal/capita/day and increased to 46 kcal/capita/day in 2007. The food balance sheets provide the best available data on food availability in Zimbabwe, however, it is not possible to assess rural-urban differences from these data.
The relationship between physical activity and chronic diseases is well established in populations in developed countries. Seminal studies such as Morris et al., (1966) in London bus drives and conductors, and Paffenbager et al., (1970) in longshore men showed that reduced physical activity was related to an increased incidence of chronic disease. The changes in diet seen in developing countries are accompanied by a reduction in physical activity, and therefore more people in developing countries may be at increased risk of chronic disease (Popkin 2009). The changes in physical activity are a result of technological advancements and mechanisation, associated with urbanisation that affects work, leisure and transportation. Consequently, urban populations in developing countries are increasingly sedentary and energy expenditure is reduced. 
Habitual energy expenditure in many populations in developing countries and particularly in rural areas, appears to be higher than intake. Mingelli et al., (1990) compared energy expenditure in 20 rural Gambian men matched for weight with 16 European men. Twenty-four hour (24 hour) energy expenditure was lower in Gambian men; this included lower basal and sleeping energy expenditure, compared with European men. The investigators concluded that the low energy expenditure was an important energy sparing and coping mechanism on a low level of dietary intake during the rainy season in Gambian men. Similarly, in Gambian women, energy intake was significantly lower than energy expenditure, and the women appeared to be in constant energy deficit during certain seasons of the year (Singh et al., 1989).
Energy intake and expenditure can be influenced by seasons, especially in rural areas in Sub-Saharan Africa. In Zimbabwe and other sub-Saharan countries, during the rainy season, high levels of energy are expended in subsistence farming; however there is limited availability of food. This is in contrast to the dry seasons where more food is available but less energy is expended (Kinabo et al., 2003). Bleiberg et al. (1980) observed that energy expenditure in rural women from Burkina Faso increased by 25% from the dry to the rainy season. A similar observation was made by Schultink et al., (1990) and Ategbo (1994) in Benin, during the rainy season the volume of activities increased considerably. Low-income urban populations may also be affected by seasonality, as the reduction in the availability of food during the rainy seasons affects the excess produced by farmers and sold in urban markets. More importantly, the prices of available food increase thus affecting energy intake of low-income urban populations during the rainy season (Kinabo et al., 2003).

Energy intake and expenditure can be measured directly (measurement of heat production), indirectly (measurement of oxygen consumption) or by non-calorimetric methods (doubly labelled water methods). However, in most epidemiological studies, energy intake and expenditure are determined through the use of questionnaires and energy is calculated using equations such as Schofield 1985. The Sub-Saharan Africa Activity Questionnaire (SSAAQ) was developed as an epidemiological tool to assess energy expenditure in people of African origin in Sub-Saharan Africa (Sobngwi et al., 2001). The questionnaire provides an alternative to measures of energy expenditure, which may be inaccessible in developing countries because of cost. Moreover, the SSAAQ was developed and validated in an African population and therefore encompasses activities that are common in urban and rural Africa.
In 2005, Merchant and colleagues developed a semi-quantitative food frequency questionnaire specifically for use in black Zimbabwean populations. This questionnaire consists of food groups commonly consumed by Zimbabwean populations in urban and rural areas. Therefore, dietary patterns can be estimated from this culturally specific questionnaire. The aim of this study was to compare energy intake and expenditure between rural and urban Zimbabweans.

9.2 Methods

Subject characteristics, ethical approval and recruitment have been described in chapter eight of this thesis.

9.2.1 Outcome measures

Energy intake and expenditure were assessed in rural and urban populations using questionnaires (appendix 8.1). Trained fieldworkers administered the questionnaires in both rural and urban locations.  

9.2.1.1 Energy intake

An adult semi-quantitative food frequency questionnaire (FFQ) was used to provide information of food consumption patterns, dietary habits over time and equivalent nutrient intakes for subjects. The questionnaire was developed and validated in a Zimbabwe population (Merchant et al., 2005).
A 24hr food recall was administered in addition to the FFQ. A trained interviewer asked each participant to recall all the food consumed in the 24 hours prior to participating in the study. Mean energy intake and intake from protein, fat and carbohydrate was compared with estimates from dietary guidelines from the FAO/WHO/UNU (1985). Dietplan 6.0 software (Forestfield Software Limited) was used to analyse the food frequency and 24hr recall. 


9.2.1.2 Total energy expenditure (TEE)

Physical activity was assessed using the Sub-Saharan Africa Activity Questionnaire (SSAAQ) (Sobngwi et al., 2001). This questionnaire was developed and validated for use in populations in Sub-Saharan Africa. Fieldworkers asked participants about their occupation and leisure time activity over the past year, month and week. Participants responded giving details of activities they undertook on a regular basis. Regular was defined as more than twice in six months (Sobngwi et al., 2001). Frequency and duration was computed for each reported activity and the cumulative time spent in occupational and leisure time activity was combined to get total activity for each participant. The metabolic cost was calculated using Ainsworth?s compendium for physical activity (Ainsworth et al., 1993). Physical activity levels (PALS) were calculated for each activity using guidelines from United Nations University (UNU). 


9.2.2 Statistical analysis

Data was tested for normality. Total energy expenditure was normally distributed, however energy intake, and % energy from protein, carbohydrate and fat were not normally distributed. The students? t test was used to compare differences in mean total energy expenditure between rural and urban men and women. The Mann-Whitney test was used to compare energy intake and % energy from protein, carbohydrate and fat.


9.3 Results

Energy intake and dietary patterns were compared between rural and urban Zimbabweans. In addition energy expenditure and physical activity patterns were compared. The results are presented in sections, the first a comparison of energy intake, secondly energy expenditure and thirdly a comparison between differences between intake and expenditure.

9.3.1 Comparison of daily energy intake and dietary patterns in urban and rural Zimbabweans.

Commonly consumed foods were identified for each population from food frequency questionnaires and are shown in table 9.1. In each category, it can be seen that Urban men and women had a greater variety of food products available to them compared with rural men and women. 


Table 9. 1 Commonly consumed foods in urban and rural Zimbabweans
RuralUrbanDairy foodsWhole milkWhole milk, semi-skimmed milk, cheese, yoghurt and ice-cream, butter and margarineFruitsWild fruits, mangoes, guavas, orangesBananas, paw paw, guava, mango, naartjie, applesVegetablesGreen leafy vegetables including spinach, kale, Carrots, cucumbers, mushrooms, kale, cabbage, peas, beans-green, butter beansMeat, eggs etcChicken, beef, goat, birds, locust, fish-driedChicken, beef, goat, fish-canned, fresh and driedBreads, cereals and starchesMaize meal ? sadza, homemade bread, soghum and millettMaize meal, rice, pasta, breadRoots and tubersSweet potatoes, potatoesSweet potatoes, potatoesBeveragesTea, traditional beerTea, coffee, fruit juices, fizzy drinks, alcohol-lager, wine and spiritsSweets and confectionaryPeanut butterCake, chocolate, sweetsDietary supplementsNot commonly usedMultivitamins
Energy intake and dietary sources of energy were compared between rural urban Zimbabweans, the results are shown in table 9.2.

Table 9. 2 Comparison of daily energy intake and sources of energy in rural and urban Zimbabweans (mean (SD))
RuralUrbanFAO/UNU/WHO (1985) recommendationM (33)F (66)M (8)F (17)Energy intake (kJ)7412 (6254)4394 (3238)9475 (5790)5272 (2867)% Protein*13.2 (7.4)12.6 (7.9)17.0 (9.3)24 (16)15%% Carbohydrate**61.1 (21.8)68.1 (18)45.5 (19.6)42.5 (14.1)50%% Fat*25.7 (18.4)19.3 (15)27.1 (12.8)32 (9.4)35%* Mean energy from protein and fat significantly (p=0.0001) lower in rural compared with Urban, **Mean energy from carbohydrate was significantly higher in rural compared with Urban


Compared with the FAO/UNU/WHO recommendations for energy, rural men and women had lower mean energy from protein, higher energy from carbohydrate and lower mean energy from fat. Whereas, Urban men and women had higher mean energy from protein, lower energy from carbohydrate and lower energy from fat. Total calories did not differ between rural and Urban men and women. However, rural men and women had significantly lower (p=0.0001) energy from protein and fat than Urban men and women. Rural men and women had significantly higher (p=0.0001) energy from carbohydrate compared with Urban men and women.


9.3.2 Comparison of energy expenditure and physical activity patterns in rural and urban Zimbabweans


The most common occupational and leisure time physical activities were compared between rural and Urban men and women and are shown in table 9.3. The majority of rural residents were engaged in labour intensive farming activity while urban residents had more sedentary occupations. Rural residents engaged in more purposeful leisure activity, which was not related to relaxation and pleasure. Urban residents were engaged in more relaxation type leisure activity.

Table 9. 3 Comparison of common physical activities in rural and urban Zimbabweans
RuralUrbanOccupational activityFarming, animal rearing, carpentry and metal work and market tradingManagerial occupations, lecturers, housewives, consultantsLeisure activityGardening, knitting, housework, church and community meetingsWatching television, reading, golf
Table 9.4 shows a comparison of total energy expenditure and physical activity levels between rural and urban men and women. Energy expenditure in rural compared with urban populations was significantly higher than urban populations. Mean PAL for rural men and women is indicative of vigorous physical activity, whereas the mean PAL for urban men and women (1.9 and 1.7 respectively) is indicative of moderate activity. Using PALS to stratify levels of physical activity. More rural men (45.5%) and women (57.6%) were extremely active compared with urban men (12.5%) and women, of whom none were extremely active.  Proportionally more urban men (25%) and women (38.5%) were sedentary compared with rural men (9.1%) and women (6.1%).


Table 9. 4 Physical activity levels in rural and urban Zimbabweans
RuralUrbanGender (n)M (55)F (66)M (8)F (17)TEE (kJ)12258 (1433)10623 (1254)13160 (2065)9976 (1558)Mean PAL2.6 (1.2)2.6 (0.78)1.9 (0.5)1.7 (0.3)Level of activityExtremely inactive <1.46.1 1.5-15.4Sedentary 1.4-1.699.16.125.038.5Moderately active 1.7-1.9918.210.637.523.1Vigorously active 2-2.421.224.225.023.1Extremely active >2.445.557.612.5-PAL= physical activity level

9.3.3 Comparison of energy balance in rural and urban Zimbabweans

An energy imbalance can lead to excessive accumulation or loss of adiposity tissue. In both cases this may impair health. Table 9.5 shows comparison of energy balance between the groups.

Table 9. 5 Energy balance in rural and Urban men and women
RuralUrbanM (33)F (66)M (8)F (17)Energy intake (kJ)7412 (6254)4394 (3238)9475 (5790)5272 (2867)Energy expenditure (kJ)12258 (1433)10623 (1254)13160 (2065)9976 (1558)Energy deficit/ excess-4846-6229-3685-4704
Results show that men and women in this study had a negative energy balance in both rural and urban areas. This was most pronounced in rural women who had a mean energy deficit of 6229kJ.


9.4 Discussion

This study compared energy balance in rural and urban Zimbabweans. Main findings show that the proportion of energy from protein and fat was higher in urban compared with rural men and women, whereas energy from carbohydrate was higher in the rural group. Energy expenditure was higher in the rural compared with the urban group. Comparison of energy balance showed that all groups were in energy deficit, and this was most pronounced in rural women.

With urbanisation the diets of populations under-going the nutrition transition, are increasingly high in fats, salt and sugars (Popkin and Nielson 2003, 2006). A key feature in the dietary change associated with the nutrition transition is the reduction in energy from complex carbohydrates and an increase in energy from fat. In this study this shift was observed between rural and urban populations. This trend is consistent with changes in populations who are transitioning from rural to urban environments in other countries in Sub-Saharan Africa. Voster et al., (2005), showed a progressive decrease in energy from carbohydrate and an increase in energy from fat in five groups of South African men and women. The rural and farm populations, who were the least urbanised, had a higher proportion of energy from carbohydrate than the urban population. The main cause of the shift in sources of dietary energy is the wide variety of food products available to urban compared with rural populations. Many of these products are fast and convenience foods that are usually high in fats, salt and sugars. Fast and convenience foods are more accessible to urban populations who have access to supermarkets and fast food chains (Schmidhuber and Shetty 2005). Consumption of convenience foods in urban areas is a consequence of the increasing number of employed women and thus there is a limited amount of time for food preparation. This was observed in this study where a greater proportion of urban women were employed compared with rural women. 

Energy expenditure in rural populations is higher than urban populations. This is a result of activities related to subsistence farming. In this study, the majority of the rural population were subsistence farmers. Subsistence farming in rural Zimbabwe is a laborious task as there is little mechanisation of farming equipment. Women undertake the larger share of the farming activity, thus rural women expend a larger amount of energy than any other group. The trend of higher energy expenditure has been observed in other populations of rural women in Sub-Saharan Africa. In the Gambia, the volume of physical activities fluctuated with season. In the dry season, energy expenditure increased because of activities related to farming. In addition, in this season food is less abundant.

Energy balance in this population was negative for all groups and was most pronounced in rural women. This implies that the population was constantly in energy deficit. However, it has been reported that energy balance in marginally nourished populations in the developing countries frequently reported very low levels of energy intake which appear to be incompatible with the high levels of physical activity associated with a rural farming life. Researchers investigating energy balance in rural Gambia suggested that the low energy expenditure was an important energy sparing and coping mechanism on a low level of dietary intake due to seasonal variation in the availability of food (Mingelli et al., 1990, Kinabo et al., 2003). This may explain the energy deficit in the rural population in this study. Data were collected during the dry season in Zimbabwe. However, this reason does not apply to the urban population. The energy deficit in the urban population may be explained by limitations in data collection tools, which are discussed in the following section.


9.4.1 Study limitations

This study used 24 hour recalls to assess energy intake, they provide a quick method to assess intake, however, cannot be used to classify a subjects usual intake as it is not necessarily representative of the subjects normal eating pattern. This same criticism can be applied to the activity questionnaire. However, both methods provide the most reliable and affordable methods for the assessment of energy intake and expenditure in epidemiological studies. 

9.5 Conclusion

This study investigated differences in energy balance in rural and urban Zimbabwean men and women. Main findings showed that sources of energy differed between the groups; urban residents received more energy from fat compared with rural residents, while energy from carbohydrate was higher in the rural population. Energy expenditure was higher in the rural population and particularly for rural women. Both populations appeared to be in energy deficit, however, this may be accounted for by seasonal availability of food in the rural population, that may also be penetrating to the urban group.


9.6 References

Ainsworth, B.E.; Haskell, W.L., Leon, A.S., Jacobs JR, D.R., Montoye, H J., Sallis, J F., Paffenbarger JR, R.S., (1993). Compendium of Physical Activities: classification of energy costs of human physical activities. Medicine & Science in Sports & Exercise, 25, pp.71-80.
Allain, T.J., Adrian, O., Wilson, Z. Gomo, A.R., Adamchak.D.A., Matenga, J.A., (1997) Diet and nutritional status in elderly Zimbabweans. Age and Ageing, 26, pp.463-470.
Ategbo, E.A.D., (1994). Food and nutrition security in northern Benin, impact on growth performance and on year-to-year nutritional status of adults. PhD Thesis: Wageningen Agricultural University, The Netherlands
Bleiberg, F., Brun, T.A., Goihman, S., (1980). Duration of activities and energy expenditure of female farmers in dry and rainy season in Upper-Volta. British Journal of Nutrition, 43, pp.71-82.

Bourne, L.T., (2000). Rural/urban nutrition-related differentials among adult population groups in South Africa, with special emphasis on the black populationSouth African Journal of Clinical Nutrition, 13, pp.
Bourne, L.T., Lambert, E.V., and Steyn, K. (2002). Where does the black population of South Africa stand on the nutrition transition? Public Health Nutrition, 5, pp.157?162,
FAO/WHO/UNU (1985). Energy and protein requirements. Geneva: World Health Organization 
Kinabo, J., Kamukama, E.,  Bukuku, U., (2003). Seasonal variation in physical activity patterns, energy expenditure and nutritional status of women in a rural village in Tanzania. South African Journal of Clinical Nutrition, 16, pp.96-102
MacIntyre, U. E. Kruger, H. S. Venter, C. S. and Vorster, H. H., (2002).  Dietary intakes of an African population in different stages of transition in the North West Province, South Africa: the THUSA study. Nutrition Research, 22, pp.239-256.
Merchant A.T., Dehghan, M., Chifamba, J., Terera, G., and Salim, Y., (2005) Nutrient estimation from an FFQ developed for a black Zimbabwean population. Nutrition Journal, 4, pp.37-46.
Minghelli, G., Schutz, Y., Charbonnier, A., Whitehead R.,, and Eric J. (1990) Twenty-four--hour energy expenditure and basal metabolic rate measured in a whole-body indirect calorimeter in Gambian men. American Journal of Clinical Nutrition, 51, pp.563-570. 

Morris, J.N., Kagan, A., Pattison, D.C., Gardner, M.J., (1966). Incidence and prediction of ischaemic heart disease in London busmen. Lancet, 2, pp.553-539. 

Paffenbarger Jnr, R.S., Laughlin, M.E., Gima, A.S., Black, R.A., (1970). Work activity of longshoremen as related to death from coronary heart disease and stroke. New England Journal of Medicine, 282, pp.1109-1114. 

Popkin, B.M., (2000). Achieving urban food and nutrition security in the developing world: Urbanisation and the nutrition transition. Focus, 7, (10).

Popkin, B.M., (2006). Global nutrition dynamics: the world is shifting rapidly toward a diet linked with noncommunicable diseases. American Journal of Clinical Nutrition, 84, pp.289 ?298. 

Popkin, B. M., (2009). The Nutrition Transition in Low-Income Countries: An Emerging Crisis. Nutrition Reviews, 52,?pp.285???298.

Prentice, A. M., (2006). The emerging epidemic of obesity in developing countries. Intenational Journal of Epidemiology, 35, pp.93-99.
Schmidhuber, J., and Shetty, P., (2005). The nutrition transition to 2030. Why developing countries are likely to bear the major burden. Food Economics - Acta Agriculturae Scandinavica, 2, pp.150 ? 166.
Schofield, W.N., (1985). Predicting basal metabolic rate, new standards and review of previous work. Human Nutrition- Clinical Nutrition, 39 C, pp.5-41.
Schultink, W.J., Klaver, W., Van Wijk, H., Van Raiij, J.M.A., Hautvast, J.G.A.J., (1990). Body weight changes and basal metabolic rates of rural Beninese women during seasons with different energy intakes. European Journal of Clinical Nutrition, 44 pp.31-40.

Singh, J., Prentice, M.A., Diaz, E., Coward, W.A., Ashford, J., Sawyer, M., Whitehead, R.G., (1989). Energy expenditure of Gambian women during peak agricultural activity measured by the doubly-labelled water method. British Journal of Nutrition, 62, pp.315-329.

Sobngwi, E., Mbanya, J.C.N., Unwin, N. C., Aspray, T. J., and Alberti K.G.M.M., (2001). Development and validation of a questionnaire for the assessment of physical activity in epidemiological studies in Sub-Saharan Africa. International Journal of Epidemiology, 30, pp1361-1368.
United Nations University (total energy expenditure and physical activity in adults: doubly labelled water data [online] available at: http://www.unu.edu/unupress/food2/uid01e/uid01e08.htm  [Accessed July 2009]
BLOOD PRESSURE, BLOOD GLUCOSE AND LIPID PROFILES IN RURAL AND URBAN ZIMBABWEANS

10.1 Introduction

In many developing countries, infectious disease and undernutrition account for a large proportion of the disease burden and mortality. Diseases such as tuberculosis, HIV/AIDS and malaria are the main cause of death in LMCs (Reddy 2002, Boutayeb 2006).  However, in 2004, it was reported that 82% of the non-communicable disease burden was attributed to populations in developing countries (WHO 2004) of which CVDs were most prevalent. Therefore, LMC experience a double burden of disease as NCDs and communicable diseases coexist (WHO/FAO 2002).  In Sub-Saharan African, the prevalence of CVD has been low. However, it has been reported that the current prevalence of some CVDs particularly in urban societies, is as high as those seen in developed countries (Voster 2002, Njelekela et al., 2003, Mufunda et al., 2006, BeLue et al., 2009).

Globally, populations of African origin are at increased risk of stroke and renal disease as a result of hypertension. In addition, higher than optimum blood glucose was found to be the leading cause of cardiovascular mortality, in most regions of the world including Sub-Saharan Africa (Danaei et al., 2006). However, the prevalence of coronary heart disease is low in most populations of African origin (Gaillard et al., 2009). It is thought that the low prevalence of CHD is related to the favourable lipid profile observed in populations of African origin. Specifically, the low serum cholesterol, low levels of low density lipoprotein cholesterol (LDL-C) and increased high-density lipoprotein cholesterol (HDL-C), which is protective against heart disease when compared with populations of Caucasian origin (Sumner 2009). 
It is well established in Caucasian populations in developed countries, that serum lipids and lipoproteins, especially serum cholesterol, triglycerides, low-density lipoprotein (LDL-C), and very-low-density lipoprotein (VLDL), are important independent risk factors for coronary heart disease whereas high-density lipoprotein (HDL-C) has a protective effect (Kannel et al., 1971, McQueen et al., 2008). Data on serum lipids is limited in many Sub-Saharan African countries including Zimbabwe. It has been reported that that lipid disorders are uncommon among black Zimbabweans (Castle, 1982 and Gomo 1985). Gomo (1985), in the only study to date to investigated the lipid profile of black Zimbabweans, found that the mean values and the range for total cholesterol and triglyceride were higher for men than for women and that concentrations of HDL-C were higher for women than for men which is consistent with findings in pre-menopausal women. More importantly the population showed a higher proportion of HDL-C than LDL-C which is believed to offer protection against CHD. The favourable lipid profile was accredited to the low fat diet in the population studied. Gomo (1985) went on to establish normal ranges for Zimbabwean men and women for blood cholesterol as shown in table 10.1

Table 10. 1 Ranges for blood lipids in Zimbabwean men and women (mmol/l)
MenWomenTotal cholesterol1.95-6.991.73-6.81HDL-C0.78-2.300.87-2.42Triglycerides0.49-1.450.40-1.36HDL-C high density lipoprotein Source: Gomo (1985).

This study was conducted over 20 years ago and Zimbabweans have undergone changes in diet and lifestyle such that these ranges may no longer be applicable. More recently in the Zimbabwe Non-Communicable Disease Risk Factors (ZiNCoDs) preliminary report (2005), 20.4% of men and 21.4% of women had raised cholesterol defined as total cholesterol ? 5.2 mmol/l, and 3.1% of men and 4.8% of women had raised cholesterol defined as total cholesterol ? 6.5 mmol/l. These data show a dramatic change in raised cholesterol in Zimbabweans comp? 5.2 mmol/l, and 3.1% of men and 4.8% of women had raised cholesterol defined as total cholesterol ? 6.5 mmol/l. These data show a dramatic change in raised cholesterol in Zimbabweans compared with the findings of Gomo (1985) and Castle (1982). However, information on levels of HDL-C and LDL-C was not differentiate between rural and urban populations.

Several studies provide data about hypertension in Sub-Saharan Africa, but very few of these provide age-standardised data, which allow comparability between studies. In 2007 Addo et al., reviewed 37 studies reporting prevalence of hypertension in 25 Sub-Saharan African countries. Using hypertension as defined by the WHO guidelines (blood pressure >140/95mmHg), they showed that the prevalence of hypertension ranged from 6.8% in rural Gambia to 25.4% in urban Tanzania. The prevalence of hypertension increased with age, and urban populations consistently had a higher prevalence of hypertension compared with their rural counterparts (Addo et al., 2007). In Zimbabwe, an analysis of all national data collected in the 1990s found that between 1990 and 1997, the national crude prevalence of hypertension increased from 1% to 4% (Mufunda et al., 2000). However, latest figures from the ZiNCoDs, (2005) showed that the prevalence of hypertension was 23% in men and 29% in women aged 25-100 years. Urban men and women had higher prevalence of hypertension compared with rural populations.
The ZiNCoDs (2005) also showed that the prevalence of diabetes defined as fasting blood glucose of (7.8mmol/l was 2.2% in men and 1.3% in women. These values indicate a very low prevalence of diabetes in Zimbabwe. However, this trend is not expected to remain the same. Wild et al., (2004) projected that the prevalence of diabetes globally will double by 2030, with the greatest increases seen in developing regions as a consequence of rapid urbanisation and population aging.
For most non-communicable diseases, and particularly CVD, early detection increases the chances of positive prognosis and for prevention strategies to be implemented. However, although cost effective in the longer term, screening is not readily available in resource-limited settings. Therefore, patients present at hospital at late stages of the disease requiring more treatment. 
Thus the aim of this study was to report on the differences in blood pressure, lipid profile and blood glucose in urban compared with rural Zimbabweans. 
10.2 Methods

Subject characteristics have been described for this population in chapter nine of this thesis.
10.2.1 Outcome measures

10.2.1.1 Health history and Smoking status

A questionnaire (Appendix fieldwork book appendix 8.1) was used to obtain previous medical history and smoking status. Participants reported previous diagnosis of certain diseases.

10.2.1.2 Blood pressure

Blood pressure was measured using an automated sphygmomanometer (Omron Digital Blood Pressure Monitor HEM-907, Omron Healthcare, Europe) on the left arm according to the guidelines set by the British Hypertension Society (O?Brien et al., 2003). The subject was seated for at least five minutes prior to measurement, relaxed and not moving or speaking. The arm was supported at the level of the heart with no tight clothing constricting the arm.  The cuff was placed with the indicator mark on the cuff over the brachial artery. The bladder encircled at least 80% of the arm (but not more than 100%). Two readings of blood pressure were taken.

Raised blood pressure was defined as systolic blood pressure ?140mmHg and diastolic blood pressure ?90mmHg (WHO/ISH 2003).


10.2.1.4 Blood lipids and glucose

Random non-fasting samples of blood glucose and blood lipids were taken. Finger prick samples of capillary blood (35 ?l) were analysed to measure non-fasting total cholesterol, total triglycerides, and HDL and glucose levels using an automated cholesterol meter (Cholestech LDX, Chol.


10.2.1.4 Blood lipids and glucose

Random non-fasting samples of blood glucose and blood lipids were taken. Finger prick samples of capillary blood (35 ?l) were analysedto measure non-fasting total cholesterol, total triglycerides, and HDL and glucose levels using an automated cholesterol meter (Cholestech LDX, Cholestech Corporation California). 
High blood cholesterol was defined as total cholesterol >5.2mmol/l. Increased low density lipoprotein cholesterol was defined as > 3.0mmol/l and decreased High density lipoprotein cholesterol as <1.3mmol/l (WHO 1978). Blood glucose outside of the range 3.0-7.0mmol/l was defined as abnormal (WHO 1999). Triglycerides as >1.7mmol/l (IDF ).


10.2.2 Statistical analysis

Data were tested for normal distribution.  Among females, systolic and diastolic blood pressure, blood glucose, total cholesterol, LDL cholesterol and triglycerides were not normally distributed. Glucose and triglycerides remained not normally distributed after log transforming. Krusal-Wallis Test was used to examine differences in systolic and diastolic BP, blood glucose and triglycerides. Among males blood glucose, total cholesterol and triglycerides were not normally distributed. These values were log transformed and only blood glucose was normally distributed. The Krusal-Wallis test was used to examine differences in total cholesterol and triglycerides. All statistical analysis was performed using SPSS.


10.3 Results

The results are presented in two sections. The first section shows results of a questionnaire that assessed history of health screening, identified any existing illness and reported smoking status. In the following section, results of blood pressure, lipids and glucose are reported. 


10.3.1 Previous health history and history of health checks

Table 10.2 shows analysis of health history and previous diagnosis of disease and smoking status.
Table 10. 2 Previous blood pressure, lipid and glucose checks
RuralUrbanUniversityM (55)F (108)M (8)F (13)M (11)F (5)Last BP check (%)Never46.424.8-5.990.9100Past month5.410.125.029.4--Past 3months8.96.4 37.511.8--Past 6months8.99.212.511.8--Past 1year10.721.125.023.59.1-Other19.628.4-17.6--Last cholesterol check (%)
Never10010062.582.410080.0Past month- ----Past 3months-----Past 6months-----Past 1year--12.55.9--Other--25.011.8-20.0Last blood glucose check (%)Never96.199.110088.2100100Past month-0.9---Past 3months-----Past 6months-----Past 1year1.8--5.9--Other1.8--5.9--Smoking status (%)Non-smoker*71.492.775100100100Smoker*28.67.325---Previous** diagnosis (%)Yes16.115.612.547.19.1-No83.984.487.552.990.9100*Never smoked, *current smoker, ** pre-existing chronic disease or any other ailment.	


10.3.1.1 Last blood pressure check

University students were least likely to have had a blood pressure test.  A large proportion of rural men (46.4%) and women 24.8% had never had a blood pressure test. In the last 12months 12.5 of Urban men and 5.9% of Urban women had a cholesterol test. Within the past 5 years 25% of Urban men and 11.8% of Urban women had a cholesterol test

10.3.1.2 Last blood cholesterol check

Blood lipid and cholesterol checks were not common in this population. All rural men and women, and university men had never had their cholesterol checked. A small proportion of university females (20%) had a cholesterol, but not within the past year. Twelve and a half % (12.5%) of Urban men had a cholesterol test in the past year and a further 25% in 5years. Among women, 5.9% had a cholesterol check in the past year and a further 11.8% within the last 5 years.

10.3.1.3 Last blood glucose check

All university students had never had a blood glucose test. In the past month 0.9% of rural women had a blood glucose test. In the past 12 months 1.8% of rural men and 5.9% of Urban women had a blood glucose test. In the past 5 years 1.8% of rural men and 5.9% of Urban women had a blood glucose test.


10.3.1.4 Smoking status

None of the university students smoked or had previously smoked. Among rural men 28.6% were current smokers and 25% of Urban men were current smokers. Amongst women, 7.3% of rural women smoked. 
Among men, cigarettes were the most common form of tobacco smoked. Rural males also smoked pipes and used snuff taken through the nose. All Urban males smoked cigarettes. Rural women took tobacco as snuff and smoked rolled up leaves common in rural Matabeleland.


10.3.1.5 Previous diagnosis of disease

A low number of participants reported previous diagnosis of disease or any underlying conditions. Among rural men and women 16.1% and 15.6% respectively reported previous diagnosis of disease. These included, high blood pressure, diabetes, asthma, pneumonia and stomach ulcers. University men and women had a low prevalence of previous disease, with only one university male reporting a previous diagnosis of bone disease. Urban women reported the highest diagnosis of previous disease (47.1%) the most common disease being hypertension. Among men, rural men reported the highest previous disease with the most common disease being hypertension followed by asthma.


10.3.2 Blood pressure, lipids and glucose

Table 10.3 shows the mean values for systolic and diastolic blood pressure, blood lipids (HDL, LDL and triglycerides) and blood glucose.

Table 10. 3 Blood lipids, pressure and glucose in rural and urban men and women 

RuralUrbanBorrowdaleUniversityGender (n)M (55)F (108)M (8)F(13)M (28)F (20)Mean systolic BP122.7 (18.8)122.9 (18.8)132.9 (11.9)135.2 (28.3)129.2  (8.36)121.4 (7.54)Mean Diastolic BP
75.6 (12.0)81.9 (12.3)81.8 (11.0)87.0 (17.6)93.4 (10.3)85.0 (8.05)% with raised BP (>140/90)1414.837.558.321.40N478761199Non-fasting blood glucose5.37 (1.24)5.63 (1.29)5.73 (1.12)6.14 (1.61)4.91 (0.75)4.41 (0.92)% with abnormal glucose out side range6.311.616.71000Total cholesterol3.48 (0.91)3.74 (.97)4.2 (0.79)4.20 (1.30)3.28 (0.57)4.13 (0.88)% with inc cholesterol <5.2mmol/l8.24.716.710011.2LDL2.29 (0.76)2.25 (0.88)2.29 (0.25)2.48 (1.61)1.91 (0.65)2.17 (0.76)% with LDL>3.0mmol/l18.815.3022.2012.5HDL0.96 (0.48)1.06 (0.39)0.95 (0.29)1.26 (0.32)1.17 (0.40)1.41 (0.27)% with HDL<1.3mmol/l87.578.816.733.366.725Triglycerides0.95 (0.45)1.16 (0.90)2.11 (1.39)2.64 (2.50)0.82 (0.31)1.90 (2.25)% with inc Tri >1.7mmol/l8.514.05027.3025


10.3.2.1 Blood pressure

Amongst the women, there were no significant differences in mean systolic  (p=0.2) and diastolic (p=0.3) blood pressure. However, 58.3% of Urban women had high blood pressure (above 140/90mmHg). Amongst men, there were no significant differences in systolic blood pressure (p=0.07). However, there were differences in diastolic blood pressure (p=0.0001). Rural males had lower diastolic blood pressure 75.6 (12.0) compared with Urban 81.8 (11.0) and University 93.4 (10.3) men.

10.3.2.2 Non-fasting blood glucose

There were significant differences in blood glucose (p=0.003). University women had significantly lower blood glucose 4.41 (0.92) compared with Urban 6.14 (1.61) and rural 5.63 (1.29) women. In addition, 11.6% of rural women and 10% of Urban women had glucose outside of the normal range (3-7mmol/l). There were no significant differences in blood glucose amongst men (p=0.4).

10.3.2.3 Non-fasting total cholesterol

There were no significant differences in total cholesterol amongst women (p=0.2). Urban men had significantly higher (p=0.05) total cholesterol compared with rural and university men. More Urban men (16.7%)  had raised cholesterol compared with rural and university men.  More University women (11.2%) had raised cholesterol compared with Urban and rural women. 
10.3.2.3.1 LDL cholesterol
There were no significant differences in LDL cholesterol in women (p=0.9) and men (p=0.4). More Urban women 22.2% had increased LDL cholesterol compared with university and rural women. Amongst men, 18.8% of rural men had raised levels of LDL cholesterol, whereas no university or Urban men had raised LDL-C. 

10.3.2.3.2 HDL cholesterol
There were significant differences in HDL cholesterol (p=0.02), University women, had significantly higher HDL cholesterol 1.41 (0.27) compared with rural 1.06 (0.39) and Urban 1.26 (0.32) women. There were no significant differences in HDL cholesterol (p=0.6) amongst men. Mean HDL cholesterol was higher in women than in men in all groups. In addition, more rural men and women had decreased HDL cholesterol, 87.5 and 78.8% respectively, compared with Urban men and women, 16.7 and 33.3% and university men and women 66.6% and 25% respectively.

10.3.2.3.3 Triglycerides
There were no significant differences in triglycerides in women (0.2). Borrowdale men had significantly high triglycerides (p=0.004) compared with rural and university men. More Urban men had increased triglycerides (50%) than rural or university men. More Urban women had increased triglycerides (27.3%) than rural (14%) and university (25%) women.	


10.4 Discussion

The aim of this study was to compare blood pressure, blood glucose and lipid profile in rural and urban Zimbabweans. The main findings show that values for these risk factors was low in this population, and particularly low for blood lipids and glucose. Proportionally more people from Urban had raised blood pressure, lipids and blood glucose.  Overall prevalence of raised blood lipids and glucose was low, however, the prevalence of decreased HDL-C was high in this population.


10.4.1 Risk factor screening

Urban men and women, were more likely to have received blood pressure, lipids and glucose tests compared with rural and university students. This may be a result of greater access to healthcare in urban compared with rural areas in Zimbabwe. Urban populations in many developing countries have greater access to healthcare facilities than their rural counterparts (Castro-Leal et al., 2000). Moreover, urban residents are more likely to afford healthcare compared to rural residents. High levels of unemployment in rural areas mean that the healthcare is largely the burden of the state. However, resources are limited, and a larger proportion of government budgets are allocated to the more prevalent infectious disease such as HIV/AIDS, TB and malaria. This finding is not unique to Zimbabwe, but is symptomatic of many low-income developing countries experiencing the double burden of disease (WHO/FAO 2002). Beran and Yudkin (2006) reviewed studies evaluating the cost of diabetes care in 25 sub-Saharan African countries, and found that although the incidence of diabetes was increasing and the prevalence in some areas was high, a small amount of the money was spent on insulin provision, and therefore insulin was often unavailable in the large city hospitals, and regularly available in rural areas in only ?ve countries. This shows that treatment of non-communicable disease is not a priority in many governments in developing countries.

10.4.2 Differences in blood pressure, lipids and glucose 

It was expected that proportionally more urban Zimbabweans would have raised blood pressure, lipids and blood glucose compared with thes. This shows that treatment of non-communicable disease is not a priority in many governments in developing countries.

10.4.2 Differences in blood pressure, lipids and glucose 

It was expected that proportionally more urban Zimbabweans would have raised blood pressure, lipids and blood glucose compared with their rural counterparts. UrbMufunda et al., 2006). The higher risk factors are a result of changes to diet and lifestyle associated with urbanisation. Urban diets are higher in fats, salts, sugars and lower in fibre compared with rural diets. In addition, people living in urban areas are increasingly sedentary as a result of mechanisation of work and leisure, and motorisation of transport. 
The finding of increased blood pressure in urban compared with rural populations is consistent with reports in other Zimbabwe populations. In the ZiNCoDs report (2005), it was reported that more urban men and women were hypertensive compared with the rural population. In addition, Addo et al., (2007) showed that in the data of 37 countries in Africa, urban residents consistently had higher prevalence of hypertension compared with rural residents.
Overall levels of blood lipids were low in this population, although Urban men and women had higher mean total cholesterol compared with rural and urban women. Interestingly the levels of low HDL-C were prevalent in all groups. This finding is in contrast to that in other populations of African origin where it has been shown that high levels of HDL-C offer protection against coronary heart disease (Gaillard et al., 2009, Sumner 2009). Gomo (1985) suggested a range for total cholesterol, HDL-C and triglycerides for Zimbabweans.  Comparisons of the ranges suggested by Gomo (1985) with the sample in the current study (table 10.4) showed that mean total cholesterol and HDL-C were higher in the Gomo (1985) sample. Triglycerides in the Urban sample were higher and outside the range suggested by Gomo. This implies that the levels of total cholesterol and HDL-C reported in this study are not outside the acceptable ranges for Zimbabweans.  The high triglycerides in the Urban group, may be a consequence of increased dietary fat compared with the rural population and the Gomo sample.


Table 10. 4 Comparison of total cholesterol, high density lipoprotein and triglycerides means with ranges suggest by Gomo (1985) 
MenWomenTotal cholesterol range1.95-6.991.73-6.81Gomo (1985) 4.5 (1.26)4.27 (1.26)Rural3.48 (0.91)3.74 (0.97)Urban4.2 (0.79)4.2 (1.3)University3.28 (0.57)4.31 (0.58)HDL-C range0.78-2.300.87-2.42Gomo (1985)1.54 (0.46)1.67 (0.40)Rural0.96 (0.48)1.06 (0.39)Urban0.95 (0.29)1.26 (0.32)University1.17 (0.40)1.41 (0.27)Triglycerides range0.49-1.450.40-1.36Gomo (1985)0.93 (0.26)0.86 (0.23)Rural0.95 (0.45)1.16 (0.90)Urban2.11 (1.39)2.64 (2.50)University0.82 (0.31)1.90 (2.25)HDL-C= high density lipoprotein cholesterol

More Urban than rural men and women had elevated blood glucose that fell outside the ranges for normal. However, overall prevalence of abnormal blood glucose was low in this population. These findings are consistent with those reported in other Zimbabwean populations. The ZiNCoDs (2005) reported a very low prevalence of elevated blood glucose measure as a fasting blood glucose range of (7.8mmol/l. It has been reported that the global prevalence of diabetes, which is caused by elevated blood glucose, was predicted to double between the years 2000 and 2030 (Wild et al., 2004). The greatest increases will be observed in developing countries experiencing rapid urbanisation and increases in life expectancy. Therefore, the low levels of elevated blood glucose in Zimbabwe may not remain at this level.


10.4.3 Study limitations

In this study the prevalence of hypertension, elevated blood glucose and lipid levels were defined on the basis of measures taken on one visit only. This may lead to over/underestimated prevalence of any of the conditions. In addition, non-fasting blood glucose and blood cholesterol were taken as it was not possible for fasting samples to be collected. However, despite this, these data provide good baseline evidence of these risk factors in rural populations compared with urban populations. 


10.5 Conclusion

The aim of this chapter was to compare blood pressure, blood glucose and lipid profile in rural and urban Zimbabweans. It was shown that urban populations had greater access to health care than the rural population. However, hypertension was higher in the urban group. Levels of blood cholesterol were low in this population, which appears to be a feature of Zimbabwean populations.


10.6 References


Addo, J., Smeeth, L., Leon, D.A., (2007). Hypertension in Sub-Saharan Africa. A Systematic Review. Hypertension, 50 pp.1012-1018.

BeLue, R., Okoror, T.A., Iwelunmor, J., Taylor, KD., Degboe, A.N., Agyemang, C.,  and Ogedegbe, G., (2009). An overview of cardiovascular risk factor burden in sub-Saharan African countries: a socio-cultural perspective 
Globalization and Health, 5:10

Boutayeb, A., (2006). The double burden of communicable and non-communicable diseases in developing countries. Transactions of the Royal Society of Tropical Medicine and Hygiene 100, pp.191?199.
Castle, W. M., (1982). Coronary heart disease risk factors in black and white men in Zimbabwe and the effect of living standards.  South African Medical Journal 61, pp.926-929. 
Castro-Leal, F., Dayton, J., Demery, L., Mehra, K., (2000). Public spending on health care in Africa: do the poor benefit? World Health Organization, Bulletin of the World Health Organization, 78, pp.66-74.

Danaei, G., Lawes, C.M.M., Hoorn, S.V., Murray, C.J.L., Ezzati, M., (2006). Global and regional mortality from ischaemic heart disease and stroke attributable to higher-than-optimum blood glucose concentration: comparative risk assessment. The Lancet,  368,  pp.1651 ? 1659.

Gaillard, T., Schuster, D., Osei, K., (2009). Metabolic syndrome in black people of African diaspora: The paradox of current classification, definition and criteria. Ethnicity and Disease, 19, pp.S2-1 ? S2-7.
Gomo, Z. A. A., (1985). Concentrations of lipids, lipoprotein, and apolipoproteinsin serum of Zimbabwean Africans. Clinical Chemistry, 31, pp.1390-1392.
Kannel, W.B., Castelli, W, P., Gordon, T., Mcnamara, P.M., (1971). Serum Cholesterol, Lipoproteins, and the Risk of Coronary Heart Disease: The Framingham Study. Annals of Internal Medicine, 74, pp.1-12.

McQueen, M. J., Hawken, S., Wang X., Ounpuu S., Sniderman A., Probst?eld J., Steyn, K., Sanderson, J. E., Hasani, M., Volkova E., Probst?eld J., Steyn, K., Sanderson, J. E., Hasani, M., Volkova E., Kazmi K., and Yusuf, S., (2008). Lipids, lipoproteins, and apolipoproteins as risk markers of myocardial infarction in 52 countries (the INTERHEART study): a case-control study. The Lance 372 pp.224?233.

Misra, A., and Khurana, L.,(2008). Obesity and the metabolic syndrome in developing countries. Journal of Clinical Endocrinology and Metabolism, 93 pp.S9 ?S30.

Mufunda, J., Scott, L.J., Chifamba, J., Matenga, J., Sparks, B., Cooper, R., Sparks, H., (2000). Correlates of blood pressure in an urban Zimbabwean population. Journal of Human Hypertension, 14, pp.65?73.

Mufunda, J Chatora, R Ndambakuwa  Y; Nyarango  P, Kosia, A; Chifamba J Filipe A, Usman A,. Sparks, H. V., (2006). Emerging non-communicable disease epidemic in Africa: preventive measures from the WHO regional office for Africa. Ethnicity and Disease. 16 pp.521?526.

Murray, C.J.L., Lauer, J.A., Hutubessy, R.C.W., Niessen, L., Tomijima, N., 
Rodgers, A., Lawes ,C.M.M., Evans, D.B.. (2003). Effectiveness and costs of interventions to lower systolic blood pressure and cholesterol: a global and 
regional analysis on reduction of cardiovascular-disease risk.  The Lancet, 
361, pp.717?725.

Njelekela, M., Sato, T., Nara,Y.,  Miki T., Kuga, S., Noguchi, T., Kanda, T., Yamori, M.,Ntogwisangu, J., Masesa, Z., Mashalla, Y., Mtabaji, J., Yamori Y.,(2003). Nutritional variation and cardiovascular risk factors in Tanzania ? rural-urban difference. South African Medical Journal, 93, pp.295-299.

O'Brien, E., Asmar, R., Beilin, L., Imai, Y., Mallion, J.M., Mancia, G., Mengden, T., Myers, M., Padfield, P., Palatini, P., Parati, G., Pickering, T., Redon, J., Staessen, J., Stergiou, G., Verdecchia, P., (2003). European Society of Hypertension recommendations for conventional, ambulatory and home blood pressure measurement. Journal of Hypertension, 21, pp.821-848.
Omran, A.R., (1971). The epidemiological transition: a theory of epidemiology of population change. Milbank Quarterly, 49, pp.509? 538.

Reddy, K.S., (2002). CVDs in the developing countries: dimensions, determinants, dynamics and directions for public health action 
Public Health Nutrition: 5, pp.231?237.
Sparks, B.T., Mufunda, J., Musabayane, C.T., Sparks, H.V., Mahomed, K., Hunter, J.M., (1996). Prevalence of hypertension among women in rural Zimbabwe: a comparison of pregnant and non-pregnant women. Central Africa Journal of Medicine, 42, pp.93?97.

Vorster, H.H., (2002). The emergence of CVD during urbanisation of Africans. Public Health Nutrition. 5, pp.239 ?243.

Walker, A. R. P.  and Sareli, P., (1997). Coronary heart disease: outlook for Africa. Journal of the Royal Society of Medicine, 90 pp.23-27.

Wild, S., Roglic, G., Green, A., Sicree, R., King, H., (2004). Global prevalence of diabetes estimates for the year 2000 and projections for 2030. Diabetes Care, 27, pp.1047?1053.
Wilson, A.O., and Nhiwatiwa, R., (1992). Does Zimbabwe need geriatric services? Central African Journal Medicine 38, pp14-16.
World Health Organization/ International Society of Hypertension (2003). Statement on management of hypertension. Journal of Human Hypertension, 21, pp.1983-1992
World Health Organization (1978). Expert Committee. Arterial Hypertension (Technical report Series No. 628). Geneva: World Health Organization 

World Health 

World Health Organization (1999). De?nition, Diagnosis and Classi?cation of Diabetes Mellitus and its Complications; Part 1: Diagnosis and Classi?cation of Diabetes Mellitus. Geneva: Department of Non-communicable Disease Surveillance, Report No.: WHO

World Health Organization and Food and Agricultural Organization (2002) Diet, nutrition and the prevention of chronic diseases; Report of a joint WHO/FAO expert commission WHO Technical Report Series 916. Geneva: WHO
ADIPOSITY AND CARDIOVASCULAR RISK FACTORS IN ADULTS OF AFRICAN ORIGIN

11.1 Introduction


The prevalence of CVD varies with ethnicity (Scarborough et al., 2010) and this variation may be as a result of differences in risk factors between groups (Must et al., 1999, Kurian and Cardelli 2007). For example, the prevalence of coronary heart disease (CHD) is low in populations of African origin. This is believed to be a result of the low prevalence of risk factors for CHD e.g. low total cholesterol and low ratio of low density lipoprotein cholesterol (LDL-C) to high density lipoprotein cholesterol (HDL-C). More importantly high HDL-C, which is associated with protection against the development of heart diseases is increased in African compared with Caucasian origin populations (Agyemang et al., 2009). However, compared with Caucasians, the prevalence of stroke and renal failure is high in many populations of African origin, this is a result of the high prevalence of risk factors for stroke, mainly hypertension. 

Overweight and obesity are defined as the excessive accumulation of adipose tissue that may impair health (World Health Organization (WHO) 1998). Must et al., (1999) showed that for both men and women in the USA, high blood pressure was the most common?overweight- and obesity-related health condition and its prevalence?showed a strong increase with increasing BMI status. The prevalence of type 2 diabetes mellitus (T2DM), gallbladder disease,?and osteoarthritis increased among both overweight and?obese men and women corresponding with the increasing BMI category. The prevalence of increased blood cholesterol level was high in both genders but showed no increase in prevalence with increasing?BMI category. However, men and women with BMIs of 25 kg/m2 or more, were more likely than persons of normal weight to have?high blood cholesterol levels. 

In a meta-analysis of 21 studies investigating the association of overweight and obesity with risk factors for CHD, Bodgers et al., (2007) reported that a total of 18 000 CHD events occurred during follow-up. The age-, sex-, physical activity-, and smoking-adjusted relative risks (RR) for moderate overweight and obesity compared with normal weight were 1.32 (1.24-1.40) and 1.81 (1.56-2.10), respectively.  It was concluded that adverse effects of overweight on blood pressure and cholesterol levels could account for about 45% of the increased risk of CHD.

Morbidity and mortality estimates associated with adiposity have been used to define cut-off points for BMI (Lew and Garfinkel 1979). Pan et al., (2004) compared the relationship between metabolic comorbidities (hypertension, hypercholesterolemia, diabetes, hypertriglyceridemia and hyperuricemia) among Asians and US Caucasians and AAs. The odds ratios for each of the selected conditions were calculated for each increment of BMI. Increments of BMI corresponded to significantly higher odds ratios in the Taiwanese than in the AAs for all five metabolic conditions studied. When the Taiwanese and the Caucasians were compared, higher odds ratios were shown for hypertriglyceridemia and hypertension, respectively. Thus, it was concluded that different BMI cut-off points were needed for Taiwanese compared with Caucasians and AAs. 
Stevens et al., (2002) attempted to establish BMI cut-off points in a group of AA women against a group of Caucasian women at a BMI of 30kg/m2, using the incidence of diabetes, hypertension and hypertriglyceridemia as indicators of disease. It was found that a large range in BMI values for AA women was associated with risks equivalent to those of the Caucasian women at a BMI of 30kg/m2. In addition, BMI cut-off points differed for the AA women in diabetes, compared to hypertension compared to hypertriglyceridemia equivalent to the Caucasian women at a BMI of 30kg/m2 (Stevens et al., 2002). Therefore, BMI may not represent the same level of risk in all ethnic groups.


Given that risk factors for CVD vary between populations of Caucasian and African origin, it is possible there severity of the risk factors associated with increasing BMI may differ between African compared with Caucasian origin populations. Therefore, the aim of this study was to identify the severity of risk factors associated with CVD in people of African origin at different levels of adiposity defined by BMI.


11.2 Methods

11.2.1 Subjects 
Participants included urban and rural Zimbabwean men and women, the characteristics of whom have been described in chapter eight of this thesis. These data were pooled together with people of African origin living in the UK, the details of whom are described in chapter six of this thesis. Therefore, the participants were men and women of differing levels of urbanisation.

11.2.2 Outcome measures

11.2.1.1 Anthropometric measures

Procedures for anthropometric measures are described in chapter eight.

11.2.1.2. CVD risk factors

Procedures for the collection of biochemical CVD risk factors are described in chapter nine 


11.2.2 Statistical analysis
Normality was assessed using the Kolmogorov-Smirnov test. Amongst men age, blood glucose, total cholesterol, HDL-C and triglycerides were not normally distributed. Among females, age, height, weight, systolic blood pressure, diastolic blood pressure, blood glucose, cholesterol, LDL and triglycerides were not normally distributed. For the normally distributed variables, differences between groups were tested using ANOVA for female groups and the students? t test for males. For the variables that were not normally distributed non-parametric tests were used. Odds ratios were calculated for the relationship between adiposity, defined by BMI and risk of high blood pressure, raised cholesterol, triglycerides and blood glucose. Statistical analysis was conducted using SPSS.


11.3 Results
Subject characteristics are shown in table 11.1 for this population of people of African origin, stratified according to BMI. 

Table 11. 1 Subject characteristics of African men and women
Risk factors (%)Normal
18.5-24.9kg/m2Overweight
25.0-29.9kg/m2Obese I
<30.0kg/m2Gender (n)M (70)F (88)M (15)F (43)F (18)Age (years)*31.7 (13.6)34.8 (13.9)45.4 (12.1)41.8 (12.3)45.4 (12.1)Height (cm) 1.73 (0.05)1.6 (0.06)1.59 (0.06)1.61 (0.05)1.59 (0.06)Weight (kg)* 64.3 (6.7)56.6 (6.4)85.4 (12.4)70.7 (5.5)85.4 (12.4)BMI (kg/m2)* 21.4 (1.6)22.1 (1.73)33.7 (4.5)27.0 (1.32)33.7 (4.5)WC* 76.2 (6.03)72.5 (6.5)95.2 (9.2)83.2 (6.4)95.2 (9.2)SYS BP (mmHg)127.0 (16.8)122.3 (18.8)122.5 (14.0)125.0 (18.8)122.5 (14.0)DIA BP (mmHg)81.7 (14.8)80.4 (12.5)82.0 (19.0)84.0 (14.7)82.0 (19.0)(Number)(46)(74)(14)(26)(14)Blood glucose (mmol/l)5.50 (1.2)5.61 (1.40)6.25 (2.0)5.46 (1.39)6.25 (2.0)Total cholesterol (mmol/l)3.56 (0.91)3.92 (1.16)3.79 (0.53)3.93 (0.82) 3.79 (0.53)LDL (mmol/l)2.20 (0.76)2.33 (1.07)1.81 (0.87)2.23 (0.79)1.81 (0.87)HDL (mmol/l)1.01 (0.53)1.16 (0.41)1.39 (0.38)1.10 (0.44)1.39 (0.38)Triglycerides (mmol/l)1.04 (0.67)1.38 (1.31)1.96 (1.88)1.45 (1.15)1.96 (1.88)SYS BP=Systolic blood pressure, DIA BP= Diastolic blood pressure, *Significant differences between groups p<0.05

Normal weight males were significantly younger (p=0.0001) than overweight males. Overweight males were significantly heavier (p=0.0001) and had significantly larger WC (p=0.0001) than normal weight males. There were no significant differences in height between the females and males. There were no significant differences in mean systolic and diastolic blood pressure, blood glucose, total cholesterol, HDL-C and triglycerides between males in normal, overweight and obese categories.
Normal weight females were significantly younger than overweight and obese females. There were no significant differences in age between overweight and obese females [F (2, 146) =7.2] p=0.0001. Overweight and obese females were significantly heavier than normal weight females [F (2, 146) = 148.0] (p=0.0001). Obese females were significantly heavier than overweight females. There were no significant differences in height between the females and males. Waist circumference was significantly larger in obese compared with overweight and normal weight females [F (2, 1460=97.6) p=0.0001]. Overweight females had significantly larger WC than normal weight females. 
There were no significant differences in mean systolic and diastolic blood pressure across the groups. There were no significant differences in blood glucose, total cholesterol, HDL-C and triglycerides between men and women in normal, overweight and obese categories.
Table 11.2 shows odds ratios of increased risk of increased WC, hypertension, blood glucose outside the normal range, increased cholesterol, decreased HDL-C, increased LDL-C and increased triglycerides in overweight and obese groups compared with normal weight.


Table 11. 2 Odds ratios for cardiovascular risk factors by body mass index category 
Weight categoryOverweight*
25.0-29.9kg/m2Obese *
<30.0kg/m2Gender (n)M (15)F (43)F (18)Increased WC45.3 4.24 108.8 Hypertensive0.32 0.92 1.14 Outside normal range (blood glucose)0.55 2.38 2.4 Increased cholesterol2.33 0.19 0 Decreased HDL-C1.54 0.40 1.95 Increased LDL-C1.54 4.18 11.15 Increased triglycerides0 1.72 3.58 *Odds ratio compared with normal weight
	
Compared with normal weight males, overweight males were 45.3 times more likely to have an increased WC, 0.32 times more likely to be hypertensive, 0.55 times more likely to have blood glucose outside the normal range, 2.33 times more likely to have increased cholesterol, 1.54 times more likely to have decreased HDL-C and 1.54 times more likely to have increased LDL-C. Overweight and normal weight males are at an equal risk of increased triglycerides. 

Compared with normal weight females, overweight females were 4.24 times more likely to have an increased WC, 0.92 times more likely to be hypertensive, 2.38 times more likely to have blood glucose outside the normal range, 0.19 times more likely to have increased cholesterol, 0.4 times more likely to have decreased HDL-C, 4.18 times more likely to have increased LDL-C and 1.72 times more likely to have increased triglycerides.

Compared with normal weight females, obese females were 108 times more likely to have an increased WC, 1.14 times more likely to be hypertensive, 2.4 times more likely to have blood glucose outside the normal range, 1.95 times more likely to have decreased HDL-C, 11.15 times more likely to have increased LDL-C and 3.58 times more likely to have increased triglycerides. Obese females were not different in risk of increased cholesterol compared with normal weight females.

11.4 Discussion

Overweight and obesity increases the risk of CVDs, independently and through their effect on other risk factors for CVD. The risk factors for CVD differ between people of African and Caucasian origin. Therefore, it is possible that the severity of CVD risk factors associated with overweight and obesity may differ between African origin and Caucasian origin groups. In this study, risk factors for CVD including hypertension, increased WC, abnormal blood glucose, increased LDL-C and decreased HDL-C were progressively worse in overweight and obese compared with normal weight groups. However, amongst women there were no differences in increased cholesterol between BMI categories, and amongst men, there were no differences in increased triglycerides between BMI categories. 
Amongst men, increased WC and increased cholesterol showed the largest difference between normal weight and overweight. Amongst women increased WC, increased LDL-C and increased triglycerides showed the largest differences between normal, overweight and obese categories. The INTERHEART Africa study showed that only five risk factors account for 89.2% of the risk for an initial myocardial infaction (MI) in populations of African origin. These include hypertension, smoking, diabetes, abdominal obesity and dyslipidaemia (Styen et al., 2005, Mensah 2008). Therefore this may explain why WC, total cholesterol, LDL-C and triglycerides showed the largest differences with increasing overweight. However, interestingly, the likelihood of increased blood pressure with increasing BMI was not significant between groups. This is unexpected as it is well established that increased blood pressure is a strong risk factor for CVD in populations of African origin (Agyemang et al., 2009). Since the majority of the people in this study comprised of rural Africans from Nkayi district in Matabeleland North in Zimbabwe. This group had low prevalence of risk factors fro CVD. This may explain why there no large differences in the likelihood of increased blood pressure between normal weight and overweight individuals. This is particularly relevant for males.

Body mass index is used to represent overweight and obesity and the comorbities associated with excess fat. Pan et al., (2004) calculated the odds ratios for 5 conditions including hypertension, diabetes, hypertriglyceridemia, hypercholesterolemia and hyperuricemia in AAs, Taiwanese and Caucasian Americans. The odds ratios for these groups were compared with the group in this study for the available measures as shown in table 11.3


Table 11. 3 Comparison of odds ratios for CVD risk factors in different ethnic groups
TaiwaneseCaucasiansAfrican AmericanOverweight males*Overweight females*Hypertension1.24 
1.11
1.07 
0.32 
0.92
 Diabetes/ abnormal blood glucose1.13 
1.15 
1.08 
0.55 2.38 Hypercholesterolemia1.10
 1.05 
1.02
 2.33 0.19 Hypertriglyceridemia1.20 
1.11
 1.05 
0 1.72 Hyperuricemia1.16 
1.13 1.09
 --*participants from the current study
Compared with Taiwanese, Caucasians and AAs, the odds of hypertension, diabetes/ abnormal blood glucose in both males and females was lower in this study population. However, hypercholesterolemia was higher in men but not women, and hypertriglyceridemia was higher in females but not men. Therefore in the population, overweight was associated with greater severity in lipid disturbances compared with Taiwanese, Caucasians and AAs.

11.4.1 Study limitations

The data from rural and urban Zimbabweans was pooled together with the data of African origin participants from the UK. This population is different in terms of socio-economic status, education, diet and lifestyle compared with the Zimbabwean group. However, it was necessary to pool these data to allow for comparison across the three levels of BMI. In addition, the UK group provides an additional level of urbanisation and population heterogeneity.


11.5 Conclusion

The prevalence of CVD and risk factors for CVD vary between populations of African and Caucasian origin. The aim of this study was to access the severity of risk factors associated with CVD in people of African origin at different levels of adiposity defined by BMI. It was shown that risk factors for CVD increasing with increasing BMI in a population of African origin. In addition, when compared with Taiwanese, Caucasians and AAs, the odds ratio for abnormal glucose was higher in overweight females and the odds ratio hypertriglyceridemia was higher in overweight males


11.6 References

Agyemang, C., Addo, J., Bhopal, R., Atkins, A.D.G., Stronks, K., (2009). CVD, diabetes and established risk factors among populations of sub-Saharan descent in Europe: a literature review. Globalization and Health 5, pp.

Bogers, R.P., Bemelmans, W.J.E., Hoogenveen, R.T., Boshuizen, H.C., Woodward, M., Knekt, P., Van Dam, R.M.,. Hu, F.B., Visscher, T.L.S., Menotti, A., Thorpe Jr, R.J., Jamrozik, K., Calling, S., Strand, B.H., Shipley, M. J., (2007). Association of Overweight With Increased Risk of Coronary Heart Disease Partly Independent of Blood Pressure and Cholesterol Levels. A Meta-analysis of 21 Cohort Studies Including More Than 300 000 Persons.
Archives of Internal Medicine, 167 pp.1720-1728.
Kurian, A.K., Cardarelli, K.M., (2007). Racial and ethnic differences in CVD risk factors: A systematic review. Ethnicity and Disease, 17, pp.143-152.

Mensah, G.A., (2008). Global burden of CVD: Ischaemic heart disease in Africa. Heart, 94, pp.836-843. 
Misra, A., Khurana, L., (2008). Obesity and the metabolic syndrome in developing countries. Journal of Clinical Endocrinology and Metabolism, 93, pp.s9-s30. 

Must, A., Spandano, J., Coakley, E.H., Field, A.E.,  Colditz, G., Dietz, W.H., (1999). The disease burden associated with overweight and obesity, Journal of the American Medical Association, 282 pp.1523-1529.

Pan, W-H., Flegal, K.M., Chang, H-Y., Yeh, W-T., Yeh, C-J.,  and Lee, W-C.,  (2004). Body mass index and obesity-related metabolic disorders in Taiwanese and US whites and blacks: implications for definitions of overweight and obesity for Asians. American Journal of Clinical Nutrition, 79 pp.31?39.

Scarborough, P., Bhatnagar, P., Kaur, A., Smolina, K., Wickramsinghe, K., Rayner, M., (2010). Ethnic differences in CVD 2010 edn. British Heart Foundation Statistics Database. British Heart Foundation Health Promotion Research Group

Steyn, K., Sliwa, K.,  Hawken, S., Commerford, P., Onen, C.,  Damasceno, A., Ounpuu, S., Yusuf, S.,  (2005). Risk factors associated with myocardial infarction in Africa: The INTERHEART Africa Study. Circulation, 112, pp.3554-3561.

Stevens, J., Juhaeri, C. J., Jones, D., (2002). The effect of decision rules on the choice of a body mass index cutoff for obesity: examples from African American and white women. American Journal of Clinical Nutrition, 75, pp.986?992.
World Health Organization (1998). Obesity: Preventing and managing a global epidemic. Geneva: WHO
Yusuf, S., Hawken, S., ?unpuu, S., Bautista, L.,  Grazia Franzosi, M.,   Commerford,P.,  Lang,C.C.,  Rumboldt,Z., Onen, C.L.,  Lisheng, L.,  Tanomsup, S., Wangai Jr, P.,  Razak, F.,  Sharma, A.M.,   Anand, S. M., (2005). Obesity and the risk of myocardial infarction in 27 000 participants from 52 countries: a case-control study. The Lancet, 366 pp.1640?1649.

THESIS SUMMARY, LIMITATIONS AND RECOMMENDATIONS


12.1 Summary 

The aim of this final chapter is to review the whole thesis and emphasise the main conclusions and recommendations. Thus it is important to first review the overall aims of the thesis. 

To establish the relationship between adiposity, (as measured by air-displacement plethysmography (ADP), bioelectrical impendence analysis (BIA) and anthropometry) and cardiovascular risk factors in different ethnic groups.
To identify field measures of adiposity, relating to cardiovascular risk in different ethnic groups. 
To compare the relationship of adiposity and cardiovascular risk factors in a single ethnic group, that of a rural and an urban population in Zimbabwe.
To identify risk factors for CVD related to adiposity in a population of African origin.


12.1.1 Summary of chapter two

Chapter two reviewed overweight, obesity and related disease. In this chapter the difference in overweight and obesity prevalence in different ethnic groups was highlighted. In addition the proxy measures for overweight and obesity were critically assessed with regards to their application in different ethnic groups.
12.1.2 Summary of chapter three

Chapter three was an historic review of body composition assumptions and the methods derived from the assumptions. The focus was on the implications of these methods when applied to different ethnic groups.
12.1.3 Summary of chapter four

What is known:
The relationship between BMI and percentagebody fat is age and gender dependant.
Its independence from ethnicity is controversial.
What this study adds:
This study demonstrated that the relationship between BMI and %BF was not the same in the ethnic groups used in this section, namely Afro-Caribbean, Asian and Caucasian.  
When the groups are matched for age, gender and BMI, ethnicity become a significant predictor of % body fat.
Matching the groups for age, gender and BMI, enabled far better comparison than was possible with the original heterogeneous groups.
This study reinforces that the universal cut-off points for overweight and obesity are not appropriate in all population groups


12.1.4 Summary of chapter five

What is known:
People of African origin have a greater density of FFM compared with people of Caucasian origin.
The classic two-compartmental model of Siri (1961) assumes a constant density of FFM of 1.100g/cm3 for all population groups and therefore may give biased estimates of %BF in African populations.
Ethnic specific equations assume a higher density of FFM and alternative two compartmental model equations have been developed. 
What this study adds:
This study demonstrated that percentagebody fat estimated by the two-compartmental model of Siri (1961) and the ethnic specific two-compartmental equations of Schutte et al., (1984) and Wagner and Heyward (2001) in Afro-Caribbean men gave significantly different results. 
In it was shown that percentagebody fat estimated by Siri (1961) and the ethnic specific equation of Ortiz et al., (1992) in Afro-Caribbean women were not significantly different.
Density of FFM was significantly lower in Caucasian than Afro-Caribbean men and women.
Compared with a three-compartmental model equation (Siri 1956), there was poor agreement in %BF for all two-compartment equations in men than in women.
Density of FFM derived from estimates of body density from air displacement and TBW from BIA gave unreliable results.
Therefore it is not recommended that TBW estimates from BIA are incorporated into multi-compartment models to be used as reference measures.

12.1.5 Summary of chapter six

What is known:
The prevalence of CVD (CVD) varies between African and Caucasian origin populations.
Differences in risk factors for CVD may account for the differences in disease prevalence.
Clustering of risk factors to form risk measures e.g. Framingham risk score or Metabolic syndrome, developed in Caucasian populations may not appropriately predict CVD risk in non-Caucasian populations.

What this study adds:
This study demonstrated that in Afro-Caribbean men and women matched for age, gender and BMI, to Caucasian men and women, total cholesterol, HDL-C and systolic blood pressure significantly predicted the Framingham risk score, whereas these variables were not significant in the Caucasian group.
It is recommended that different clusters risk measures are developed for non-Caucasian groups.


12.1.6 Summary of chapter seven

Chapter seven is a review, which described the nutrition transition with a particular focus on the changes that have occurred in Zimbabwe. Available data was used to highlight the changes in diet, physical activity associated with urbanisation.


12.1.7 Summary of chapter eight

What is already known:
There are differences in the prevalence of overweight and obesity between rural and urban populations in developing countries.
A limited amount of prevalence data exists in many developing countries including Zimbabwe.
Anthropometry provides a vital tool for describing the overweight and obesity patterns in low resource settings.
What this study adds:
This study increases the evidence of differences in overweight and obesity prevalence, between rural and urban Zimbabweans.
It was shown that urban men and women were significantly heavier, had higher BMI, percentagebody fat and had larger skinfolds compared with rural men and women.
This study showed that women in both urban and rural areas were overweight in Zimbabwe
It is recommended that further studies are carried out to assess the effect of overweight in rural men and women


12.1.8 Summary of chapter nine

What already known:
Differences in energy balance between rural and urban populations in developing countries have been reported.
With urbanisation, energy intake from fats is increased while energy from carbohydrates is reduced.
With urbanisation energy expenditure decreases as a result of sedentary lifestyles
Energy expenditure in rural populations varies with season and these populations may go through periods of energy deficit not experienced by urban populations.

What this study adds:
This study provides further evidence of differences in energy intake, energy sources, energy expenditure between rural and urban populations in Zimbabwe.
It was shown that as the population is urbanised the sources of energy shift, with an increase in energy from fats and a reduction in energy from carbohydrates. 
Both urban and rural populations recorded an energy deficit.
This may be related to the use of a single 24hour recall to estimated daily energy intake
This study reinforces the need for the use of multiple 24hour recalls in order to capture true intake


12.1.9 Summary of chapter ten

What is known:
Changes in diet and physical activity patterns may be associated with changes to blood pressure, blood lipids and blood glucose.
In Zimbabwe a limited amount of data exists on the differences in blood pressure, blood lipids and glucose.
One study suggested ranges for total cholesterol, HDL-C and triglycerides in Zimbabwean men and women. 
The study (Gomo 1985) is twenty years old and may not reflect current trends in Zimbabwe.

What this study adds:
This study provides evidence of differences in blood pressure, lipids, and glucose between rural and urban populations in Zimbabwe.
Values for total cholesterol and HDL-C fell within the suggested ranges for Zimbabweans for both rural and urban men and women. 
Urban men and women had higher triglycerides which fell outside the suggested range
It is recommended that more studies are conducted in large populations living in urban areas to ascertain the change in triglycerides noted in this study.

12.1.10 Summary of chapter eleven

What is known:
Overweight and obesity are associated with increased risk of CVD.
Overweight and obesity act as independent risk factors, but also enhance the effect of other risk factors.
The relationship between overweight, obesity and CVD morbidity and mortality has been used to define cut-off points for BMI.
There are differences in the risk factors of CVD between populations of African compared with Caucasian origin populations.

What this study adds:
This study shows that amongst men increased WC, increased cholesterol, decreased HDL-C, increased LDL-C were more likely in overweight compared with normal weight African men. 
Amongst women increased WC, increased LDL-C and increased triglycerides were more likely in overweight and obese than normal weight  African women.
This study demonstrates differences in the severity of risk factors for CVD in an African origin compared with, Taiwanese, AAs and Caucasians.


12.2 Limitations

During the course of the study the following limitations were recognised
Samples sizes
In chapters eight, nine and ten, it was intended that a similar number of urban and rural participants would be recruited to this study. However, in 2008 while undertaking this study, a cholera epidemic occurred in the urban areas of Zimbabwe, with its epicentre in Harare. Therefore, recruitment of participants was in urban areas was stopped. This significantly influenced the number of urban Zimbabweans recruited. 
The small number of urban participants recruited cannot be considered representative of the majority of urban Zimbabweans. This group represents an affluent and highly educated population. As such these findings cannot be generalised to other urban Zimbabweans. However, this study does show differences in urban and rural population that have been reported in other African countries.
 In chapter eleven, the data from rural and urban Zimbabweans was pooled together with the data of African origin participants from the UK. This population is different in terms of socio-economic status, education, diet and lifestyle compared with the Zimbabwean group. However, it was necessary to pool these data to allow for comparison across the three levels of BMI. In addition, the UK group provides an additional level of urbanisation and population heterogeneity. 


12.3 Recommendations for future work

The recommendations for future work have been included in each chapter. These include
Matching for comparability in studies that are comparing different ethnic groups.
The use of universal cut-off points for overweight and obesity are not appropriate in all population groups and not recommended.
It is not recommended that TBW estimates from BIA are incorporated into multi-compartment models to be used as reference measures.
It is recommended that different clusters risk measures are developed for non-Caucasian groups.
In epidemiological studies, such as the one carried out in Zimbabwe, larger samples of both rural and urban populations are needed.
It is recommended that further studies are carried out to assess the effect of overweight in rural men and women
When assessing dietary intake, one 24 hour recall may not sufficiently capture energy intake.
It is recommended that more studies are conducted in large populations living in urban areas to ascertain the change in triglycerides noted in this study.
It is recommended that the severity of risk factors associated with increased BMI in African populations are explored in larger samples.


12.4 Conclusion

This thesis explored the relationship between adiposity and cardiovascular risk in different ethnic groups. The initial laboratory studies compared field measures of adiposity and cardiovascular risk in three ethnic groups, Afro-Caribbean, South Asian and Caucasian. These laboratory studies were followed by a population study was undertaken in Zimbabwe. 
The first laboratory study demonstrated that the relationship between body mass index (BMI) and percentage body fat (%BF) differed between Afro-Caribbeans, Asians and Caucasians.  In addition it demonstrated the importance of matching for age, gender and BMI; this enabled a better comparison between the groups than was possible with heterogeneous groups. More importantly it was shown that ethnicity was a significant predictor of %BF.  This study questioned the use of universal BMI cut-off points to define overweight and obesity as they may not represent the same levels of %BF in all ethnic groups.
The equations used in the estimation of body composition were compared between Afro-Caribbean and Caucasian men and women. It was shown that, %BF estimated by the two-compartmental model of Siri (1961) and the ethnic specific two-compartmental equations of Schutte et al., (1984) and Wagner and Heyward (2001) in Afro-Caribbean men gave significantly different results. In addition, it was shown that percentage body fat estimated by Siri (1961) and the ethnic specific equation of Ortiz et al., (1992) in Afro-Caribbean women were not significantly different. Moreover when compared with a three-compartmental model equation (Siri 1956), there was poorer agreement in %BF for all two-compartment equations in men than in women. This showed that equations used to predict %BF may be biased when used in different ethnic groups and by gender. 

The basis of the ethnic specific equations is that density of fat free mass differs between ethnic groups. Using a modified three-compartment equation, which incorporated TBW estimated by BIA, it was demonstrated that density of fat free mass was significantly lower in Caucasian than Afro-Caribbean men and women. However, when the TBW estimates from BIA were compared with an established range for TBW the estimates were outside of the established range and significantly lower, Therefore, it is not recommended that TBW estimates from BIA be incorporated into multi-compartment models to be used as reference measures in field studies.
The second laboratory study aimed to explore the differences in CVD risk in Afro-Caribbean and Caucasian men and women matched for age, gender and BMI. It was demonstrated that total cholesterol, HDL-C and systolic blood pressure significantly predicted the Framingham risk score in Afro-Caribbeans, whereas only HDL-C significantly predicted the Framingham score in the Caucasian group. This study reinforced the fact that risk factors for CVD contributed differently to CVD risk between ethnic groups.
The laboratory studies enabled the critique of the methods, testing and preparation for the population study. The population showed that there were differences in overweight, obesity and risk of CVD in three population groups resident in Zimbabwe. A pattern emerged, amongst women; all variables showed the urban group at greatest risk, followed by the rural women and then university women. Amongst men, the urban group was at greatest risk followed, however, interestingly there were small difference between rural and university men. However, rural men had a significantly larger waist circumference than university men.  This population study provided an evidence base for future studies in Zimbabwe that aim to explore differences in overweight, obesity and risk of CVD.

12.4.1 Public health implications

The main contribution of these studies is that they were able to demonstrate differences in body composition, mainly adiposity, and risk of related disease between different ethnic groups in the UK and between groups within the same population in Zimbabwe. These studies reinforced the need for cautious use of measurement of adiposity, in particular the use of universal measures and cut-off points for the identification of disease risk in all population groups.
In Zimbabwe these data, although limited, show evidence of differences in overweight and obesity prevalence and risk of associated disease between population groups. Moreover, women in both urban and rural areas are at greater risk than men. This follows a trend that has been reported in other countries in Sub-Saharan Africa and gives evidence of a population undergoing the nutrition transition. It is likely that these trends towards overweight and obesity will continue in Zimbabwe as larger numbers of the population become urbanised. Urbanisation results in changes in diet and lifestyle that promote overweight and obesity. Consequently, this country is likely to experience an increase in the prevalence of obesity-related disease, such as type 2 diabetes mellitus and cardiovascular disease. For a country that is already experiencing a high burden of infectious disease (mainly HIV/AIDS), this will result in a double burden of disease, which will be a burden on government spending in healthcare.
Investment in preventative measures such as campaigns to promote healthy eating and physical activity, particularly amongst urban populations and women will help in reducing the burden on overweight and obesity. Moreover, a political commitment to build non-obesegenic environments, will ensure the sustainability of efforts to reduce overweight and obesity.


APPENDICES
APPENDIX 2.1 :Ethnicity

Defining ethnicity

The terms race and ethnicity are commonly misused within the scientific literature including epidemiological studies (Bhopal 2009). Although related, the concepts of race and ethnicity are different. Race refers to a group?s biological homogeneity as defined by a few phenotypical features such as skin colour, hair form and facial features. While ethnicity refers to its shared characteristics, including geographical and ancestral origins, but particularly cultural traditions and languages (Bhopal 2004). The characteristics of ethnicity, unlike race, are not fixed and change over time and with context (Bhopal 2004).
Human beings were grouped into continental races i.e. Africans, Caucasians (Europe, North America and Australia), Asians, Pacific Islanders and Native Americans (Cooper and David 1981). However, it was noted that majority of population differences (90-95%) occurred within populations and not between continental/racial groupings. Moreover, the genes responsible for different physical characteristics, such as skin colour and facial features, that underpin race, were few and rarely related to behaviour or disease (Ashari et al., 2002, Pearce et al., 2004).
The term race has therefore been slowly rejected in favour of ethnicity within the scientific literature particularly in Europe  (Billinger 2007). However, it has been argued that race remains a relevant concept and the term should not be rejected (Oppenheimer 2001). Consequently, a large vocabulary has emerged for the definition of ethnicity and race. For example, ethnic group, ethnic origin, socio-cultural group, race, racial group, ancestry, family origin and national origin were all used to define populations in this respect (Aspinall 2007). These terms are unclear and may be used inappropriately. Current debate calls for the standardisation of terms relating to ethnicity and race (Bhopal 2004, Aspinall 2007, Oppenheimer 2001).

Ethnicity as an epidemiological variable

Senior and Bhopal (1994) identified the following problems with the use of ethnicity as an epidemiological variable
Firstly, ethnicity is a difficult variable to measure. It is not clearly defined changes with time and context. Health researchers tend to adopt the official categorisations such as those used in the census (Smart et al., 2008). This practice is not recommended as it aligns scientific research to state bureaucracy and which may be influenced by social, migration and discriminatory factors (Smart et al., 2008). Most importantly, official categorisation may result in the social influencing the biological, with differences in disease prevalence and risk identified within the official boundaries as opposed to there being genuine differences (Kahn 2006).
In the UK 2001 census the categories were developed following studies in which populations self-defined their ethnicity. Although self-assessment is the favoured form of categorisation (Risch 2002), it changes over short periods of time and is not subject to the control of an investigator. These characteristics make it a complex variable to use in studies of health and disease (Senior and Bhopal 1994).
Secondly, the use of broad terms such as Asian, Afro-Caribbean and Caucasian to describe populations is misleading as it assumes heterogeneity within these groups (Bhopal and Donaldson 1998, Agyemang et al., 2005). The terms are too broad and mask the differences within the population groups that may be as a result of socio-economics and lifestyle differences (Aspinall 2003, 2004).
Thirdly, the problem of ethnocentricity, where one?s own culture is viewed as the standard against which other populations are compared for example where the minority groups are compared with the majority population. However, there is an underlying assumption that the majority population is heterogeneous.
And finally unclear or unproven hypothesis using preconceived assumptions of ethnic differences may affect true estimates of prevalence. In some cases differences may be a consequence of another variable, but attributed to ethnicity. For example, differences may be a consequence of socio-economic variables rather than ethnic difference. There are numerous published reports of ethnic differences in the pattern of diseases that have been little studied beyond the initial observation of difference (Senior and Bhopal 1994, Fausto-Sterling 2008). 

References

Afshari, R., Bhopal, R.S., (2002). Changing pattern of use of ethnicity and race in scientific literature. International Journal of Epidemiology, 31, pp.1074-1076.
Agyemang, C., Bhopal, R., Bruijnzeels, M., (2005) Negro, Black, Black African, African Caribbean, African American or what? Labelling African origin populations in the health arena in the 21st century. Journal of Epidemiology and Community Health 59, pp.1014-1018.
Aspinall, P.J., (2007). Approaches to developing an improved cross-national understanding of the concepts and terms relating to ethnicity and race. International Sociology, 22, pp.41-70.
Aspinall, P.J., (2004). Collective terminology to describe the minority ethnic population: the persistence of confusion and ambiguity in usage. Sociology, 36, pp. 803-816.
Aspinall, P.J., (2003). Who is Asian? A category that remains contested in population and health research. Journal of Public Health Medicine, 25, pp. 91-97.
Bhopal, R., (2009). Medicine and public health in a multi-ethnic world Journal of Public Health, 31, pp.315-321.
Bhopal, R., (2004). Glossary of terms relating to ethnicity and race: for reflection and debate. Journal of Epidemiology and Community Health, 58, pp.441-445.
Bhopal, R., Donaldson, L., (1998). White, European, Western, Caucasian, or what? Inappropriate labelling in research on race, ethnicity and health. American Journal of Public Health, 88, pp.1303-1307.
Chaturvedi, N., (2001). Ethnicity as an epidemiological determinant- crudely racist or crucially important. International Journal of Epidemiology. 30, pp.925-927. 
Fausto-Sterling, A., (2008). The bare bones of Race. Social Studies of Science, 38, pp.657-694.
Kahn, J., (2006). Genes, race and population: avoiding a collision of categories. American Journal of Public Health, 96, pp.1965-1970.
Oppenheimer, G.M., (2001). Paradigm lost: race, ethnicity and the search for anew population taxonomy. American Journal of Public Health, 91, pp.1049-1055.
Pearce, N., Foliaki, S., Sporle, A., Cunningham, C., (2004). Genetics, race and health. British Medical Journal, 328, pp.1070-1072.
Risch, N., Burchard, E., Ziv, E., Tang, E., (2002). Genome Biology, 3. Pp.1-12.
Smart, A., Tutton, R., Martin, P., Ellison, G.T.H., Ashcrof, R., (2008). Social Studies of Science, 38, pp.407-422,
Senior, P.A., Bhopal, R., (1994). Ethnicity as a variable in epidemiological research. British Medical Journal, 309, pp.327-330.


APPENDIX 3.1: THE BRUSSELS CADAVER ANALYSIS SERIES

None of the classic cadavers described in chapter three included any skinfold thickness measurements or extensive anthropometry (Clarys et al., 1984, Clarys and Marfell-Jones 1986, Martin and Drinkwater 1991). In 1985, Clarys and colleagues published data from the first Brussels cadaver analysis series (BCAS). They conducted an extensive series of measurements including anthropometry, radiography, photogrammetry and densitometry on 25 elderly Belgian male and female cadavers. In addition they attempted to provide normative data on the weights and densities of skin, adipose tissue, muscle, bone and organs. Two further studies followed in the Brussels cadaver analysis series; Clarys et al., (1986) measured body composition of the limb segments and derived prediction equations for segment weights of skin, adipose tissue, muscle and bone. The third study investigated the relationship between body composition estimated by computed tomography and values obtained by dissection and weighing tissues, these were incomplete dissections except for three whole body analysis (Janssens et al., 1994). In total 25 cadavers, 13 male and 12 female were analysed in the BCAS.
It was noted by the BCAS investigators that, the methods used to preserve and analyse the cadavers differed significantly between their studies. For example some of the cadavers were preserved with embalming fluid while others were not. If the embalming fluid were distributed homogenously throughout all the tissues it would have had no effect on the tissue masses expressed as %ages of body weight. However, they noted that adipose tissue, muscle and undifferentiated tissues retained more embalming fluid than skin or bone. The retention of embalming fluid would have tended to inflate the relative weights of adipose, muscle and undifferentiated tissues as compared with bone and skin. Therefore, the findings of these analysis need to be interpreted with caution, given that the relative weights of these tissues in embalmed cadavers is uncertain. (Clarys et al., 1985, 1999, 2005).
Moreover, the results of the BCAS are questionable when applied to in vivo. Clarys et al., (2005) measured 18 elderly male and 22 elderly female Belgians aged 55-92years.  They found the overall morphology of the living to be similar to the dead and concluded that their results were applicable in vivo. However, this finding is expected, as the living and the cadavers were from the same population group i.e. elderly Belgians. Moreover, this finding does not support the application of the BCAS knowledge to younger, non-Caucasian ethnic groups. The BCAS proposed the adipose tissue free weight (ATFW) concept as a normalising agent that would facilitate the comparison of body composition between groups. It compares the weight of the whole body less the weight of all the adipose tissue. When the normalising agent was applied to their sample it eliminated the gender differences. 
Differences in non-adipose tissue weight have been reported between ethnic groups. Asians have been reported to have less ATFW when compared with their Caucasian counterparts, resulting in their smaller frame size (Deurenberg et al., 2000). In addition African-Americans have been reported to have greater skeletal weight and consequently greater skeletal muscle than Caucasians (Wagner and Heyward 2000). Therefore, a normalising agent may not reduce the differences between ethnic groups. 
The BCAS added substantially to the overall number of whole body cadavers dissected. Cadaver analysis has added substantially to the understanding of human body composition.  It is a complicated and difficult undertaking. In addition to securing the cadavers, specialist equipment and a dedicated laboratory are required. The process is time consuming, labour intensive and can be distasteful. In more recent times ethical considerations have limited this type of study (Clarys et al., 2005). However, the nature of their sample skewed the data available on cadaver analysis considerably. The sample is overwhelmingly elderly and Caucasian. 


References


Clarys, J.P., Martin, A.D., Drinkwater, D.T., (1985). Gross tissue weights in the human body by cadaver dissection. Human Biology, 56, pp.459-473.
Clarys, J.P., Martin, A.D., Marfell-Jones, M.J., Janssens, V., Caboor, D., Drinkwater, D.T., (1999). Human body composition: A review of adult dissection data. American Journal of Human Biology, 11, pp.167-174.
Clarys, J.P., Marfell-Jones, M.J., (1986). Anthropometric prediction of component tissue masses in the minor limb segments of the human body. Human Biology, 58, pp.761-796.
Clarys, J.P., Provyn, S., Marfell-Jones, M.J., (2005). Cadaver studies and their impact on the understanding of human adiposity. Ergonomics, 48, pp.1445-1461.
Janssens, V., Thys, P., Clarys, J.P., Kvis, H., Chowdhury, B., Zinzen, E., Cabri, J., (1994). Post-mortem limitations of body composition analysis by computed tomography. Ergonomics, 37, pp.207-216.


APPENDIX 4.1: UK ETHNIC CATEGORIES, CENSUS (2001)

WhiteWhite BritishWhite IrishWhite otherMixed Asian or Asian BritishIndianPakistaniBangladeshiOther AsianBlack or Black BritishBlack CaribbeanBlack AfricanBlack otherChineseOther


APPENDIX 8.1 FIELDWORK BOOK


ADIPOSITY AND CVD RISK FACTORS: A COMPARISON BETWEEN AN URBAN AND RURAL POPULATION IN ZIMBABWE


Professor Lucie Malaba (University of Zimbabwe)
Nonsikelelo Mathe (Buckinghamshire New University)


SUBJECT INFORMED CONSENT
PROTOCOL TITLE: Adiposity and CVD Risk Factors: A Comparison between an Urban and Rural population in Zimbabwe
NAME OF RESEARCHER: Professor Lucie Malaba and Nonsikelelo Mathe
PHONE: 04 308772
PROJECT DESCRIPTION:
Overweight and obesity are related to heart disease (including hypertension and stroke), diabetes and some types of cancer. Although more people are obese and overweight in western countries, more people in Zimbabwe are becoming overweight and obese. Not a lot is known about the relationship between body fat and the development of heart disease in people of African origin. This study will look at this relationship.  

PURPOSE OF RESEARCH STUDY: To understand the relationship between body fatness and risk of heart diseases in rural and urban populations in Zimbabwe.
PROCEDURES INVOLVED IN THE STUDY:

Cholesterol and blood sugar levels: One of your fingertips will be pricked and a drop of blood will be used to determine cholesterol and blood sugar levels.  This is another way of estimating your risk of heart disease. 

Food frequency, physical activity and demographic questionnaires- These questionnaires will be used to collected data on your food consumption, physical activity patterns and other lifestyle factors.

Blood pressure: A blood pressure monitor will be used to measure blood pressure while you?re sitting comfortably.  Blood pressure is used to estimate the long-term risk of heart disease.

Height, weight, waist and hip circumferences: Height will be measured using a wall-mounted ruler (or stadiometer) and you will be weighed on electronic scales. A tape measure will be used to measure your waist and hip circumferences.  

Skinfold thickness: Some body fat is found just beneath the skin and by measuring this fat we can estimate how much fat there is in all the body.  A small fold of skin is gently pinched together and measured using callipers. This will be done on seven places on your body. You will need to loosen your clothing to have these measurements taken. 

RISKS: Participating in this study possess very absolute minimum level of risk of harm as none of the measurements taken expose you to harmful substances. In addition sterile disposable needles will be used for the fingerprick procedure. Trained enumerators 
DISCOMFORT: The fingerprick procedure may cause discomfort to some participants, however, this is limited as the prick is shallow and the does not cause a laceration in the skin. 
Participants may also feel discomfort during the skinfold calliper measurements. Investigators have been trained to pinch gently and to take readings within two seconds.
PERSONAL BENEFITS: You will receive feedback on the levels of blood cholesterol, blood sugar, blood pressure and estimated percentagebody fat.
STUDY WITHDRAWAL:  You may choose not to enter the study or withdraw from the study at any time without loss of benefits entitled to you.
CONFIDENTIALITY OF RECORDS: Your involvement in and your particular data from this trial will remain strictly confidential. Only researchers involved in the investigation will have access.
PROBLEMS/ OUESTIONS: If you are at all concerned about this please contact Professor Lucie Malaba or Nonsikelelo Mathe 04308772
AUTHORIZATION:
I have read the information on the research in which I have been asked to participate and have been given a copy to keep. I have had the opportunity to discuss the details and ask questions about this information.
PARTICIPANT?S SIGNTURE???????.DATE???????????
PARTICIPANT?S NAME??????????????????????..

INTERVIEWER?S NAME??????????????????????.

INTERVIEWER?S SIGNATURE????????????????????


DEMOGRAPHICS AND PARTICIPANT INFORMATION
Name:Subject ID:ResidenceUrbanPeri-UrbanRuralDOB:17-2425-3435-4445-5455-6465+GenderMaleFemaleMarital StatusSingleMarriedWidowedDivorced/separatedOtherEducational statusNo formal educationPrimary educationSecondary educationUniversity educationOtherCurrent EmployedEmployment statusUnemployedJob descriptionStudentSelf-employedOtherCurrent approxmonthlyhouseholdincomeOtherNo. of dependentsChildrenAdultsOther


HEALTH ASSESSMENT 
Have you ever been told by your doctor, nurse or other health professional that you have any of the following conditions:

Yes
No

Asthma


Bone disease (specify)


Bowel disease (ulcers)


Cancer (specify)


High blood cholesterol


Diabetes


Heart disease


High blood pressure


Obesity


Stroke


Others (specify)

SPECIFY???????????...
???????????????...
???????????????...

Are you currently on treatment for any of the following conditions:

    
Yes
No

Asthma


Bone disease (specify)


Bowel disease (ulcers)


Cancer (specify)


High blood cholesterol


Diabetes


Heart disease


High blood pressure


Obesity


Stroke


Others (specify)

SPECIFY???????????.
???????????????.
???????????????.

1
2

Yes
No       

a) Have you ever had your blood pressure checked? 
 

If your answer to this question is yes, please answer part b) of the question.     b) When was the last time you had your blood pressure checked?
1
within the past 1 month


2
within the past 3 months


3
within the past 6 months


4
within the past 1 year


5
Other (specify) ???????
????????????.?.

a) Cholesterol is a fatty substance found in blood. Have you ever had your blood cholesterol checked? 
1
2

Yes
No       


If your answer to this question is yes, please answer part b) of the question.     b) When was the last time you had your blood cholesterol checked?
1
within the past  6 months


2
within the past 12 months


3
within the past 2 years


4
within the past 5 years


Other (specify)????????? 
????????????...??.

a) Have you ever been told after blood sugar tests that you have diabetes?
1
2

Yes
No       


If your answer to this question is yes, please answer part b) of the question.     1
Less than 10 years old


2
between 10 and 20 years old


3
between 20 and 30 years old


4
between 30 and 40 years old


5
more than 40 years old


6


Specify age at diagnosis
??????????

b) What age (how old) were you when you were told you had diabetes?a) Are you currently on treatment for diabetes?
1
2

Yes
No       

If your answer to this question is yes, please answer part b) of the question.     b) Does your diabetes treatment include
1
insulin injection


2
Tablets only


3
Tablets and special diet


a) Are you currently on treatment for heart disease?
1
2

Yes
No       


If your answer to this question is yes, please answer part b) of the question.     b) Does your treatment include any of these?
1
BP tablets


2
Other heart disease tablets?


3
Special diet (specify) ???????????????...

4
Physical