Obesity in adult populations is recognised as a growing epidemic worldwide1 and appears to be increasing more rapidly in developing countriesReference Popkin and Gordon-Larsen2. Importantly, obesity is associated with non-communicable diseases, including non-insulin-dependent diabetes, hypertension, dyslipidaemia, cardiovascular diseases, certain types of cancers and psychological problems1. The disproportionate increase in the developing world has been attributed to the ‘nutrition transition’, a shift from traditional diets and lifestyle to the pattern of Western societiesReference Popkin and Gordon-Larsen2. Its serious implications for the health of these populations are clear through the link between overweight and chronic non-communicable diseases.
Little is known about diet and its contribution to adiposity in the African Diaspora. We have previously reported on the variation in diet across three populations that share a common genetic ancestry but different environments – rural and urban Cameroon, Jamaica and Afro-Caribbeans the UKReference Mennen, Jackson, Sharma, Mbanya, Cade and Walker3. Our investigation of adult Jamaicans of African descent regarding the relationships between body mass index (BMI) and diet showed that BMI was not explained by energy and that higher protein intake was associated with increased BMI in females onlyReference Jackson, Walker, Forrester, Cruickshank and Wilks4. In the present study we investigated the relationship of diet to overweight and obesity across these different populations.
It was hypothesised that, relative to the degree of Westernisation, a gradient in prevalence of excess weight would exist: rural Cameroonians would exhibit the lowest prevalence, followed by urban Cameroonians, Jamaicans and the UK sample. We also hypothesised that the macronutrient content of the diet, in particular dietary fat, would be independently related to weight status.
Materials and methods
Sample and study design
The investigation was a multi-centre collaborative study of hypertension and diabetes in four populations of African origin. The sampling scheme has been described elsewhereReference Cooper, Rotimi, Ataman, McGee, Osotimehin and Kadiri5. In summary, in urban Cameroon, rural Cameroon and Jamaica (Spanish Town) communities were sampled according to the ‘probability proportionate to size’ methodReference Lemeshow and Robinson6. Equal proportions of men and women in four age categories (25–34, 35–44, 45–54 and 55–74 years), determined to be eligible (see eligibility criteria below), were approached to be included in the survey. Recruitment was done house to house. In the UK (Manchester) random samples of Afro-Caribbeans were recruited from population registers.
Subjects were recruited to the study if they were 25–74 years old, not pregnant and of African origin according to at least two of the following criteria: (1) ancestry – at least three grandparents should be black; (2) observed race – a subjective assessment by the interviewer; and (3) self-assignment.
The response rate in urban and rural Cameroon was 95% and 98%, respectively, 62% in Jamaica and 80% in the UK.
Measurements
Dietary assessment
Dietary intakes were assessed using previously validated food-frequency questionnaires (FFQs) developed to assess diets of adults in all sitesReference Jackson, Walker, Cade, Forrester, Cruickshank and Wilks7, Reference Sharma, Cade, Jackson, Mbanya, Chungong and Forrester8. Each site developed an FFQ to assess habitual nutrient intakes. All FFQs assessed diet over the previous year and were interviewer-administered. In Jamaica, the FFQ was validated against twelve 24-hour recalls over one yearReference Jackson, Walker, Cade, Forrester, Cruickshank and Wilks7. The UK questionnaire was calibrated by comparison with a 24-hour recall and 4-day weighed intakeReference Sharma, Cade, Jackson, Mbanya, Chungong and Forrester8. The validity of the questionnaires was comparable to those used in other populationsReference Rimm, Giovannucci, Stampfer, Colditz, Litin and Willett9, Reference Martin-Moreno, Boyle, Gorgojo, Maisonneuve, Fernandez-Rodriquez and Salvini10.
Anthropometry
Body weight (without shoes) was measured in clothing to the nearest 0.5 lb on a bathroom type digital scale. Height was measured without shoes on a floor-standing stadiometer fitted with spirit level, to the nearest 0.1 cm. BMI was calculated and the World Health Organization's classification was used to determine overweight (BMI ≥ 25.00 kg m− 2) and obesity (BMI ≥ 30.00 kg m− 2). Information on age, education level, marital status and employment status was obtained by questionnaire.
Statistical analyses
Sociodemographic characteristics were compared according to weight status using chi-square statistics. Analysis of variance and non-parametric statistics (Mann–Whitney U-test) were used as appropriate.
Dietary intake variables were expressed as grams per day or as a percentage of total energy from fat, carbohydrate and protein. Alcohol was expressed as grams per day, percentage of energy or as a dichotomous variable (yes/no).
Nutrient intakes were adjusted for total energy by computing residuals from regression analyses, with energy intake as the independent variable and nutrient intake as the dependent variableReference Willett11. Residuals were added to the expected nutrient value for the mean energy intake of the sample to obtain a score adjusted to the average energy intake. Energy intakes and adjusted mean intakes were grouped in tertiles.
Multiple logistic regression analyses were used to examine the effect of diet composition (tertiles of adjusted macronutrient intake and fibre as grams per day) and sociodemographic variables on the risk of overweight (BMI ≥ 25.00 kg m− 2) and obesity (BMI ≥ 30.00 kg m− 2) at each of the four sites. All analyses were conducted separately for men and women. Age and energy intakes were entered in all models; macronutrients, social and demographic variables (marital status, education and employment) offered stepwise. Final models were calculated with the combined sites, with site entered in the model. In the final model where diet and site were significant, we examined the effect of interaction between both on the odds of overweight and obesity. Statistical analyses were performed using the Statistical Package for the Social Sciences, version 12 (SPSS Inc., 2004). Statistical significance was achieved when P < 0.05.
Results
Exclusions
Using a priori exclusion criteria of energy intakes < 800 or >5500 kcal, 155 subjects were excluded (rural Cameroon, n = 59; urban Cameroon, n = 53; Jamaica, n = 39; UK, n = 4). The final sample comprised: n = 686 subjects from rural Cameroon, n = 975 from urban Cameroon, n = 937 form Jamaica and n = 257 from the UK. The data were collected between 1992 and 1995.
Characteristics of the sample
Demographic and anthropometric characteristics of the study participants by gender are shown in Table 1. The mean age of men and women studied was 46 years, but was somewhat lower in urban Cameroonians (38 years). Fewer persons from rural Cameroon and Jamaica compared with those from the UK and urban Cameroon had attained a secondary education or higher. Jamaicans were less likely to be married or in a union. Among females, the prevalence of overweight and obesity was lowest in rural Cameroon and increased across the sites; however, among males the second lowest prevalence of overweight was found in Jamaica.
Sociodemographic and anthropometric characteristics of four African populations
SD – standard deviation; BMI – body mass index.
* Significantly different across sites (analysis of variance or chi-square test): P < 0.001.
† Significantly different between genders (analysis of variance or chi-square test): P < 0.001.
Energy, nutrients and BMI
Energy and macronutrient intakes were compared among BMI categories. Among males and females with BMI < 25 kg m− 2, total energy intake was highest among rural and urban Cameroonians and lowest in men from the UK and women in Jamaica (Table 2). With the exception of rural Cameroonian males with BMI of 25–29 kg m− 2 who reported slightly lower intakes of energy from fat and higher intakes of carbohydrate, males consumed similar proportions of macronutrients among the BMI categories. Macronutrient intakes were similar across BMI categories among females in Jamaica. Women in the UK reported diets in which energy from fat tended to increase and that from carbohydrate tended to decrease with increasing BMI. Only two women in rural Cameroon were obese (BMI ≥ 30 kg m− 2) and they were not considered in the analyses.
Energy intakes and percentage contribution of macronutrients to energy by categories of BMI and site
BMI – body mass index; %E – percentage of energy.
Data are expressed as mean ± standard deviation.
* Dietary intakes significantly different across sites (analysis of variance): P < 0.001.
† Dietary intakes significantly different among BMI categories (analysis of variance): P < 0.001.
Factors associated with overweight within sites
Among men few dietary variables were associated with overweight. Increased energy intake in urban Cameroonian men increased the risk of being overweight but the relationship was inverse among men from the UK (Table 3). Protein was associated with overweight in UK men only, with the consumption of 80 g or more of protein daily (tertile 3) associated with increased risk of being overweight (odds ratio (OR) = 7.43; 95% confidence interval (CI) 2.23–15.76) compared with tertile 1 (72 g or less of protein daily) (Table 3). Middle-aged men in urban Cameroon and the UK were more likely to be overweight. Jamaican men who had attained tertiary education were 5.77 (95% CI 2.00–13.27) times more likely to be overweight than men with secondary education or less. No variables were related to overweight in rural Cameroonian men.
OR (95% CI) for the likelihood of overweight (BMI=25.00–29.99 kg m−2) by dietary and sociodemographic factors of adults of African populations
OR – odds ratio; CI – confidence interval; BMI – body mass index.
* Age and energy were entered in the model; carbohydrate, protein, fat, fibre, education and marital status were offered.
† Energy-adjusted nutrients as suggested by WillettReference Willett11.
‡ Non-significant variable; not included in the model.
In all sites, women's energy intake was not associated with overweight. Jamaican women who consumed 87 g or more of protein daily (tertile 3) were at increased risk of being overweight (OR = 2.06; 95% CI 1.25–3.38) compared with those in tertile 1 (75 g or less daily), while women in the UK whose daily intake of fat exceeded 46 g (tertile 3) were more likely (OR = 8.06; 95% CI 1.45–29.31) to be overweight than those in tertile 1 (36 g or less). Urban Cameroonian women aged 41–59 years were more likely to be overweight than younger women (Table 3). Tertiary-educated women in rural Cameroon also showed increased risk of being overweight compared with women with secondary or less education.
Factors associated with obesity within sites
Table 4 shows the variables associated with risk of obesity. Dietary variables were associated with obesity only in the UK. Higher protein intakes (tertile 3) were related to increased risks of obesity in UK males, whereas carbohydrate intakes above 283 g (tertiles 2 and 3) lowered the risk of obesity among these men. Middle-aged (41–59 years) urban Cameroonian men were more likely to be obese than younger men. Jamaican men who were single were less likely to be obese.
OR (95% CI) for the likelihood of obesity (BMI≥30.00 kg m−2) by dietary and sociodemographic factors of adults of African populations*
OR – odds ratio; CI – confidence interval; BMI – body mass index.
* Rural Cameroon excluded from analyses.
† Age and energy were entered in the model; carbohydrate, protein, fat, fibre, education and marital status were offered.
‡ Energy-adjusted nutrients as suggested by WillettReference Willett11.
§ Non-significant variable; not included in the model.
Similar to men, risk of obesity increased with age in urban Cameroon women aged 41 years or older. No variables were associated with obesity among Jamaican women. Consumption of more than 36 g of fat per day (tertiles 2 and 3) among women in the UK was associated with increased risk of obesity.
Cross-site comparison
Aggregated analyses of all sites were conducted with a variable included for site. Obesity was rare in rural Cameroon and this site was omitted from analyses of obesity. Rural Cameroon was the reference category for analyses of overweight and urban Cameroon for analyses of obesity.
Among males, no dietary variables were related to overweight and obesity. Risks of both overweight and obesity were increased in the age range 41–59 years compared with younger men, and older men (60–74 years) had increased risk of obesity but not overweight (Table 5). Men in urban Cameroon, Jamaica and the UK were at increased risk of overweight compared with rural Cameroon. Compared with males in urban Cameroon, Jamaican men were at reduced risk of obesity whereas risk of obesity was not different in UK males.
OR (95% CI) for the likelihood of overweight (BMI=25.00–29.99 kg m−2) and obesity (BMI≥30.00 kg m−2) by dietary and sociodemographic factors – combined sites
OR – odds ratio; CI – confidence interval; BMI – body mass index.
* Age and energy were entered in the model; carbohydrate, protein, fat, fibre, education and marital status were offered.
† Energy-adjusted nutrients as suggested by WillettReference Willett11.
‡ Non-significant variable; not included in the model.
Among women, increased risks of overweight were related to being 41–59 years of age, higher protein intakes (89 g or more daily) and having attained a secondary level education. Assessment of interaction terms revealed that the effect of protein intake on the odds of being overweight did not differ with site. Women in the second tertile of energy consumption were at lower risk of overweight. Women in urban Cameroon were seven times more likely, and Jamaica and the UK 10 and 17 times more likely, respectively, to be overweight than women in rural Cameroon. Females aged 41–59 years were at greater risk of obesity than younger women and when compared with urban Cameroonian women, women from Jamaica and the UK were at increased risk of being obese.
Discussion
The present study is the first of which we are aware to examine the influence of diet and sociodemographic factors on risks of overweight and obesity in adults of African origin in environments ranging from rural Africa to Western society. Dietary intakes were generally not related to overweight and obesity in developing countries but appeared to influence the risks of excess adiposity in the UK. Associations of macronutrients with risks of overweight and obesity differed between genders and across countries.
Diet and excess weight
Sustained imbalance between energy intake and energy expenditure for extended periods contributes to the onset of weight gainReference Stunkard, Chadwick and Cardew12, Reference Aronne13. Energy intakes differed among the sites but were associated with overweight among men only. Energy intake was positively associated with overweight in urban Cameroon but negatively associated in the UK and sites aggregated. The inverse relationship of energy intake and overweight observed in the UK may be due to the underestimation of energy intakeReference Kromhout14 or dietingReference Parker, Gonzalez and Derby15, Reference Hu, Pekkarinen, Hanninen, Tian and Jin16 by overweight subjects. Of note, in an earlier report we showed that underreporting, as determined by varying levels of the ratio of energy intake to estimated basal metabolic rate (EI/BMRest)Reference Schofield, Schofield and James17, ranged from 4 to 13% in urban Cameroon but showed high levels in the UK (28–66%)Reference Mennen, Jackson, Cade, Mbanya, Lafay and Sharma18. Underreporting may therefore have contributed to the negative association of energy intake and overweight found in the UK and sites combined. The absence of associations between energy intakes and overweight and obesity in rural Cameroon and Jamaica is in agreement with other cross-sectional investigationsReference Myers, Klesges, Eck, Hanson and Klem19, Reference Dreon, Frey-Hewitt, Ellsworth, Williams, Terry and Wood20 but in contrast with experimental diet studies. In cross-sectional studies weight status may influence intakes and obscure any relationship of dietary contribution to weight outcomes. In spite of the care taken to eliminate under- and over-reporters, our data may also be subject to methodological bias, such as the ability to estimate dietReference Willett11, as previously discussedReference Mennen, Jackson, Cade, Mbanya, Lafay and Sharma18.
An unexpected finding was the association of higher protein intake with overweight. Greater protein intake was associated with overweight among men in the UK and Jamaican women and with obesity in men from the UK. This is inconsistent with other studies that associate healthy body weights with higher protein intakesReference Paeratakul, Popkin, Keyou, Adair and Stevens21 through suggested mechanisms of increasing satiety and facilitating weight lossReference Anderson and Moore22. It may be that overweight subjects in our study, in an attempt to lose adiposity, increased protein intakes, a regime widely promoted for weight lossReference Parks23, Reference Freedman, King and Kennedy24. The higher protein intakes associated with overweight/obesity may be a marker for attempted weight loss. Associations with protein intakes may also be confounded by social factors. Although we adjusted for education as a proxy for social class, confounding by other aspects of social class may remain.
The results for carbohydrate are similar to other cross-sectional studies. Greater carbohydrate intake predicted decreased likelihood of obesity among men in the UK, similar to negative associations of body weight and carbohydrate intakes in populations of industrialised countriesReference Astrup, Meinert Larsen and Harper25, Reference Slattery, McDonald, Bild, Caan, Hilner and Jacobs26, although an association of carbohydrate-rich diets with high glycaemic index and weight gain has also been suggestedReference Scali, Siari, Grosclaude and Gerber27, Reference Minehira, Bettschart, Vidal, Vega, Di Vetta and Rey28.
The importance of dietary fat in the regulation of body weight has received much attention but remains controversialReference Willett29, Reference Bray and Popkin30. Epidemiological studies in countries undergoing the nutrition transition have shown positive relationships between obesity and dietary fat intake but inverse associations with body weight in many Westernised populationsReference Astrup, Ryan, Grunwald, Storgaard, Saris, Melanson and Hill31, Reference Lissner and Heitmann32. The energy density of fat is higher than that of carbohydrate and satiety signals arising from its ingestion are weaker, thereby allowing passive over-consumption of fat-rich energy-dense dietsReference Blundell, Cotton, Delargy, Green, Greenough and King33–Reference Hill and Prentice35. In our study greater dietary fat intakes increased the risk of overweight and obesity in women in the UK only. Dietary fat was not associated with overweight in Cameroonians who consumed approximately 43–49% of energy as fat, and their low rate of overweight, despite the highest energy intakes and high percentage of energy from fat, is almost certainly explained by high physical activity. Energy intakes and percentage energy from fat were slightly lower in urban than rural Cameroon but overweight and obesity substantially higher, suggesting that lower physical activities due to more sedentary lifestyles may be a more critical influence on weight than intakes in this setting.
Our results differed by site. Overall mean intakes of nutrients differed appreciably within countries and increased the possibility that site-specific variability in intakes of macronutrients were large enough to detect significant associations. There were few associations between diet (energy and macronutrients) and weight in Cameroon and Jamaica but in the UK several associations were observed which were consistent with previous studies except for protein. We also analysed all sites together, potentially creating a more heterogeneous sample to provide insights into the factors affecting weight. The only association was between protein intake and overweight in females. The results suggest that energy expenditure may be a more important determinant of weight status than diet in developing counties and that diet becomes more important where the majority of the population has a sedentary lifestyle.
Non-dietary factors
Several studies provide evidence of the importance of socio-economic factors to weight gain in developing and developed countries and our results were consistent with previous associations. A positive association with age and overweight and/or obesity was noted among developing (urban Cameroon – males and females) and developed (UK – males) countries. Overweight in Jamaica and Cameroon was positively related with higher education, a proxy for social status, as has been reported in other transitional countries1, Reference Benejellounn36, Reference Kain, Vio and Albala37. However, we did not find the inverse association with education in the UK. It is possible that this association may not be consistent across all ethnic groups. We are unaware of previous studies exploring this association among Afro-Caribbeans in the UK. The inverse relationship of not being married or in a stable union (Jamaican men) to weight gain was similar to our findingsReference Rissanen, Heliovaara, Knekt, Reunanen and Aromaa38.
Body image has been identified as important in the aetiology of obesity and social norms are likely to influence reporting of habitual intakesReference Worsley, Baghurst and Leitch39. Our earlier reportReference Mennen, Jackson, Cade, Mbanya, Lafay and Sharma18 showed that the prevalence of underreporting was lowest in rural Cameroon, somewhat higher in urban Cameroon and Jamaica and highest in the UK, suggesting trends in cultural influences on obesity.
There are limitations to this study. Estimates of nutrient intakes from FFQs are not preciseReference Rimm, Giovannucci, Stampfer, Colditz, Litin and Willett9, Reference Cade, Thompson, Burley and Warm40 and have potential for measurement error; however, the FFQs were calibrated for the nutrients examined in this reportReference Jackson, Walker, Cade, Forrester, Cruickshank and Wilks7, Reference Sharma, Cade, Jackson, Mbanya, Chungong and Forrester8. The cross-sectional nature of our study prohibits defining a temporal sequence between diet and weight gain. For example, the possibility exists that real changes to diet occur in response to weight gain, and hence the reported intakes could lead to misclassification of dietary intake and unexpected associations with weight status.
Our results show that associations between diet and overweight/obesity are not generalisable among populations. Given the high rates of overweight in populations of African origin in industrialised countries, longitudinal studies are needed to examine the relative influence of diet and lifestyle factors, including physical activity, on weight gain.
Acknowledgements
The authors gratefully acknowledge financial support from the European Union, contract number TS3*CT92-0142.
