In this cross-sectional study, we assessed the prevalence of obesity and cardio-metabolic risk factors in apparently healthy urban adults in Benin and explored whether birthplace, length of urban residence and SES were associated with these factors, after controlling for age and sex. We also assessed the association of several modifiable lifestyle factors, taken individually or jointly, with these risk factors. We considered the cardio-metabolic risk factors individually rather that their clustering as metabolic syndrome because different definitions of the syndrome are used [32
], and rates may vary considerably depending on the sets of criteria. Furthermore, the value of the metabolic syndrome as predictor of CVD beyond its component abnormalities remains controversial [33
The observed high rate of obesity, particularly among women, is consistent with previous studies in urban Africa [34
]. The influence of environmental, behavioural, psychosocial, and genetic factors on obesity is well recognised [36
]. Sedentary lifestyles are common in urban women compared with men, as we verified using a novel approach consisting of 24-hour recalls of activities. Most women in our study were only involved in activities that are not physically demanding. The average score for physical activity was significantly lower in women compared to men and this may explain the difference in the prevalence of obesity between the two groups. Cultural values and the positive social attitudes towards fatness among women in Africa [37
] are also conducive to feminine obesity as our study confirmed (data not shown).
Our findings show that the risk of obesity increased significantly with rising socioeconomic status. Subjects in the upper SES group have higher access to food and they may maintain a positive energy balance over a prolonged period of time, while periodic food shortage may be common among the poor [39
]. Similar to our results, a positive relationship between SES and obesity has been reported in Cameroon [40
]. Our findings suggest that the study population is still in the early stages of the nutrition transition since excess weight is currently seen primarily among the affluent, before progressively shifting to lower-income groups, as demonstrated in middle-income developing countries at later stages of the nutrition transition [41
]. Based on national data, the burden of obesity appears to shift towards the poorer groups as the country's gross national product reaches the level of upper middle income countries [41
]. However, these are aggregated data and the shift of the obesity burden towards the poor may be expected to take place even in low income countries, at least in large cities.
The fact that SES is more closely associated with overall obesity than abdominal obesity in multivariate analyses is an intriguing observation. The WHO cut-off levels for waist circumference or BMI as used in the present study may not be appropriate for African populations. Indeed, it is recommended to define specific BMI or waist circumference cut-offs for different race-ethnicity groups [42
]. Race-ethnicity-specific WC cut-offs have been proposed in the USA [44
] but these may not necessarily apply to Africans. Lower BMI cut-offs for overweight and obesity have been suggested as alternative public health action points for Asian populations [43
], but relevant data are not available for Africans.
We observed a high prevalence of hypertension, in spite of the fact that previously diagnosed subjects were excluded from the study. A high prevalence of hypertension was also reported in other African countries [45
]. Clearly, hypertension is a major public health problem in sub-Saharan Africa. Although evidence suggests that people of African origin are more susceptible to hypertension [13
], both genetic and environmental factors are intertwined. Unlike for obesity, we did not find that socio-economic status was associated with hypertension. This suggests that hypertension may affect all segments of a population, even in the early stages of the nutrition transition. However, the results of this study indicate that a longer duration of urban residence, independent of age and sex, was associated with a higher risk of hypertension. This was also observed in a black population of the Cape Peninsula, South Africa [10
]. Similarly, a positive rural-urban gradient for the prevalence of hypertension was observed in a population-based survey conducted in Tanzania [47
]. Social deprivation, financial constraints and pressure associated with city living is suspected to add to the risk of hypertension [8
]. Urbanisation also plays a role in the occurrence of hypertension through psycho-social stress as previously found in Tanzania [48
]. However, we did not collect data on stress, which is difficult to measure. The psycho-social determinants of hypertension need further study in urban populations of Africa.
The low prevalence rate of diabetes and hypertriglyceridemia in the present study is noteworthy. A lower propensity to an adverse blood lipid profile in people of African origin was suggested in a previous extensive review on the issue [49
]. Regarding diabetes, the rate was low even if subjects excluded because of previously diagnosed diabetes were taken into account. In fact, the prevalence of diabetes would have reached 2.5% if they had been included in the study (data not shown).
As urbanisation is associated with changes in diet and physical activity [8
], being born in a city could be a potential risk factor for obesity and related metabolic abnormalities. Although we found that birthplace was significantly associated with dietary patterns in a previous study conducted in the same population [20
], it did not show a significant association, however, either with obesity or with other metabolic risk factors in the present study. This is at variance with a study conducted among Mexican adults living in the United States, which showed that Mexican-born men and women had a lower risk of obesity than their U.S-born counterparts [50
]. This suggests that the influence of early exposure to urban life on obesity or related risk factors may vary according to the context. However, as indicated earlier, the population under study is still probably in the early stages of the nutrition transition and it may be possible to find an influence of early life exposure to urban life on CVD risk factors in the future.
Single lifestyle behaviours were correlated with overall lifestyle score, with the exception of smoking. Lifestyle behaviours indeed tend to cluster together, as reported in China and the USA [11
]. For example, men who were physically inactive were also more likely to drink heavily in the present study. However, we did not find any significant association between alcohol consumption and smoking, at variance with several studies reporting such an association [51
]. This may be ascribed to the low prevalence of smoking in the study population (2.5%). Data on smoking status was based on self-reported information, typically collected via questionnaire, which may suffer from reliability problems. Probably one should determine biomarkers of tobacco smoking, i.e serum or plasma level of cotinine, the main metabolite of nicotine, to have more accurate information on smoking status [53
Of the single lifestyle factors examined, physical activity was the most strongly associated with a lower likelihood of overall obesity, abdominal obesity and hypertension. This is in line with previous studies that assessed the association between lifestyle factors and metabolic abnormalities [54
]. Our results showed that young subjects were more active than old ones. This is not surprising as the prevalence of most of the CVD risk factors increased with age (data not shown). We were not able to assess the effect of lifestyle behaviours on diabetes and hypertriglyceridemia because of the small number of cases.
Paradoxically, dietary quality did not show a significant association with obesity and other CVD risk factors, which is in contrast with the nutrition transition theory [3
]. One possible explanation is that the population under study is still in the early stage of the nutrition transition and the diet is still low in fat and sugar, as previously reported [20
]. The lack of association between diet and the risk of CVD could also be ascribed, at least partly, to the diet quality assessment method. The dietary scores that we developed and used in several settings [20
] are the first to integrate the more recent WHO recommendations for the prevention of chronic diseases. Several diet quality indexes have been developed and subsequently modified and adapted [56
]. However, only few showed a significant relationship with health risk [57
]. Diet quality indexes still need to be used and interpreted with care as recently suggested because their development is based on empirical choices [56
]. Location-specific food-based dietary guidelines are urgently required for culturally and economically relevant nutrition communication and as a basis for evaluating the quality of local dietary patterns.
Unlike Kim et al [11
], we gave the same weight to all individual components of the lifestyle score, which assumed that they contributed equally to the score. However, ideally, the weighting of risk factors should be determined based on extensive data on the relative health risk associated with each lifestyle component in longitudinal studies. Unfortunately, such data do not exist for African populations. We found that the likelihood of overall obesity, abdominal obesity, and hypertension decreased significantly as the overall lifestyle score improved. These results suggest that in this population, the majority of cases of obesity and other related metabolic abnormalities could be curbed by the adoption of healthier lifestyles, particularly as regards physical activity.
To our knowledge, there is no universal definition of urbanisation status. Proxy measures such as birthplace and length of urban residence were used to measure urbanisation status in this study. Length of urban residence was recorded by summing the time lived in a city, from birth until the time of data collection. A similar approach, but using a more detailed questionnaire, was used in a previous study conducted in Cameroon [9
]. Birthplace was used because it is closely related to acculturation [58
It has been suggested that multiple indicators of SES should be used when assessing the influence of socioeconomic factors on health [59
]. This issue was addressed in the present study because our SES score was computed based on three different sets of indicators – education, main occupation, and household assets. However, even studies that include multiple standard SES measures may be fall of potentially important socioeconomic influence on health [56
]. It is clear that some unmeasured socioeconomic factors may have affected obesity or other CVD risk factors in our study population. Economic status impinges on health outcomes. In affluent societies, income is the most commonly used measure of economic status, while its measure poses a significant problem in developing countries because household income or expenditure levels are often unavailable or unreliable [60
]. We did not use income as a direct measure of economic status because a large part of the population engages in informal work. As indicated by Houweling et al [30
], expressing income in monetary value in countries where a large part of the population work in self-subsistence agriculture or the informal sector can be extremely time-consuming and suffers important reliability problems. Household assets are a better reflection of economic status than income in developing country settings, and were used as a proxy measure of income. Education was included in the SES score as it is related to health. Education includes several aspects, and years of formal education was used in the present study because of its potential effect on health [61
Our study has several limitations. First, the cross-sectional design does not allow any inference to be drawn with respect to the causal relationship among variables. Second, the study sample is only representative of urban adults of Cotonou, and thus findings may not apply to the whole urban population of Benin. The study may probably lack statistical power due to the modest sample size. The conclusions of the study should therefore be interpreted with caution. These limitations notwithstanding, our study provides useful data on the prevalence of obesity and other CVD risk factors among adults of Cotonou, and their socioeconomic and lifestyle correlates.