Abdominal vs. overall obesity among women in a nutrition transition context: geographic and socio-economic patterns of abdominal-only obesity in Tunisia

Based on a large national random sample of Tunisian women, we found a much higher prevalence of abdominal than overall obesity, similar to results in the few other large-scale studies using national WC data in the MENA region, e.g., Iran [19] or Oman [20] (although not in comparable age classes). Originally, the 88 cm WC “high-risk waist circumference” cut-off value was chosen by the World Health Organization to correspond to a BMI of 30, on the basis of a study in the Netherlands [21]. The large discrepancy in the prevalence of abdominal vs. overall obesity in our study could then result from increases in WC across the whole BMI range over the last decades, as reported in other settings [6]. There could also be ethnicity issues [9], and some authors have proposed a different cut-point of WC ≥ 85 cm to define abdominal obesity among Tunisian women [22]; but, if applied, this would result in an even higher prevalence of abdominal obesity (n = 2964, 68.4% [65.7-71.4]). Other anthropometric indices have been put forward to assess abdominal adiposity, e.g., the waist-to-hip ratio (WHR) ≥0.85, which would result in a similarly higher prevalence of abdominal (n = 2961, 56.2% [53.4-59.0]) vs. overall obesity, or waist-to-height ratio (WHtR) ≥0.6, which was proposed more recently [23], with which the prevalence of abdominal obesity would be lower (n = 2964, 42.6% [39.9-45.4]). Some authors have reported larger seasonal variations in WC vs. BMI, with the difference between the proportion of abdominal vs. overall obesity being somewhat higher in winter than in summer, although in a very different context [24]. However, the present survey took place in the summer months and the discrepancy we observed between the proportion of abdominal and overall obesity was 10-fold that attributed to seasonal variation by these authors.

The higher prevalence of abdominal vs. overall obesity was mostly due to the high proportion of AO women. Physiological, socio-economic, or lifestyle factors affecting overall obesity in LMIC are well documented [1,25], and some studies deal with correlates of WC, but evidence for why one would preferentially develop abdominal but not overall obesity is not plentiful. As we studied women aged 35 to 70, age could be a factor, since it is thought to be related to preferential accumulation of abdominal fat [26], but even among the younger women in our study, one out of five had AO. The relatively high proportion of postmenopausal women in this population could also be involved, as menopause has been shown to be linked to an accelerated accumulation of central body fat [27], but although we also observed a somewhat higher proportion of AO among postmenopausal women, the prevalence was high in both categories. Reproductive history may also be a factor [28], as mean parity in our population was higher than in the population from which the original WC cut-points were derived [21]. But increases in WC (adjusting for BMI) have been observed in other countries independently of higher parity [29], and also we observed a quite high proportion of AO women even in the lowest parity category. A few authors have suggested that lifestyle factors such as sedentary behavior, high energy intake, total and type of fat intake, or lack of sleep could be linked to abdominal obesity independently of overall obesity, but evidence is generally scarce [30-32]. This could nevertheless be in accordance with the nutrition transition that Tunisia is experiencing [13,33]. Smoking [34,35] or alcohol consumption [31] have also been suggested, but these factors concern a tiny minority of Tunisian women. History of nutrition over the life course such as rapid infant weight gain [36] or exposure to severe undernutrition during the prenatal period [37] have been hypothesized to shift body fat distribution toward abdominal adiposity. They could be significant factors behind the large proportion of AO women in our study, especially in this population with birth dates ranging from 1935 to 1970 (a time frame which includes the troubled period before World War II, the war itself, and the pre- and postindependence periods). Finally, genetics or epigenetics are of course also possible factors [38], with genetic variability of adipose tissue deposits, including anatomical location, possibly interacting with past exposure to various burdens of infectious diseases [39].

On the whole, we found a marked but mostly similar socio-economic pattern of overall and abdominal obesity, and as for the concordance of the two types of obesity at the subject level, almost no independent association of socio-economic factors with AO status, with the exception of a rather mild association with profession. As discussed by some authors [13,40], women working outside the home may benefit from factors related to social and intra-household roles, such as reduced food stimuli, exposition to a healthier lifestyle, or desiring nicer body image, all of which would generally render them less prone to obesity (especially those with a higher job level). Nevertheless, the association of professional activity with abdominal-only obese status would appear to be a somewhat different issue; indeed, as we observed in the present study, only those with a “worker/employee” type of job appeared to be less prone to abdominal-only obesity than nonworking women, while it was not so for the “upper/intermediate” category. This could be linked to the possible association of WC with physical activity (independently of BMI) [30]. Indeed, the employee/worker category that initially grouped “lower-level” jobs with regard to socio-economic position, de facto comprises mostly “manual” jobs (detailed data not shown).

The unadjusted association with urban/rural area of residence was quite strong as often observed among women in LMIC [5], but was observed similarly for both types of obesity. Thus there was no association of urban vs. rural of residence with AO status, either adjusted or unadjusted. Concerning geographic variability, women from the more developed eastern regions were both more overall and abdominal obesity-prone than women from the less developed western regions, as was also observed for Tunisian adolescents [41]. However, as observed in some other countries, regional variability was much larger for abdominal obesity [42]. There was a much higher prevalence of AO status in the eastern vs. western regions, and this could result from a contextual effect linked to the general level of obesity (higher in the eastern regions) [43]. However, although the prevalence of both types of obesity was also much higher in urban vs. rural areas or in the higher vs. lower quintiles of welfare, no association between AO and these two factors was found.

Regional differences in lifestyle factors could also be involved. However, there is ample evidence that in such a nutrition transition context, there are huge differences in lifestyle factors, e.g., between the urban and rural environment or between different levels of household welfare (including evidence in Tunisia, although in a different age class [33]), but we did not observe any association of the urban/rural variable or the household welfare proxy with AO. Also, in such a nutrition transition situation, these lifestyle factors are to a great extent determined at a higher level of causation, by area of residence (urban vs. rural) or subject- and/or household-level socio-economic factors that are adjusted for in regional comparisons. One would need to hypothesize that residual adjusted differences in lifestyle among regions could explain the large residual regional contrasts in rates of AO. This is all the more unlikely since, as mentioned above, evidence linking these factors to abdominal obesity independently of BMI is quite scarce. It is also generally acknowledged that these factors contribute substantially less to variation in fat distribution than nonmodifiable factors such as ethnicity and genetics [44].

Indeed, over the course of history, many different populations have mixed and been assimilated to varying degrees in what is currently defined as Tunisian territory. Thus, beyond the diverse cultural influences, the relatively small population of 10 million Tunisians is also genetically quite heterogeneous, including genetic features linked to nutrition transition-related NCDs such as type 2 diabetes [45-47]. Such genetic differences could also partly explain regional differences in concordance between abdominal and overall obesity [38]. This would need to be confirmed by appropriate genetic assessment of the geographic disparities.

Regional differences in the history of nutrition over the life course and exposure to severe undernutrition during the prenatal period [36,37] (discussed above with respect to the high overall proportion of AO) may also explain the regional variability of the rate of AO. They could be significant factors, especially in this population whose birth dates ranged from 1935 to 1970, during which the eastern regions underwent more rapid socio-economic development than the western regions. Intergenerational and/or epigenetic factors could also be involved.

Although there is yet no definitive epidemiological evidence, endocrinal disruptors have been hypothesized to be specifically linked to abdominal obesity due to effects on hormonal factors, which, in turn, may influence lipogenesis toward more abdominal fat accumulation [48,49]. The fact that the more developed and industrialized regions of the east had the highest proportions of AO women would be in accordance with such a hypothesis, but data on endocrine disruptors in Tunisia are almost nonexistent.

The cross-sectional design has limitations regarding the analysis of the dynamics of body fat distribution and its cofactors over the life course [50]. Like in a number of studies pertaining to risk factors of chronic diseases, the 35- to 70-year age class was chosen, but not having included younger adults is a limitation. Like in most large-scale epidemiological studies, overall and abdominal adiposity were assessed by anthropometric proxies only. Apart from issues related to measurement techniques [51], one drawback of WC is that it does not distinguish between different types of abdominal fat accumulation, such as visceral vs. subcutaneous adipose tissues that may be linked to adverse health outcomes in different ways [52,53]. Beyond our characterization of body shape as AO or OO from the internationally acknowledged BMI and WC, some authors have proposed specific indices, such as the ABSI (a body shape index), whose use is not yet standard [54]. Lifestyle factors such as dietary intake or physical activity were not adjusted for in the models, but as discussed above, in such a nutrition transition context these factors are mostly determined at a higher level of causation by individual or socio-economic factors, and these were adjusted for in our comparisons. To our knowledge, this is the first large-scale study in a nutrition transition situation that compares abdominal and overall obesity both at the population and subject level based on a national representative sample. Using multinomial regression to analyze AO and OO status vs. concordance at the subject level and their relationships with environmental, physiological, and socio-economic factors is also original.