Body mass index and waist circumference trajectories across the life course and birth cohorts, 1996–2015 Malaysia: sex and ethnicity matter

In this study, the age trajectory of BMI increased steeply during adulthood and plateaued after age 60. The increasing trajectory in early adulthood was similar to previous APC studies on obesity [12] and mean BMI [9,10,11, 14,15,16,17]. However, their trajectories varied in the older age groups. A few studies reported a decreasing [8,9,10,11,12], whereas others reported a plateauing trend [14,15,16]. Our findings are consistent with the existing knowledge of decreased BMI among older people due to loss of skeletal muscle mass [18], reduced appetite, and malnourishment/poor nutrient intake [19]. In contrast, monotonic increasing age trajectories of waist circumference, which concurred with findings from a Norwegian study [20], were observed in both sexes. These increasing age trajectories can be attributed to the age-related increase in body fat, as studies have shown that the percentage of body fat increases with age until 80 years old [21] by about 1% per decade [22].

The monotonic increasing cohort trajectories of BMI and waist circumference observed in the present study were consistent with most studies reported elsewhere [6, 8, 9, 12, 14, 16, 23]. Cohort effects refer to inter-generational differences in experiences, exposures, behaviors, and socioeconomic factors between individuals of different birth cohorts. These differential exposures could have affected the habits and health behaviors in their formative years and persisted throughout their life course, thus leading to divergent health outcomes between different generations.

For instance, older cohorts born before the industrialization (the 1970s) and globalization era (1980s–1990s) may have experienced higher food scarcity, more labor-intensive occupations, and more physically active modes of transportation [24]. Therefore, they were less likely to be exposed to obesogenic risk factors such as uptake of energy-dense food, physical inactivity and sedentarism. In contrast, the progressively obesogenic environment may have predisposed the younger generations (particularly the Millennials or Generation Y born between 1981–1996) [8] to higher BMI and waist circumference than their counterparts in preceding cohorts. Possible reasons include increased availability, accessibility, and affordability of energy-dense nutrient-dilute foods, shift from labor- to capital-intensive occupations, increased screen time, and reduced active commute.

In the present study, a more profound increasing age trajectory of BMI [14] was observed among women, particularly as they age. This finding concurred with those reported in a cohort study among the rural population of The Netherlands. Evidence from a series of cross-sectional surveys in England demonstrated a more pronounced inverse association between BMI and height in older adults and women than in men, with little change over time [25]. Therefore, this increasing sexual dimorphism in the BMI–height associations could have led to the observed sex divergence of BMI trajectories over the life course, particularly during late adulthood.

On the other hand, other studies in high-income and upper-middle-income countries did not observe differences in age trajectories by sex for mean BMI [10, 11, 15] and prevalence of obesity [12]. These findings were in keeping with those observed from the Global Health Observatory, where equitable obesity rates between men and women were observed among most developed nations [26]. According to Grantham & Henneberg’s estrogen hypothesis, the preponderance of men’s exposure to environmental estrogen-like substances, such as xenoestrogen in soy products and polyvinyl chloride, that are related to the superfluous nature of developed nations, could have “feminized” the men, resulting in equitable obesity rates between men and women [27], particularly among those born in more recent cohorts.

We also observed sex variations in age and cohort trajectories of BMI and waist circumference: sex divergences in age and cohort trajectories of BMI but sex convergences in age and cohort trajectories of waist circumference. Sex dimorphism in fat distribution [28, 29] and physiological differences [30] may explain these variations. Since women tend to accrue more weight after menopause, they may be more likely to have higher BMI than men, thus explaining the increasing sex divergences in BMI trajectories as they age. On the contrary, men tend to store fat in visceral adipose tissue (VAT) in the deep abdominal region, compared to premenopausal women who preferentially keep excess fat in the subcutaneous adipose tissue (SAT) depots surrounding low extremities such as hips and thighs [31], thus predisposing men to greater waist circumference than premenopausal women. However, as age advances, SAT decreases, and VAT increases with age [22], with women having almost double increases in mean waist circumference than men [32]. Therefore, sex differences in VAT diminished as they age, resulting in the sex convergence in waist circumference over the life course.

The present study revealed that the age and cohort trajectories of BMI and waist circumference varied by ethnicity. Chinese had the least pronounced increasing trajectories of BMI. A recent APC study using four national longitudinal cohort studies also observed persistent ethnic differences in BMI trajectories across the life course between the Black, Hispanic and White [9]. Another APC study among the New Zealand population found ethnic differences in BMI trajectories between the Maori and non-Maori populations, with an increasing cohort trajectory only observed among the Maori [11]. For waist circumference, despite an almost similar magnitude of trajectories between Chinese and Indians, Indians had the highest overall waist circumference. This finding concurred with that reported in a local study where Indians have a greater likelihood of abdominal obesity than other ethnic groups across the life course [33].

Ethnic heterogeneities across the life course and birth cohorts can be attributable to the well-established ethnic variations in total body fat (TBF) percentage [34] and SAT and VAT fat depots [35]. Besides, the unique dietary habits among ethnic groups could also be the contributing factors. A greater preference for healthy-based over the Western-based (high in fat, sugar, and salts) food pattern among Chinese adolescents than their Malay counterparts [36] may likely explain healthier BMI and waist circumference trajectories among the Chinese, particularly those born more recently. In contrast, higher content of carbohydrates, saturated fatty acids, and trans fatty acids in the Indian diet may contribute to higher waist circumference across ages and cohorts [35].

Besides, genetic predisposition to certain diseases could also contribute to ethnic differences. Several ethnic-specific single nucleotide polymorphisms (SNPs) or single gene mutations associated with obesity have been identified [37]. In addition, racial/ethnic differentials in C-reactive protein (CRP) levels, a known risk factor of abdominal obesity [38], could be at play. Studies among the multi-ethnic US and Canadian populations had unequivocally reported that the Chinese had the lowest mean CRP level compared to other ethnicities such as Europeans, South Asians, aborigines [39], Caucasians, African Americans, and Hispanics [40]. These findings likely explain the more favorable obesity trajectories among the Chinese in the present study, who had the least profound increasing BMI and waist circumference trajectories, particularly among older adults and those born in more recent cohorts.

It is worth noting that due to the APC identification problem, strong assumptions must be made to discern the APC effect. These assumptions or constraints, however, cannot be made on the basis of the data [41]. To determine either cohort or period is more likely to be the driving temporal factor of obesity, one could compare the age trajectories produced and deduce which seems more plausible [41]. In our case, we would argue that the age trajectories of BMI and waist circumference predicted from the primary models that assumed no period effect are theoretically plausible compared to those predicted from the alternative models that assumed no cohort effect.

The age trajectories of BMI and waist circumference predicted from the primary models concurred with those reported in previous studies, where mean BMI [42] and waist circumference [43] increased with age until 60 to 70. Such trajectories are also consistent with ageing-related body composition changes such that body fat develops up to the eighth decade of life and reduces afterward [44].

On the other hand, the age trajectories of BMI, predicted from the alternative model, resembled the parabolic age trajectory of obesity among the U.S. adult population in a previous HAPC study that explicitly assumed no cohort effect [45]. As also argued by Bell and Jones [41], while BMI is known to be negatively associated with advancing age due to sarcopenia and survival bias; however, the relatively sharp and early decline in BMI and waist circumference at age 50, as observed in the alternative model (Appendix XI), are rather unlikely. Furthermore, the less profound age trajectories from the alternative models depict a much lower predicted mean BMI and waist circumference (in fact, well below the BMI cut-off point of 25.0 kg/m² for overweight and waist circumference cut-off of 90 cm for men) among the Malaysian adult population, which is, again, unlikely given the fact that about 45% of Malaysia adults are overweight (Appendix II).