Conceived and designed the experiments: RB AAM. Performed the experiments: RB. Analyzed the data: RB AAM. Wrote the paper: RB AAM.
The authors have declared that no competing interests exist.
The identification of biological and ecological factors that contribute to obesity may help in combating the spreading obesity crisis. Sex differences in obesity rates are particularly poorly understood. Here we show that the strong female bias in obesity in many countries is associated with high total fertility rate, which is well known to be correlated with factors such as low average income, infant mortality and female education. We also document effects of reduced access to contraception and increased inequality of income among households on obesity rates. These results are consistent with studies that implicate reproduction as a risk factor for obesity in women and that suggest the effects of reproduction interact with socioeconomic and educational factors. We discuss our results in the light of recent research in dietary ecology and the suggestion that insulin resistance during pregnancy is due to historic adaptation to protect the developing foetus during famine. Increased access to contraception and education in countries with high total fertility rate might have the additional benefit of reducing the rates of obesity in women.
Recently the number of obese people on earth exceeded for the first time the number of people who do not get enough to eat
There are enormous differences among countries in obesity rates, from less than one percent of adults in Ethiopia and Cambodia to more than sixty percent of adults in Nauru and the Cook Islands
Variation in and relationships among life history traits such as lifespan, reproductive effort and weight gain can be understood by studying them at a variety of scales, from longitudinal studies on individual subjects to large international data sets. Although international data are by their nature very coarse in resolution, they do tend to capture a wider range of variation in economic and cultural factors than more focussed experimental, longitudinal or neighbourhood-level studies. As such they are an indispensible tool for identifying the range of phenotypically plastic strategies that humans are capable of, and generating hypotheses for more direct testing. This is particularly true for questions that involve sex differences. For example, Maklakov
The striking pattern of high female obesity relative to male obesity in many nations requires explanation
We used standardised obesity data from the World Health Organisation's Global Database on Body Mass Index (WHO 2010) which includes the results of a large number of surveys and studies. For many countries, male and female obesity rates were available from the same study. Where this was not the case, data for adult men and women were always obtained from samples within 3 years of one another. To avoid confounding effects of temporal trends in obesity, we also only used data from surveys post 1998, and for countries where there were multiple surveys post-1998 we used the most recent. There were suitable female obesity data for 137 countries, but suitable male data for only 94 of these countries. Data represent percentage of adult (older than 15 years) men and women with Body Mass Index greater than or equal to 30.0, the standard WHO definition of adult obesity rate. We used the Central Intelligence Agency World Factbook (
We estimated pairwise Pearson's correlations and fitted multiple regressions in JMP 7.0.2. When building a multiple regression to explain female obesity, we fitted male obesity (and vice versa) to represent all of the broad factors such as food availability and national diet that cause obesity in general, and then fitted other variables to explain the sex difference in obesity. We used Mallows
Although male and female obesity rates are strongly correlated (
Female obesity | Male obesity | |||
r | N | r | N | |
Male obesity | 0.91 |
94 | ||
GNI PPP | 0.51 |
136 | 0.33 |
93 |
Population density | −0.04 | 135 | −0.10 | 92 |
Urbanisation | 0.55 |
135 | 0.44 |
92 |
Female education | 0.53 |
121 | 0.44 |
85 |
Male education | 0.43 |
121 | 0.31 |
85 |
Contraception | 0.36 |
115 | 0.06 | 74 |
Total Fertility Rate | −0.36 |
136 | −0.05 | 93 |
Infant mortality rate | −0.55 |
136 | −0.38 |
93 |
Latitude | 0.31 |
136 | 0.11 | 93 |
Gini Index | −0.039 | 112 | −0.052 | 75 |
***P<0.0001
**P<0.001
*P<0.01
Multiple regression analysis tells quite a different story. The first and by far the most important predictor of female obesity is male obesity rate which explains 83.0 percent of the variance in female obesity. The best multiple regression model (the smallest model for which
Bubble sizes represent Total Fertility Rate.
The best regression for the smaller subset of data that included Gini includes male obesity, TFR, contraception use and income inequality (R2adj = 0.86, F4,55 = 90.4, P<0.0001). This model includes far fewer countries because of the number of missing values for the Gini index. Higher female obesity was associated with less contraceptive use (
In all of the other multiple regression models that we tried in which male obesity was the first term fitted and the second term added was not TFR, the second term we added indicated that obesity was associated with socioeconomic disadvantage and low status of women. This comes about because high total fertility rates, infant mortality, low GNI PPP, high Gini index, few years of female education, low contraception use, and low urbanisation all tend to be correlated as a suite of traits. The associations between these traits are well documented, both within and among countries
Male obesity was strongly correlated with female obesity, but once these effects had been statistically controlled for (
Our results are consistent with smaller-scale studies that document an association between low income, material deprivation, food insecurity or minority status and increased obesity in women but not (or less often) in men
Several recent studies across a variety of countries and circumstances from rural Iraqi women to middle-income Mexicans to Americans of all ethnicities and incomes suggest that parity (the number of times a woman has given birth) is positively associated with increased obesity risk
It appears, therefore, that the role of parity as a trigger for excessive weight gain may be a combined effect of the nutritional and demographic transitions. We predict that the countries with the greatest female bias in adult obesity will be those in which a large proportion of families have escaped from hunger, yet women still have high total fertility rates. If this is true, then access to contraception and the education of women may be just as important in combating obesity as they are in curbing population growth
Our results also suggest that high income inequality within countries may elevate the incidence of obesity in women but not in men, and that these effects are additional to the effects of parity. The positive contribution of the Gini index to the multiple regression analysis for female obesity suggests that female obesity is governed not only by average wealth, but also by the variation in wealth within societies. Inequality of income is known to be an important determinant of the levels of violence, risky behaviours, accidental death, mental illness, anxiety, and teenage pregnancy within societies
Evolutionary perspectives and large scale correlative studies can play an important role in generating mechanistic hypotheses for health problems like obesity and type 2 diabetes
Brooks, Simpson and Raubenheimer
Many thanks to Stephen Simpson for helpful discussions and comments on the manuscript.