Strengths and weaknesses

The strength of our analysis is that it tests a pre-existing hypothesis using a global data set of routinely collected data from 184 countries with the capacity to adjust for potential confounding factors such as economic activity, prevalence of malnutrition and latitude. The availability of data from 53 countries from the African continent also permits a subset analysis, which permits further testing of the associations observed within a slightly less heterogeneous group of nations.

Our analysis has a number of limitations that require consideration. The ecological study design is one that has a variety of methodological problems compared with other epidemiological study designs.20 However, when testing the hypothesis that a population-based stressor may impact on a population-based outcome, it may be the best study design available to identify major population-level determinates of disease.21 However, the use of routinely collected data makes the generalisability of any associations observed restricted to comparisons within countries, and other study designs will be required to explore these associations within national borders. The associations that we have observed between both high prevalences of HIV/AIDS and TB and a low sex ratio at live birth are not evidence of a causal relationship, although we cannot exclude this possibility for one of the exposures. There are other factors in common between countries with a high prevalence of both of these infectious diseases, which we were unable to adjust for in our analyses that may shed more light on the causal pathways involved. Such factors include the genetic/ethnic mixture of the national populations,22 ,23 the preference of certain populations for male offspring,24 ,25 increased paternal age26 and local geography, which will substantially modify the environment. It is possible that some of these exposures that are intrinsic to or associated with regional cultures may explain some of the associations that we report, and future studies of environmental factors that may modify sex ratio would be enhanced by consideration of both the ethnicity of individuals concerned and cultural factors including religious beliefs. The nature of our data set does not permit these factors to be considered in this analysis, although one study design that may increase understanding of the associations of interest could be a time series analysis of national data over the period that HIV and TB prevalence changed. The absence of an association between either HIV or TB prevalences and the sex ratio from the non-African countries was not surprising, as the range of prevalences for both infectious diseases in these regions were relatively low, thus limiting the power of these analyses.

To our knowledge, this is the first study to present data on the association between the prevalence of either HIV/AIDS or TB and the sex ratio at live birth. There have been many studies utilising routinely collected data and examining the association between acute and chronic environmental exposures and sex ratio at live birth. Our data are consistent with an early paper by Willard and Trivers12 who used data from mammals to speculate that the sex ratio at live birth decreases in times of scarcity, another challenge to the health of populations. However, they are less consistent with the observations that the sex ratio increases during times of extended war6 and famine15 and also appears to be increased in individuals with hepatitis B infection.3 As these exposures are indisputably population stressors, it is clear that untangling the relationships between exposures and the human sex ratio is a challenging area of investigation, and the nature, duration and intensity of the exposures may all be factors that warrant consideration in future studies. Reports that higher levels of psychological stress4 and exposure to severe life events27 during early gestation are associated with a lower sex ratio at live birth may be relevant in providing a potential biological mechanism to explain our observations.

From a practical perspective, as well as suggesting that the sex ratio at live birth may have potential as a sentinel health indicator, our observation of 1%–2% decrease in sex ratio in countries with a high prevalence of HIV/AIDS and TB has implications for the demographic structure of these countries in the future, particularly if these patterns of lower levels of male birth are compounded over decades or longer. This is a relatively large size of association that is of a similar magnitude to the change reported in economically developed countries by Davis et al2 over the second half of the 20th century.