Discussion
This review offers a systematic look at sex differences over a wide geographical area. The studies included in the meta-analysis show that boys aged 0–59 months are much more likely to be wasted, stunted and underweight using anthropometric case definitions than girls. This indicates sex differences in susceptibility to undernutrition. The reasons currently provided for these differences vary and are often speculative rather than informed by direct evidence.
When stratified by region, the results also showed that boys are more likely to be wasted, stunted or underweight than girls. There were however some exceptions where ORs were reduced or reversed for boys with respect to undernutrition, in East Africa, Central America, South and South East Asia. The differences in Central America were based solely on one study, with a limited sample size and therefore need to be interpreted with caution. Our analysis potentially masks some of the complexities of regional variations in sex differences, particularly in South and South East Asia as many studies from these regions did not qualify for inclusion in the meta-analysis due to insufficient data. It is possible these differences might be under or overestimated. In reviewing the individual studies identified in the main search, results from this region are inconsistent and often conflicting compared with those coming from other regions of the world, such as Africa, which show a more consistent pattern of male disadvantage, a finding resonating with other studies.5 21 The inconsistencies in findings for parts of South and South East Asia, however, may be explained in part by well-described social preferences for men,22 and warrant further investigation. Such differences have also been described for under-5 mortality, with excess female child mortality for certain diseases, and according to socioeconomic status, birth order and family composition.23–26
These findings challenge commonly held assumptions within the nutrition community that girls are more likely to be affected by undernutrition. Recent studies focused on the relationship between wasting and stunting have also highlighted similar findings showing boys are more likely to be concurrently wasted and stunted than girls4 27–29 and have identified this as an unexpected finding.
We found that even where sex differences are reported, they are not always acknowledged or explored. Just over a quarter of studies (28%) did not provide any discussion on reported differences and 14% cited similar findings but did not consider causes. Where explanations for sex differences in the prevalence of undernutrition were offered, nearly half (49%) of the studies reviewed offered explanations related to social reasons or based on speculation or preconceived supposition rather than evidence. The search criteria used (which filtered articles to those which use terms related to sex or gender in the abstract) might have introduced some bias here with a potential overestimation of studies that report and explore the issue of sex differences.
When stratified by age, the meta-analysis also shows that boys are at higher risk across all age groups, though again, our analysis potentially masks some of the complexities in age as detailed analysis of different age groups was not possible. While the results for age show that boys are more likely to be stunted than girls, the ORs are lower in the older age group compared with the younger group. Limited data in the 24–59 month age category, especially for wasting and underweight, however mean results must be interpreted with caution. These tentative results might indicate any sex-specific risks differ at different ages: further study is warranted. Two studies exploring concurrent wasting and stunting28 29 found it to be a condition that affects children below 30 months more than it does older children, and found that sex ratios in undernourished children change with age, with a higher susceptibility for boys up to 30 months that then disappeared. Alongside other studies,30 they suggest that sex hormones, specifically testosterone, luteinising hormone and follicle-stimulating hormones might play a role in this. Selection effects might also contribute to this, whereby if boys are more likely to die than girls, the remaining pool of boys would represent healthy survivors.
Adair and Guilkey31 studied children in the Philippines and found men were more likely to become stunted in the first year of life (using the NCHS reference), but women were more likely than men to become stunted in the second year. They suggest differences in parental caregiving behaviours may partly account for this finding, but these were not measured in the study. Bork and Diallo32 also found evidence of interaction between age and sex in that the deficit in boys compared with that in girls increased between the first and second years of life, regardless of the indicator used. The differences in height status were however sensitive to the growth reference chosen; they were greater when assessed using the 2006 WHO growth standards than when using the NCHS growth reference.
Sex differences in undernutrition may vary not only by geographical area, but also over time. When diseases causing undernutrition known to be more severe among girls, such as measles, whooping cough and tuberculosis, disappear because of vaccination, lower transmission and better feeding, the disadvantage of boys might increase. Conversely, if efficient nutrition programmes are conducted, the disadvantage of boys might be reduced over the years.
Interpretation of these findings into implications for practice and policy is limited at this stage but does warrant consideration and some degree of change. As a minimum, the systematic collection and reporting of disaggregated data by age and sex should be included in the design of programmes and assessments in all settings. Where differences are observed, particularly in programme admissions, these should be interpreted in light of sex differences in population burden in order to draw conclusions as to whether programmes are proving equally accessible to boys and girls, and then the potential causes of these differences should be considered and/or investigated. At present, boys’ vulnerability to undernutrition is rarely a consideration in the design of nutrition programming, nor the formulation of policy. Moreover, some high-level international nutrition policies explicitly focus on the vulnerability of women and girls (eg, The Scaling Up Nutrition Movement Road Map for 2016–2020 Khara et al
28). Similarly, the recent Inter-Agency Standing Committee guidance on gender in humanitarian action33 recognises the inequity in food intake that may be faced by women and girls in crises but makes no reference to higher levels of undernutrition among boys. The absence of any reflection on gender, or the misuse of the term to highlight solely the health of women and girls, is likely to unintentionally reinforce inequalities in health.7 In the Nutrition for Growth 2020 summit (https://nutritionforgrowth.org/) and beyond, a major focus will be on inequities in undernutrition and how they affect different groups in different locations. The emerging findings from this review have significance in ensuring consideration of these sex differences through an equity lens.
Strengths and limitations
One of the strengths of this study lies in the systematic approach that was chosen and its primary objective to review sex differences in undernutrition over a wide geographical area. However, there are areas where bias has potentially been introduced.
First, screening for studies to be included in this study was conducted by only one of the authors. While we employed systems to ensure contentious articles were discussed among two or more authors, we recognise that not using double screening is a limitation.34
Second, the search strategy looking for explicit mention of sex or gender in the abstract might have biased towards studies that reported on sex and gender in the abstract, or towards studies that found a significant difference, and therefore sex differences might be under-reported or over-reported in this study. Likewise, the search may have limited the analysis as there are potentially missed studies which include sex as a variable in analysis but without focusing on mention of sex in the study abstracts. Similarly, there may be a degree of publication bias whereby sex differences are simply not considered or reported.
The search criteria also encompassed a large number of studies with differing objectives meaning a limited degree of homogeneity. Few studies directly assessed the true relation between sex and undernutrition. This analysis is therefore potentially biased by healthy survivors—those children that have survived to be included in studies. We do not believe however that our results would be significantly different considering the evidence presented on male vulnerability. We also recognise the possibility of an overlap in data used from sources such as Demographic and Health Surveys (DHS). By comparing the dates and locations of included studies, we have not been able to establish any overlap. Unidentified overlap, if it occurs, is therefore likely to be minimal in our review and unlikely to affect overall conclusions. Where multiple studies are available from the same country, we have established these to be from different regions and times, therefore their inclusion as independent studies is justified. We hope that our review will stimulate future work to explore not just intercountry differences but also intracountry/regional differences as this would help understand biological versus social reasons for any difference in male/female risks.
While this analysis included some secondary DHS data, the subject in question could benefit from a systematic analysis of DHS, Multiple Indicator Cluster Surveys and or nutrition survey data. Though it is not believed that the outcome of the ORs of sex differences would be different, further analysis might help improve understanding of some of the complexities of age, context, dual burdens of undernutrition and sex differences and the implications for programmers. This might also help towards explaining some of the between-study heterogeneity that we identified but were unable to explain with our analysis.
The rigour of findings of the analysis is limited in relation to age as the grouping and degree of available data potentially masks some of the differences at various stages of the lifecycle, similarly geographical differences might be biased towards studies included through the search.
The absence of data on other anthropometric measures, such as Mid-Upper Arm Circumference (MUAC), is also a potential limitation. In considering the implications of the differences highlighted here, in addition to biological and social explanations, it is necessary to consider how we measure and define undernutrition and whether sex differences are an artefact of the indices in use. The WHO growth standards describe the physiological growth within optimal environmental conditions and are separated by sex. These reference data from healthy well-nourished populations resolve sex differences to zero by expressing data as z-scores calculated using the appropriate male and female subset of the reference population. However, it is unclear if we would expect sex differences in undernutrition expressed in this way to be zero, when the distribution of weight and height in both sexes has been shifted away from the healthy reference range. Likewise, it is unclear if the loss of the same amount of body weight in a girl or boy would have the same physiological effect. If boys cope worse than girls when exposed to food shortages or disease and infection, this potentially highlights increased vulnerability over and above what is already accounted for by the standards.
In comparison, MUAC cut-offs are unadjusted and do not differentiate by sex or age (between 6 months and 5 years). This absence of adjustment may lead to a preferential inclusion of girls in programmes compared with what would be obtained if sex-specific standards were used as girls tend to have lower MUACs than boys. Though it has been shown to be a good predictor of mortality, sex differences in using MUAC to define undernutrition have not been widely studied.
Finally, the number of studies identified in the overall search that qualified for the meta-analysis was low. This was mainly due to a lack of presentation of disaggregated data. A recent Lancet series on gender equality, norms and health, highlighted the need for accurate disaggregated data.35
Implications for future research
This study is a step towards better understanding of sex differences in undernutrition and highlights the need to consider potential implications for policy and practice. Future research should aim to unpack the complexities related to age, biological and social risks (including gender norms) and context. In particular, we note that current papers are conjectural about reasons for observed differences. Any hypotheses should be more directly tested in future studies to further our understanding of sex differences in the context of undernutrition and subregional variations in order to determine the implications of these differences for programme staff and policymakers.
Future research will focus on a more detailed analysis of factors affecting outcomes for boys and girls such as epidemiological, demographic and social differences, explore the consequences of sex, age, and behavioural differences in nutritional outcomes and mortality. The impact of using differing anthropometric measurement and indices should also be explored to better understand how differing methods detect the most vulnerable children and explore how substantial sex differences are.