Abstract

Background

Stunting remains a considerable public health problem globally, and sex differences in prevalence have been documented. While many risk factors for stunting have been identified, few studies examine how these factors may contribute to sex differences. We test whether: (1) boys and girls are differentially exposed to stunting risk factors, (2) boys and girls respond differently to similar exposures, and (3) these associations are sensitive to the growth measures used.

Methods

Data comes from 7486 Zambian infants, aged 0鈥�23 months, participating in the SUN (Scaling Up Nutrition) 2.0 Program baseline survey. Surveys and dietary recalls were collected from primary caregivers, and anthropometry was collected for caregivers and children. Stunting was defined as height-for-age z-score (HAZ) <-2. Descriptive statistics and adjusted multilevel logistic regression models controlling for age and province were used to identify sex differences in exposures and risk factors for stunting, respectively. Interaction terms between exposure and infant sex were added to test for sex differences in response. Sensitivity testing with alternate measures of infant size, including height-for-age difference (HAD), was conducted.

Results

Boys were more likely to be stunted than girls (34.5% vs. 25.7%, respectively). Numerous maternal, care giving, diet, and household characteristics were associated with the risk of stunting, but few sex differences were seen in exposure to these factors. Only one risk factor鈥攎aternal marital status鈥攕howed evidence of moderation by sex at the p鈥�<鈥�0.05 level, while a limited number of risk factors did so at the p鈥�<鈥�0.10 level. At all ages, boys were longer than girls, and there were no sex-specific differences in the patterns of HAZ or HAD with age. Results were robust to alternate specifications.

Conclusions

Our results show that Zambian boys have lower mean HAZ scores than girls and a greater prevalence of stunting throughout the first two years We do not find strong evidence that infant feeding practices, environmental exposures, or care giving differ consistently between boys and girls or that boys and girls respond differently to these exposures. Our results instead indicate that further investigation of prenatal factors and/or measurement issues is needed.

Background

Malnutrition among women and children is a major public health problem worldwide. Globally, 148听million children under-five years of age (22.3%) were stunted and 45听million children (6.8%) were wasted in 2022 [1]. Despite global improvements from the 1990s, the burden of malnutrition, especially stunting, remains high in Africa, where the stunting rate is 30%. Malnutrition during childhood is associated with both short-term morbidity and mortality [2, 3] and longer-term functional, reproductive impairment [4], and a risk factor for later obesity and cardiometabolic disease [5], making improvements in child nutrition critical for short and long-term health.

Similarly to other African countries, Zambia has a persistently high proportion of undernourished children. 35% of children under age 5 are stunted [6], and this proportion is significantly higher for boys (38%) than girls (31%). Such sex differences in stunting prevalence have been found in other studies [7], particularly those from Sub-Saharan Africa and in households of lower socioeconomic strata [8,9,10]. While these patterns have been widely documented, less research has examined why this difference between boys and girls exists. Research in other contexts suggests that feeding or child care practices may differ by child sex [11, 12], with boys more likely to receive non-breastmilk foods and liquids earlier than girls [10, 13] or to be given routine medical care [14]. Alternatively, researchers have suggested that the higher prevalence of stunting in boys may stem from greater biological vulnerability. Boys have higher energy needs than girls due to their larger body sizes and more rapid growth beginning in utero [15,16,17]. Boys also have less competent immune systems [18, 19], leading to higher morbidity and mortality from common childhood infections such as lower respiratory infections, diarrhea, and malaria [20, 21]. Together, these higher energy needs and more frequent morbidity may place boys at greater risk for stunting and acute malnutrition [11]. Conversely, another line of research suggests that instead of, or in addition to, social differences in child care or biological differences in vulnerability, sex differences in the prevalence of stunting may stem from measurement issues, such as study design [22], the growth standards used [10, 23, 24], and reliance on height-for-age cut-points to assess growth faltering [25, 26].

We test the proposed pathways contributing to sex differences in stunting in a large sample of Zambian infants and young children, under 24 months of age. First, we examine whether boys and girls are differently exposed to the wide range of socioeconomic, environmental, and behavioral risk factors for stunting identified in the literature [26] and included in the UNICEF Conceptual Framework on the Determinants of Maternal and Child Nutrition [27]. Next, we use interaction models to test whether boys and girls respond differently to these exposures, providing potential evidence of sex differences in biological vulnerability. We then examine two specific putative mechanisms, feeding practices and measurement issues, previously described in the literature. We use age-stratified analyses to examine whether feeding practices contribute to differential risk of stunting in earlier vs. later infancy, a pattern seen in some studies [10, 13]. To address concerns that sex differences in stunting may stem from the use of cut-points and/or sex-specific differences in the performance of growth standards across populations, we repeat our analyses using height-for-age difference (HAD) from the WHO Mutlicentre Growth Reference Study (MGRS) [28] median instead of z-score based stunting cut-points.

Methods

Sample and study design

The data come from the Zambia Scaling Up Nutrition (SUN)/First 1000 Most Critical Days Programme (MCDP II), which was initiated in 2018 with the aim of reducing stunting among children less than 24 months old.

The SUN/MCDP II baseline survey [29], undertaken by the USAID-supported SUN Learning and Evaluation project in 2018/2019, in collaboration with the National Food and Nutrition Commission, working with line Ministries, collected representative data in 30 districts where the SUN/MCDP II was operating.

The household survey targeted 7,500 households, 250 households in each of 30 districts, with children under 2 years. The household sample was drawn using stratified two-stage sampling: first,听random selection based on Probability Proportional to Size sampling of 10 enumeration areas (EA) in each of the 30 districts, and second,听in each EA, random selection of 25 households with children 0鈥�23 months based on a sampling frame of all households with children in that age group. Five additional households were also randomly selected per EA as potential replacements for instances of non-response or non-availability of residents. In cases where a selected EA had fewer than 25 households or a sample could not be drawn due to non-availability of residents, an EA was randomly selected from the neighboring EAs, and all households in that EA were listed. Household lists from both EAs were combined, and a sample of 25 households was randomly selected from that list.

In each selected household, the biological mother or primary caregiver of the child was interviewed, and anthropometric measurements (height/length and weight) of the child and biological mother were taken by a trained interviewer. If a household had more than one eligible child, the data collection system randomly selected one child. In total, 7,486 eligible household with children under 24 months of age were reached (7,177 biological mother-child pairs and 309 other primary caregiver-child pairs). Household survey data were electronically collected using a structured questionnaire programmed in Survey Solutions software in English and seven local language translations (Lunda, Luvale, Bemba, Chinyanja, Chitonga, Silozi, and Kaonde).

Ethical approval

The baseline survey protocol received ethical approval from the University of Zambia Biomedical Research Ethics Committee. Informed consent to participate was obtained from the parents or legal guardians of participants under the age of 16.

Measures

Sociodemographic data

Maternal, child and household characteristics were assessed through administered surveys. Mothers reported their age and that of their child, their education level (none, preschool, primary, junior secondary, senior secondary, or higher) and marital status (single, monogamously married or cohabiting, polygynously married, or widowed/separated/divorced), and the economic activity of the household head (farmer, formal employment (salaried), informal employment (self-employed), business man/woman (commercial), charcoal burning, casual labour, student, none or other). Due to low numbers in some categories, these economic activities were grouped into four categories (none, farmer, informal employment, formal employment).

Anthropometry

Child length was measured using standard protocols. Stunting was defined as a height/length-for-age z-score of <-2 in comparison to the WHO MGRS standard [30]. Height-for-age difference was also calculated by subtracting a child鈥檚 length from the age- and sex-specific WHO median [31]. Child birth weight was collected from maternal report and verified from health records. Maternal height and weight were collected using standard protocols. Mother鈥檚 stature was considered short if their height was <鈥�145听cm. Maternal BMI was calculated as weight(kg)/height(m)², and their BMI was classified as moderate/severe thinness (BMI鈥�<鈥�17.0), underweight (17鈥夆墹鈥塀MI鈥�<鈥�18.5), normal (18.5鈥夆墹鈥塀MI鈥�<鈥�24.99), overweight (25鈥夆墹鈥塀MI鈥�<鈥�30), and obese (BMI鈥夆墺鈥�30).

Diet measures

Mothers or primary caregivers completed a 24-hr dietary recall for the index child. This recall was used to define breastfeeding status and child diet quality. Children were considered breastfed if they had received breastmilk in the previous 24-hours. Children under the age of 6 months were considered exclusively breastfed if they had not received non-breastmilk solids or liquids in the previous 24-hours. Conversely, children under 6-months who received any non-breastmilk solids or liquids in the 24-hour recall were considered to received other foods and liquids early and not be exclusively breastfed. Prelacteal feeding was defined as the introduction of any non-breastmilk liquids or solids provided to infants prior to the initiation of breastfeeding.

Three indicators were created for children aged 6鈥�24 months using the WHO/UNICEF IYCF indicators [32]: minimum diet diversity, minimum meal frequency, and minimum acceptable diet. The list of foods gathered from the recalls were grouped into food groups based on WHO IYCF guidelines [33], and children were assigned a score of 1 for each food group consumed. Children were considered to meet the recommendations for minimum diet diversity if they consumed at least 5 of the 8 identified food groups (breastmilk, grains, roots and tubers; pulses; dairy products; flesh foods; eggs; vitamin-A rich fruits and vegetables; and other fruits and vegetables). Minimum meal frequency was defined by breastfeeding status and child age. Breastfed children were considered to meet the minimum meal frequency if they consumed at least two meals of solid, semi-solid or soft foods at ages 6鈥�8 months or 3 meals from 9 to 24 months. Non-breastfed children aged 6鈥�24 months were considered to meet the minimum meal frequency if they consumed at least 4 meals per day.听Minimum acceptable diet was defined as meeting the minimum diet diversity and minimum meal frequency for breastfed children aged 6-24 months and as meeting the minimum diet diversity and minimum meal frequency as well as at least two milk feeds for non-breastfed children aged 6-24 months.听

We also looked at the consumption of key food groups in the past 24-hours, including iron-rich foods, vitamin A rich foods, and animal source foods. Iron-rich foods were defined as consumption of foods, including flesh foods rich in iron, foods fortified with iron designed for infants and young children, or foods fortified with iron in the home. Vitamin-A rich foods include vegetables, roots and tubers such as pumpkin, carrots, squash or sweet potatoes that are yellow or orange on the inside or fruits such as ripe mango, ripe papaya, peaches, passion fruit, and red palm oil fruit or pulp. Finally, animal source foods include meats, fish, eggs, and dairy products.

Maternal diet quality was assessed based on whether she met the FAO Minimum Dietary Diversity for Women. Women were considered to have met this indicator if they had consumed five or more of the following food groups: grains, white roots and tubers, and plantains, pulses (beans, peas and lentils), nuts and seeds, dairy, meat, poultry and fish, eggs, dark green leafy vegetables, other vitamin A-rich fruits and vegetables, other vegetables and other fruits in the past 24听h.

Child care and health

To assess for diarrhea, mothers were asked whether their child had experienced 3 or more loose stools in one day in the past 2 weeks. In addition, mothers who answered in the affirmative were also asked whether they had taken their child to the clinic. Mothers were also asked about routine preventative health care services including whether their child had received deworming treatment or vitamin A supplementation in the past 6 months. Maternal participation in community-level nutrition-specific interventions, such as social and behavior change communication (SBCC) messaging around maternal diet and child feeding (e.g. diet while breastfeeding, feeding a sick child and exclusive breastfeeding) and growth monitoring, were also assessed.

Mothers were also asked about specific feeding practices and perceptions that could impact infant feeding, including their use of gripe water and their perception of child appetite. Mothers reported if they had ever given their child gripe water, an herbal mixture [34]. Gripe water is used in many SSA contexts as a treatment for colic or perceived abdominal pain and less commonly as a prelacteal food [35, 36]. Mothers were asked whether they thought their child鈥檚 appetite was poor, average/moderate, or good.

Household measures

Food insecurity: The Household Food Insecurity Access Scale was used to assess food insecurity. This 9-item scale assesses respondents鈥� perception about food-related vulnerability and behavioral responses to inadequate quality or quantity of food in the past 30 days [37]. Responses are coded based on frequency of occurrence (0-never to 3鈥�>鈥�10 times in past 30 days) and summed for a range of 0鈥�27. Households were classified as food secure, mildly food insecure, moderately food insecure, or severely food insecure based on their patterns of responses following UNAID guidelines [37].

Household resilience to lean seasons and environmental shocks was measured using the FAO鈥檚 HH Coping Strategy Index (CSI; [38]) as a proxy. Households were considered more resilient if they did not need to undertake certain coping strategies in the previous 12 months. Responses to the 12-item scale were weighted for severity and a total score calculated for the household. Households were categorized as more resilient (0鈥�40%) or not resilient (>鈥�41%) based on their total score.

Household water, sanitation and hygiene measures were assessed through surveys and interviewer observation. Houses were considered to practice essential hygiene if they met three conditions: (1) had access to an adequate quantity of water from a safe source with correct treatment and storage, (2) used a covered toilet/latrine for disposal of feces, and (3) had hand washing facilities with soap and water within a one-minute reach. Household food handling was defined as unsafe or safe based on three domains (processing, preparation and storage) based on USDA鈥檚 Basics for Handling Food Safely (). Exposure to animal feces was coded based on observation of evidence for animal presence (feathers or feces) in the child鈥檚 play area or maternal report that animals are present within the surrounding area, at least once per week.

Data analysis

Descriptive analysis was used to test for differences in household, maternal and child characteristics by child sex using t-tests for continuous variables and chi-square tests for categorial variables. Multilevel logistic regression models adjusted for child age were used to test first whether putative risk factors were associated with child stunting in the pooled sample. An interaction term between the exposure and child sex was next added to these models to test whether child sex moderated the association with stunting. Variables that showed evidence for potential effect modification (interaction p-value鈥�<鈥�0.10) were then entered into sex-stratified models adjusted for all stunting risk factors. All models include random intercepts for enumeration area (level 2) and district (level 3) to account for correlation among clustered observations.

Analysis testing feeding practices were similarly structured. Sex differences in diet were first tested with descriptive tests, then with multilevel logistic models in the pooled sample, and finally with the inclusion of an interaction term between the feeding practice and sex. Analyses were initially stratified by 6-month age groups (0鈥�5, 6鈥�11, 12鈥�17 and 18鈥�24) to account for developmental differences in feeding practices. However, few differences were seen in the later age groups and the subsequent analysis focused on infants under 6 months and those older than 6 months.

Finally, we ran similar pooled and interaction models using a continuous measure of HAD to test the sensitivity of our results to the use of the 鈭掆��2 LAZ cut point for stunting and to account for differences in the growth patterns of boys and girls relative to the WHO MGRS standards.

Sensitivity analysis

We re-ran our models with several alternate outcomes to test whether our results were robust to measures of stunting and undernutrition. We ran similarly specified models with continuous LAZ, severe stunting (LAZ<-3.0), and underweight (WAZ<-2.0).

Results

Sex differences in stunting

Boys were significantly more likely to be stunted across all ages (Fig.听1).

Fig. 1

Patterning of stunting with age

Sex differences in exposures

Relatively few exposures differed by child sex (Table听1). Boys were less likely to have low birth weight and were more likely to have had diarrhea in the past two weeks and to have received vitamin A supplementation. Boys under 6 months of age were less likely to be exclusively breastfed and were more likely to receive gripe water. No sex differences were seen in maternal or household exposures.

Table 1 Differences in maternal, child, and household exposures by child sex

Risk factors for stunting

To test whether child sex moderates the effect of exposure to stunting risk factors, we first tested whether exposures were associated with stunting in both sexes and then added an interaction term between the exposure and sex (Table听2). Many child, maternal, and household characteristics were associated with the risk of stunting for both sexes. Among these factors, maternal stunting, adolescent mothers, low birth weight, moderate and severe household food insecurity, and rural residence were all associated with an increased risk of stunting. Conversely, having a mother with a secondary or higher education, having a mother receive education on exclusive breastfeeding or complementary feeding, meeting the minimum diet diversity, meeting the minimum acceptable diet, consuming iron rich foods, receiving vitamin A supplementation, consuming gripe water, having a head of household with formal employment, living in a household practicing essential hygiene, living in a household with safe food handling, or living in a household with greater resilience were all associated with a reduced risk of stunting.

Table 2 Evidence for moderation by sex in risk factors for stunting

Moderation by child sex

Despite these many factors associated with stunting, only a few factors (perception of child appetite, education about exclusive breastfeeding, exclusive breastfeeding, and use of gripe water) showed evidence for moderation by sex at the p鈥�<鈥�=鈥�0.10 level and only one (maternal marital status) at the p鈥�<鈥�=鈥�0.05 level. The association between maternal marital status and stunting differed in boys and girls from mothers in monogamous marriages/partnerships and mothers who were widowed, separated, or divorced (Fig.听2a). Girls with monogamously partnered or widowed, separated, or divorced mothers had significantly lower risk of stunting than those with single mothers, an association not seen in boys. Boys perceived as having an average appetite were more likely to be stunted than those perceived as having a low or higher appetite (Fig.听2b). Conversely, girls perceived as having an average appetite were less likely to be stunted than those perceived to have low or high appetite. Compared to boys, whose probability of stunting did not differ by exclusive breastfeeding status, girls under 6 months of age who were exclusively breastfed were more likely to be stunted in minimally adjusted models (Fig.听2c). While the sample of children receiving gripe water is small (18.6%), the risk of stunting is comparatively lower for boys who received gripe water (26.7%) than those who did not (36.7%); while significant, the magnitude of difference for girls is less pronounced (25.5% vs. 21.4%; Fig.听2d). Of these interactions by sex, only the use of gripe water in boys and exposure to exclusive breastfeeding education for girls remained significant in sex-stratified models that fully adjusted for other child, maternal and household characteristics associated with stunting risk (Supplemental Table 1).

Fig. 2

Evidence for moderation by child sex

Predicted probability from model adjusted for age and province for variables that showed evidence of moderation in Tables听2 and 3; a: maternal marital status, b: caregiver perception of child appetite, c: exclusive breastfeeding to 6 months for children鈥�>鈥�6 months of age; d:use of gripe water, e: water intake before 6 months of age

Table 3 Age stratified analysis of sex differences in feeding exposures, association with stunting, and potential moderation by sex

Age-stratified analysis of infant and child diet

We examined differences in diet before 6 months and from 6 to 24 months between boys and girls. Boys under 6 months of age were more likely to receive water or any solid foods in the previous 24听h compared to girls; however, neither of these dietary exposures was associated with stunting in the pooled sample. Water intake showed only marginal moderation by sex (p鈥�=鈥�0.08) before 6 months of age. While boys under 6 months were more likely to receive water in the previous 24听h, girls who consumed water in the previous 24听h had a lower predicted probability of stunting than those who did not (20.2% vs. 13.1%, respectively) while no differences were seen in boys (26.2% vs. 26.6%, respectively; Fig.听2e).

Height-for-age difference

In addition to being longer than girls at all ages, boys also had a more negative height-for-age difference from the WHO MGRS median at all ages (Fig.听3; Supplemental Table 2). There was no significant interaction between age and sex, suggesting that the slopes of HAD with age do not differ between boys and girls. Finally, results from regression models testing the exposures associated with HAD showed few differences from the primary analysis of stunting (Supplementary Table 3). Like stunting, HAD is significantly associated with multiple maternal, infant, and household exposures. In contrast to the results for stunting, no factors were significantly moderated by sex.

Fig. 3

Sex differences in infant length for age z-score (LAZ) and length/height for age difference (HAD)

Sensitivity analysis

Results from sensitivity analysis with alternate outcomes were largely consistent with our primary results (Supplemental Table 4). Few differences were seen in the exposures for our alternate measures of growth faltering (severe stunting, continuous LAZ, and underweight) or in evidence for moderation by sex. Unlike the primary analysis with stunting, the interaction terms between sex and exclusive breastfeeding, perception of child appetite, and maternal marital status were not significant in models with severe stunting, continuous LAZ, or WAZ. Conversely, a larger effect was seen for the interaction between sex and the use of gripe water and severe stunting. As with our analysis of HAD, no significant interaction between age and sex was seen for continuous LAZ.

Discussion

Similarly to previous studies in Zambia [6, 39, 40] and other countries in Sub-Saharan Africa [7, 8], we found a high overall prevalence of stunting in children under age 2, with boys significantly more likely to be stunted than girls during the first two years of life. Our study extends previous research on sex differences in the prevalence of stunting by testing whether exposure to key environmental and behavioral risk factors for stunting differs for boys and girls, and whether boys and girls respond differentially to similar exposures. Although numerous factors were associated with the likelihood of stunting, exposure to most of these factors did not differ between boys and girls. Similarly, few differences were seen in the response to these exposures by sex, providing little evidence of differential vulnerability of boys and girls. Instead, examination of the length of boys and girls across the first two years of life indicate that boys, while being longer than girls at all ages, have more negative LAZ scores from birth continuing across the first two years. No sex differences were seen in the patterns of LAZ or HAD with age. This suggests that rather than being due to exposure to riskier environments or poorer feeding or care practices than girls in infancy and early childhood, the higher prevalence of stunting seen in boys may derive from either prenatal factors and/or comparison to the WHO MGRS growth standards.

We found a number of maternal, child, and household risk factors for stunting in our pooled sample of boys and girls, including a number that have been previously identified in Zambia [39,40,41,42,43,44,45] including maternal stunting, younger maternal age, low maternal education, low birth weight, poorer dietary diversity with fewer iron rich foods, lack of vitamin A supplementation, household food insecurity, poorer essential hygiene practices, lower wealth or rural residence. We also identified several additional behavioral and environmental factors associated with lower risk of stunting, including use of gripe water, safe food handling practices and household resilience, that, to our knowledge, have not previously been examined in Zambia.

Yet, despite the large number of risk factors for stunting that were identified in our analysis, relatively few sex differences were seen in exposure to these factors. Girls were more likely to be born at low birth weight than boys, while boys were more likely to have experienced diarrhea in the previous two weeks. These findings are similar to much other research showing that boys are born at higher birth weight, but also tend to have higher morbidity from common childhood illnesses like diarrhea and respiratory infections [20, 21]. Also similar to previous literature [13, 46,47,48], we found some sex differences in feeding and caring practices. In our sample, boys were more likely to receive solid foods or water before 6 months of age while girls were more likely to be exclusively breastfed. Previous research has documented sex differences in infant feeding practices; for example, research in the Philippines found that girls were exclusively breastfed for a longer length of time than boys [13], while, in Senegal, boys were more likely to have consumed solid foods before 6 months of age [10]. These differing feeding practices have been attributed to differences in the perceived needs of boys and girls [13], with boys being viewed as hungrier and fussier than girls who are seen as less demanding and in need of complementary foods [49].

As with these sex differences in feeding practices, we also saw differences in a few caregiving practices. Boys were more likely to receive vitamin A supplementation, a sex difference in routine health seeking behavior that has also been documented in many regions [12, 50, 51], and to be given gripe water, a practice that may be linked to perceived differences in infant temperament since gripe water is often given for the management of infant distress or colic [36, 52, 53]. While these differences in feeding and child care practices have the potential to contribute to sex differences in growth trajectories and stunting risk [10, 12, 22, 49], only the use of gripe water showed evidence for sex-specific risk for stunting in our stratified adjusted models, where it was associated with a lower risk of stunting in boys. Our findings suggest that, while differences in care and feeding practices may exist, they only contribute marginally to differences in the prevalence of stunting in this sample.

We also found limited evidence that boys and girls responded differently to exposure to risk factors for stunting. In our interaction models, only exclusive breastfeeding, maternal perception of child appetite, use of gripe water, education about exclusive breastfeeding, and maternal marital status showed some evidence of moderation by sex. Only the use of gripe water and exposure to exclusive breastfeeding education remained significant in models adjusting for other stunting risk factors, with boys who receive gripe water and girls whose mothers receive exclusive breastfeeding education having a lower risk of stunting than other boys or girls, respectively. While we do not have qualitative data on maternal perceptions of their children鈥檚 growth and/or needs, these findings may reflect differing perceptions of child needs by mothers and/or differing feeding practices in response to these perceptions [10, 12, 22, 49]. Exploratory work in Zambia has described cultural norms around weaning that differ for boys and girls, something seen in other SSA contexts [54].

At the household level, we found that the risk of stunting differed by maternal marital status for boys and girls. Girls whose mothers were monogamously partnered or who were widowed, separated or divorced have significantly lower risk of stunting than those with single mothers in the minimally adjusted models. This association was not seen in boys and may reflect differences in maternal autonomy and/income. Previous work has found that greater maternal education, a marker of greater maternal autonomy in childcare, is associated with a lower likelihood of stunting for girls, but not boys [8, 9, 55]. Previous work in SSA also documents that sex differences in stunting were moderated by socioeconomic status (SES) and maternal education. Sex differences in the prevalence of stunting were more pronounced in children in lower SES quintiles compared to those in higher quintiles, though the significance and direction of these differences varied across studies [8].

Overall, our results do not provide strong evidence that feeding practices, environmental exposures or child care practices differ consistently between boys and girls in Zambia or that boys and girls respond differently to these exposures. The sex differences we find are relatively small in magnitude, vary in direction, and are unlikely to account for the wide disparities in the prevalence of stunting seen between boys and girls. Our sensitivity analysis using length, LAZ, and HAD, shows similar patterns to our main analysis, supporting a relatively small effect of sex differences in exposures or vulnerability on infant and young child growth and nutritional outcomes. Further, we see no differences in the patterns of length, LAZ or HAD by sex across the first two years of life. Instead, our results more clearly suggest that the sex differences seen in stunting prevalence may stem from their smaller size of boys relative to the MGRS reference standards at birth and throughout the first two years of life.

The smaller size of boys than girls at birth relative to the WHO MGRS standards may stem from unmeasured prenatal or postnatal factors that compromise the growth of boys in utero and make them more vulnerable to even subtle differences in postnatal environmental insults. Previous research documents that boys tend to grow larger and more quickly beginning in utero and across infancy and early childhood [15, 56, 57]. The greater energy needed to support this growth may consequently limit boys鈥� ability to respond to undernutrition or other pre- or postnatal stressors [58], making them more sensitive to energetic or pathogenic exposures postnatally [11, 59]. Other non-assessed variables, such as micronutrient intake, could also contribute to differences in stunting through nutritional and infectious pathways. Along with their higher energy needs due to more rapid growth, boys may also have a higher demand for micronutrients and be at greater risk of deficiency [60]. For example, zinc requirements are around 10% higher for boys during infancy [61] and zinc deficiency is significantly associated with both infectious illness and poor linear growth [62, 63]. Further, differences in immune function, linked to differential inheritance of immune-related genes on the X chromosome and to the immune-modulating functions of sex steroids levels [19, 64], may place boys at greater risk of infection with consequences for their growth [2, 65,66,67]. Partnered with gender-based differences in weaning practices or routine medical care, this biological vulnerability may contribute to a greater risk of stunting for boys. However, we find relatively little evidence for this greater biological and/or social vulnerability in this sample. The patterns of growth of boys and girls are similar, there are few sex differences in exposures or response to adverse exposures, and the differences we do find do not all favor girls, as has been previously documented in an examination of care and feeding practices and child health outcomes in SSA [68].

Alternatively, the higher prevalence of stunting in Zambian boys may reflect a poorer fit of the MGRS standards for boys than girls in this sample. Boys tend to have a higher prevalence of stunting than girls when compared to the WHO MGRS despite their greater mean length/height values than girls [67]. In the WHO MGRS, standard deviations are smaller for boys than girls at all ages, so even at longer lengths boys are more likely to be defined as stunted. Coupled with the lower sexual dimorphism seen in length in our sample, these measurement issues may be leading to the higher estimates of stunting in boys despite the similar patterns of length and HAD across the first two years of life. Prior research has shown that sex differences in estimates of stunting, and undernutrition more generally, are sensitive to the reference standard used [23, 69,70,71]. Comparisons of the WHO MGRS, NCHS and British growth references in a large sample of Chinese infants, for example, documented higher LAZ scores in girls than boys from 6 to 24 months according to WHO growth reference, but lower scores in girls than boys compared to the British reference and mixed results compared to the NCHS [24]. Similarly, research among infants in Senegal also showed that sex differences in the prevalence of stunting were sensitive to the growth reference used and were higher with the WHO growth standard compared to the NCHS growth reference [10] due to the differing standard deviations for LAZ with age for boys and girls between the growth references.

Along with the reference used, limitations in measurement and analytic methods may also affect the interpretation of sex differences in the vulnerability to stunting. The use of cross-sectional analysis, as we employ, may lead to different interpretations of the relative growth of girls and boys [13, 22]. In cross-sectional analysis, girls in Bangladesh, for example, had higher average z-scores than boys from 6 to 23 months, despite longitudinal assessment showing that girls growth falters across the time period into childhood [22]. While we examined patterns of growth in length, LAZ and HAD monthly in our sensitivity analysis, we are not able to track the growth of individual children in our sample and it is not clear whether the cross-sectional patterns we see would accurately characterize the growth trajectories of individual children. Recent research has also documented that estimates of sex differences in stunting are sensitive to the age period studied with boys having lower LAZ in first 24 months, with differences reduced and sometimes reversed by 40 months of age [67]; thus, it is not clear that our findings of greater stunting in boys would persist past 2 years of age. Additional concerns have been raised that large-scale analysis, such as our comparison of sex differences for boys and girls across Zambia, obscures inequalities that may exist at the sub-national [14] or household-level [72]. To account for potential regional differences in risk factors, which may also include sociodemographic and/or cultural factors, we ran mixed models controlling for district, but finer grained analysis may be warranted in future research. Further, qualitative data around maternal perceptions of the growth and feeding needs of boys and girls would provide greater context for our findings. Finally, our results should be considered exploratory and interpreted with caution. We tested many potential risk factors in separate models for ease of interpretation; however, the large number of variable tested could have led to spurious associations. We provide Bonferroni correction for multiple testing in Table听2, which reduces the number of significant associations and none of the interaction terms were significant at this level of significance. Nevertheless, our findings are consistent across our alternate measures of growth (severe stunting, continuous LAZ, HAD, and WAZ) lending support for our fundings.

Along with these limitations, our study has several strengths. We were able to address gaps in the literature, by testing sex differences in exposure to a broad range of risk factors for stunting from the individual to the household levels. Our large sample size allowed us to test interactions between sex and exposure to these risk factors in the association with stunting and to conduct age-stratified analysis. Our sensitivity analysis allowed us to test our findings across multiple measures of growth addressing concerns that have been raised with using 鈭掆��2 LAZ scores as a cut-point in samples where most children may be exposed to growth faltering [25, 26]. Finally, our analysis identifies a few risk factors not regularly examined in studies of child stunting, such as the use of gripe water, consumption of iron rich foods, and household resilience, and may identify new avenues for intervention.

Conclusion

In sum, we find few differences in feeding practices, pathogenic exposures, or caregiving between boys and girls and find little evidence for differential vulnerability of boys and girls to these exposures. Our results suggest that Zambian boys begin life with a lower LAZ score than girls and continue to have more negative LAZ scores throughout infancy and a greater prevalence of stunting. These results may suggest that boys are more vulnerable to prenatal environments and that even the small 鈥渉its鈥� we observe have greater impacts on stunting risk. Alternatively, our analysis of LAZ, HAD, and length indicates that patterns of growth are similar for boys and girls across infancy and that boys鈥� greater prevalence of stunting may result from sex-based differences in the performance of WHO MGRS growth standard. In either case, our results show that stunting rates are high for both girls and boys and worsen across infancy, indicating that continued work is needed to improve child feeding practices, the child care environment, and household resources for all children.