High double burden of child stunting and maternal overweight in the Republic of the Marshall Islands

Abstract Many low‐ and middle‐income countries are faced with a double burden of malnutrition characterized by a stagnating burden of undernutrition and an increasing prevalence of overweight and obesity often observed both at population and household levels. We used data from the 2017 National Integrated Child Health and Nutrition Survey in the Republic of the Marshall Islands to explore the prevalence of overweight mother‐stunted child pairs (mother–child double burden, MCDB). We used bivariate analysis, multivariate logistic regression, and multinomial logistic regression analysis to explore associations between child‐, maternal‐, and household‐level variables and both stunting and MCDB and other types of maternal–child pairs. Our results indicate that nearly three out of four mothers were overweight or obese and one in four households is home to an overweight mother with a stunted child. The risk of child stunting and of MCDB were largely associated with maternal characteristics of lower maternal height, maternal age at birth, years of education, and marital status and household economic status as measured by wealth index and number of household members. These findings support the growing body of evidence showing that the coexistence of high maternal overweight and child stunting (MCDB) has linked root causes to early life undernutrition that are exacerbated by the nutrition transition.

This analysis examines the prevalence of MCDB in RMI and explores associated child, maternal, and household variables that influence the risk of child stunting alone and of MCDB compared with other types of maternal-child pairs. As urbanization and economic growth continue to progress in RMI, identification of households at risk for MCDB and an understanding of the influencing variables will guide policy makers to best address the growing divide of malnutrition in the country. Briefly, the sample was based on the 2011 RMI census with 2017 projections used as the sampling from which a two-stage, stratified cluster sampling approach was used to for the selection of the survey sample. A total of 50 clusters (communities) stratified for urban and rural areas were selected using probability proportion to size on all islands in RMI with a population over 100 people. In each selected cluster, a household listing was carried out to identify household with children under 5 years of age and 12 households in each of the 50 clusters were randomly sampled as the secondary sampling unit for a total of 600 children. Several clusters had less than 12 households with children under 5 years of age necessitating a reduction in the total sample size to 581 households. Household eligibility was based on a child 0-59 months of age living in the household.

| Key variables
Child, maternal, and household characteristics included in the analysis were selected based on the UNICEF conceptual framework for the determinants of child undernutrition (UNICEF, 2013) and on previous examinations on drivers of MCDB (Oddo et al. 2012, Kosaka & Umezaki, 2017 bs_bs_banner Discordance in nutrition status of mother-child pairs is a more extreme form of the double burden of malnutrition. It is commonly assumed that mothers and their children should be less likely to differ in their nutrition status as they have shared access to food, sanitation, hygiene, and other resources. However, in low-and middle-income countries, the nutrition transition, along with increasingly sedentary lifestyles, has changed eating and activity patterns in both children and their mothers with increased consumption of energy dense and nutrient poor foods for all members of the household. These foods contribute to increased overweight in adults and fail to help young children meet their linear growth potential (Tzioumis & Adair, 2014).
The mother-child double burden (MCDB) of malnutrition is an established concern globally. In a systematic review of the prevalence of the mother-child double burden in 50 countries, the prevalence was less than 10% for nearly all countries with the exception of Guatemala (20%), Egypt (16%), and Bolivia (11.5%; Kosaka & Umezaki, 2017).
MCDB has linked root causes to early life undernutrition that are further exacerbated by the nutrition transition (Delisle, 2008;Popkin, Adair, & Ng, 2012;Gillespie & Haddad, 2001). Increased risk of MCDB is associated with increasing urbanization (Food and Agriculture Organization [FAO], 2006), socio-economic status (Doak CM., 2005;Jehn M., 2009;Lee J., 2010), household food insecurity (Gubert MB., 2016;Ghattas, 2014), and poor dietary quality (Delisle, 2008). Maternal short stature, indicative of early undernutrition of the mother, is strongly associated to increased risk of child stunting (Garrett J.L., 2005), maternal overweight (Sichierei R., 2003;Bosy Westphal A., 2009), and of MCDB (Ramirex-Zea M., 2014; Oddo VM., 2012). There are several proposed mechanisms explaining the link between early childhood stunting and later overweight and obesity in adulthood. Children stunted during the first 1,000 days are likely to undergo metabolic and physiological changes in response to undernourishment, including impaired fat metabolism (Frisancho 2003, Friedman SM 2006, accumulation of excess body fat (Sawaya, 2003;Sawaya, Grillo, Verreschi, Carlos da Silva, & Roberts, 1998), higher energy intake for body weight , lower resting metabolic rate (Grillo LP. 2005), and lower energy , Friedman SM 2006. In obesogenic environments, these metabolic and physiological adaptations for energy and fat conservation can lead to overweight and increase susceptibility to develop noncommunicable diseases (NCDs) later in childhood and into adolescence and adulthood Sawaya, 2003;Sawaya et al., 1998 classified as stunted when their height-for-age Z-score < −2 SD and mothers with a BMI ≥ 25 kg/m 2 or BMI ≥ 30 kg/m 2 were classified as overweight and obese respectively. MCDB was defined by four categories: (a) stunted child-overweight mother pairs (corresponding to MCDB pairs), (b) stunted child-nonoverweight mother pairs, (c) nonstunted child-overweight mother pairs, and (d) nonstunted child-nonoverweight mother pairs.
Potential covariates for child undernutrition were examined at the child, maternal, and household levels. Variables collected for children included child sex, child age (recoded as a dummy variable to allow for nonlinear associations), incidence of diarrhoea in the previous 2 weeks, and being left in inadequate care in the previous week.
Mothers' characteristics included current height as a proxy for prior stunting experience, marital status, level of schooling, and age at child's birth. Maternal height was defined by three categories: maternal height <150 cm, ≥150 cm and ≤159.9 cm and ≥160 cm (Ozaltin, Hill, & Subramanian, 2010

| Statistical methods
Bivariate analyses were conducted to determine the association between known determinants of child undernutrition (UNICEF, 2013) with stunting and MCDB. For MCDB, multinomial logistic regression was conducted to determine the associated risk of MCDB with influencing variables from the child, maternal, and household levels compared with other types of maternal-child nutrition status pairs. Results are expressed as relative risk ratios (RR) with 95% confidence intervals. For child stunting, binomial logistic regression was conducted to determine the associated odds of stunting with predicting variables from the child, maternal, and household levels.
Results are expressed as adjusted odds ratios ( removed from the analysis as they did not live with their mother. There were no significant differences in the prevalence of stunting, child sex, area, or wealth index quintile between these children, and 729 children (464 households) were included in the analysis ( More than one-third (37.1%) of children were stunted with the highest prevalence of stunting (60.0%) among children 18-23 months of age (Table 2). The prevalence of wasting and overweight among children was low (4.0% and 3.3% respectively). Among maternal-child pairs, 25.2% were MCDB with an overweight or obese mother and a stunted child (Table 3).

| Influencing factors for stunting
Variables associated with increased odds of stunting in children include older age of the child, larger number of household members, lower household wealth index, shorter maternal height, and unmarried or cohabitating maternal status (Table 4). Compared with mothers taller than 160 cm, mothers who were shorter than 150 cm had a 6.6 greater odds of having a stunted child (AOR = 6.62; 95% CI = 2.50, 17.51), and mothers whose height was 150-159 cm tall had a 4.5 greater likelihood of having a stunted child (AOR = 4.55; 95% CI = 1.66, 12.50). Mothers who were never married or cohabitating were more likely to have a stunted child compared with married mothers (AOR = 1.72; 95% CI = 1.07, 2.77). Children from lower wealth quintiles were more likely to be stunted compared with children from the wealthiest quintile; the highest odds of stunting were observed among children from the poorest wealth quintile (AOR = 3.13; 95% CI = 1.54, 6.37).

| DISCUSSION
The high double burden of malnutrition in RMI is found not only at the population level but at the household level as well. Our analysis indicates that one in four households in RMI is home to an overweight mother with a stunted child. Stunting in children surpasses the WHO threshold for very high prevalence (30%), peaking during the period of complementary feeding from 6 to 23 months of age. Our analysis indicates that nearly three out of four mothers were overweight or obese, and nearly all mothers (93.7%) were shorter than 160 cm. Wealth index quintiles were developed using principal component analysis summarizing housing variables (e.g., roof, walls, and floor) and asset variables (e.g., television, radio, and car). The prevalence of child overweight (3.3%) and wasting (4.0%), however, was low, and few mothers (2%) were underweight. As the prevalence of maternal overweight and obesity in the survey was very high, MCDB pairs were markedly different from households with a stunted child/nonoverweight mother pair or nonstunted child/nonoverweight mother pairs. Compared with pairs with nonoverweight mothers, increased risk of MCDB was associated with older maternal age, attainment of higher education, and compared with stunted child/nonoverweight mother pairs, lower parity. These factors may reflect the risk of maternal overweight rather than child stunting with older, higher education mothers more likely to be overweight regardless of having a stunted child. In comparison to nonstunted/overweight mother pairs, however, MCDB pairs are significantly poorer.
Child stunting, regardless of the nutrition status of the mother, is influenced by shorter maternal height and indicators of poverty in RMI. Risk of MCDB, where the child is stunted and the mother is overweight, increases with maternal age and higher education and remains associated with poorer households. Although not measured in this study, the high prevalence of MCDB in RMI is likely linked with the nutrition transition in the country. Previous research has shown that low consumption of nutrient-rich foods and high consumption of nutrient-poor foods by women and children are strongly associated with both an increased risk of stunting in children (Arimond & Ruel, 2004;Ihab et al., 2014;Rah et al., 2010;Sawadogo et al., 2006) and a higher risk of overweight and obesity in their mothers (Aitsi-Selmi, 2015).
Globally, the coexistence of maternal overweight and child stunting (MCDB) is linked to early life undernutrition exacerbated by the nutrition transition (Delisle, 2008;Popkin, Adair, & Ng, 2012;Gillespie & Haddad, 2001). In a context of nutrition transition, children stunted during the first 1,000 days are likely to undergo metabolic and physiological changes in response to undernourishment that increase their susceptibility to becoming overweight and developing related NCDs later in childhood, adolescence, and adulthood Sawaya, 2003;Sawaya et al., 1998). As children who are stunted after 2 years of age have missed the critical window for optimal growth, the high burden of early stunting in RMI has broader implications for the health and economic capacity of the entire population.
There are a few limitations to our analysis. We investigated the effects of various household, maternal, and child factors on child stunting and MCDB using a cross-sectional study design, which prevents the inference of causal relationships. Additionally, an insufficient sample size precluded a representative analysis of the effects of low birthweight and infant and young child feeding practices on the risk of stunting as these indicators were collected in subsamples of the total population. Further research should determine the effects of these important indicators on child stunting in the RMI context.  Abbreviations: AOR, adjusted odds ratio; BMI, body mass index; MCDB, mother-child double burden; RR, risk ratios. a Multivariate logistic regression performed with baseline comparison group as children who were not stunted. Multinomial logistic regression performed with MCDB pairs and baseline comparison groups of (a) nonstunted child nonoverweight mother, (b) stunted child non-overweight mother, and (c) nonstunted child overweight mother. Results are presented as AOR with 95% CI. All estimations are adjusted for two-stage cluster randomized survey design and nonresponse.
b Stunted defined as height-for-age Z-score (HAZ) < −2 SD below median. c Non-stunted child and a nonoverweight mother pair defined as child having a height-for-age Z-score (HAZ) ≥ −2 SD above median and a mother with a BMI <25.
d Stunted child and a nonoverweight mother pair defined as child having a height-for-age Z-score (HAZ) < −2 SD below median and a mother with a BMI <25. e Non-stunted child and an overweight mother pair defined as child having a height-for-age Z-score (HAZ) ≥ −2 SD above median and a mother with a BMI >25.

| CONCLUSION
The public health action in RMI has focused on the prevention and treatment of overweight, obesity, and NCDs in adolescents and adults.
However, the high prevalence of child stunting and MCDB found in our analysis represents a major public health challenge for the country.
Preventing childhood stunting is vital for ensuring children's physical and cognitive development, preventing future overweight, NCDs, and associated consequences in later life, and for achieving greater social and economic gains for the country. Universal and targeted measures are recommended to address the driving factors to child stunting and maternal overweight, which include addressing poor maternal nutrition during pregnancy and increasing access, affordability, and demand to nutrient-rich foods for young children and their mothers, especially among the most vulnerable social groups. The findings of high double burden of malnutrition in RMI 2017 have brought about high-level commitment by the nation's president to mobilize resources for promoting services and support to mothers and children during the 1,000 days from conception to age 2 years.
Continued high-level advocacy with integrated and synergistic action across programmes is needed to reduce the double burden of stunting and overweight among children and women in RMI.

ACKNOWLEDGMENTS
None declared.

CONFLICTS OF INTEREST
The authors declare that they have no conflicts of interest.

CONTRIBUTIONS
The authors' responsibilities were as follows-JLB and SG designed the study; JLB acquired and analysed data; JLB, SG, UP, and WE interpreted the data; JLB drafted this manuscript; and JB, SG, UP, JA, FdB, and WE provided critical intellectual feedback to help revise the manuscript. All authors have read and approved the final manuscript.