Childhood stunting in relation to the pre- and postnatal environment during the first 2 years of life: The MAL-ED longitudinal birth cohort study

Abstract

Background: Stunting is the most prevalent manifestation of childhood malnutrition. To characterize factors that contribute to stunting in resource-poor settings, we studied a priori selected biological and social factors collected longitudinally in a cohort of newborns.

Methods and findings: We enrolled 1,868 children across 7 resource-poor settings in Bangladesh, Brazil, India, Nepal, Peru, South Africa, and Tanzania shortly after birth and followed them for 24 months between 2 November 2009 and 28 February 2014. We collected longitudinal anthropometry, sociodemographic factors, maternal-reported illnesses, and antibiotic use; child feeding practices; dietary intake starting at 9 months; and longitudinal blood, urine, and stool samples to investigate non-diarrheal enteropathogens, micronutrients, gut inflammation and permeability, and systemic inflammation. We categorized length-for-age Z-scores into 3 groups (not stunted, ≥-1; at risk, <-1 to -2; and stunted, <-2), and used multivariable ordinal logistic regression to model the cumulative odds of being in a lower length-for-age category (at risk or stunted). A total of 1,197 children with complete longitudinal data were available for analysis. The prevalence of having a length-for-age Z-score below -1 increased from 43% (range 37%-47% across sites) shortly after birth (mean 7.7 days post-delivery, range 0 to 17 days) to 74% (16%-96%) at 24 months. The prevalence of stunting increased 3-fold during this same time period. Factors that contributed to the odds of being in a lower length-for-age category at 24 months were lower enrollment weight-for-age (interquartile cumulative odds ratio = 1.82, 95% CI 1.49-2.23), shorter maternal height (2.38, 1.89-3.01), higher number of enteropathogens in non-diarrheal stools (1.36, 1.07-1.73), lower socioeconomic status (1.75, 1.20-2.55), and lower percent of energy from protein (1.39, 1.13-1.72). Site-specific analyses suggest that reported associations were similar across settings. While loss to follow-up and missing data are inevitable, some study sites had greater loss to follow-up and more missing data than others, which may limit the generalizability of the findings.

Conclusions: Neonatal and maternal factors were early determinants of lower length-for-age, and their contribution remained important throughout the first 24 months of life, whereas the average number of enteropathogens in non-diarrheal stools, socioeconomic status, and dietary intake became increasingly important contributors by 24 months relative to neonatal and maternal factors.

Fig 1. Modified version of the UNICEF malnutrition conceptual hierarchical framework and the maternal and household factors and childhood environmental exposures included in our analyses.

AAT, alpha-1 antitrypsin; AGP, alpha-1-acid glycoprotein; ALRI, acute lower respiratory infection; MPO, myeloperoxidase; NEO, neopterin.

Fig 2. Length-for-age stratified by age and site.

Site-specific length-for-age trajectories with age were smoothed using a smoothing spline. Sites include BGD, Dhaka, Bangladesh; BRF, Fortaleza, Brazil; INV, Vellore, India; NEB, Bhaktapur, Nepal; PEL, Loreto, Peru; SAV, Venda, South Africa; TZH, Haydom, Tanzania.

Fig 3. Categories of length-for-age stratified by exact month of age and site.

In this figure, not stunted (length-for-age Z-score [LAZ] ≥ −1) is represented in green, at risk of being stunted (−2 ≤ LAZ < −1) is represented in yellow, and stunted (LAZ < −2) is represented in orange. Sites include BGD, Dhaka, Bangladesh; BRF, Fortaleza, Brazil; INV, Vellore, India; NEB, Bhaktapur, Nepal; PEL, Loreto, Peru; SAV, Venda, South Africa; TZH, Haydom, Tanzania. The vertical broken line represents 6 months of age.

Fig 4. Interquartile cumulative odds ratios of being in a lower length-for-age category (at risk or stunted) at enrollment, 12 months, and 24 months for 5 risk factors, as obtained from the multivariable ordinal logistic regression model.

In (A), we show adjusted interquartile cumulative odds ratios and corresponding 95% CIs. The interquartile cumulative odds ratio is calculated for the 75% and 25% percentiles of the risk factor. In (B), we show site-specific estimates represented by triangles.

Fig 5. Site-specific estimates of the interquartile cumulative odds ratios of being in a lower length-for-age category (at risk or stunted) at enrollment, 12 months, and 24 months for 5 risk factors, as obtained from the multivariable ordinal logistic regression model.

We contrast site-specific adjusted interquartile cumulative odds ratios and corresponding 95% CIs to the overall population interquartile cumulative odds ratios and 95% CIs. Overall interquartile cumulative odds ratios and corresponding 95% CIs at enrollment, 12 months, and 24 months are represented by black lines and grey (enrollment), blue (12 months), and red (24 months) shading. Site-specific interquartile cumulative odds ratios and corresponding 95% CIs are represented by triangles and horizontal bars. Overlap between the 95% CIs of site-specific and overall population interquartile cumulative odds ratios suggests that the associations were relatively similar across settings. BGD, Dhaka, Bangladesh; BRF, Fortaleza, Brazil; INV, Vellore, India; LAZ, length-for-age Z-score; NEB, Bhaktapur, Nepal; PEL, Loreto, Peru; SAV, Venda, South Africa; TZH, Haydom, Tanzania.

Fig 6. Age-specific cumulative odds ratios of being in a lower length-for-age category (at risk or stunted) for gut inflammation and permeability biomarkers.

Stool AAT concentration (A), lactulose:mannitol Z-score (B), and AGP concentration (C). In all panels, we show the interquartile cumulative odds ratio (75th percentile versus 25th percentile) for each risk factor as a function of age. The black line represents the mean estimate, and the grey shading represents the 95% pointwise confidence interval. AAT, alpha-1 antitrypsin; AGP, alpha-1-acid glycoprotein; OR, odds ratio.