Causes and consequences of child growth faltering in low-resource settings

Growth faltering in the form of stunting (chronic malnutrition) and wasting (acute malnutrition) is common among young children in low-resource settings and contributes to child mortality and adverse adult outcomes. Despite recognition of its public health importance, preventive interventions in low- and middle-income countries (LMICs) have had limited success at scale. Nutritional interventions across fetal and early childhood periods show benefits but often small effects on population-level stunting and wasting, and WASH interventions have not improved growth in several large trials. This suggests a gap in understanding optimal timing and targets for interventions. The study aims to identify principal causes, age windows, and subgroups for preventive focus by pooling longitudinal cohort data across LMICs to estimate relationships between child, parental, and household characteristics and growth outcomes from birth to 24 months, and to assess how early growth faltering relates to later severe faltering and mortality.

Prior evidence indicates high global burdens of stunting (22%) and wasting (7%) among children under 5, concentrated in LMICs, with substantial mortality attributable to both. Nutritional interventions (education; macro- and micronutrient supplementation during pregnancy and complementary feeding; breastfeeding promotion) yield beneficial but often modest improvements in growth. Large randomized trials of WASH interventions in Bangladesh, Kenya, and Zimbabwe found no effects on child growth. Meta-analyses show small but significant impacts of complementary feeding and mixed results for breastfeeding promotion on growth. Observational and experimental studies have explored links between infections (e.g., diarrhea), environmental enteropathy, and growth with varying conclusions. Overall, literature underscores the need to better pinpoint modifiable determinants, critical windows, and synergistic intervention strategies, potentially beginning before conception and during pregnancy.

The study conducted pooled longitudinal analyses of 33 cohorts from 15 LMICs (south Asia, sub-Saharan Africa, Latin America, and one European cohort) initiated between 1987 and 2014. Inclusion criteria: (1) LMIC setting; (2) children enrolled between birth and 24 months with repeated length and weight measurements; (3) not restricted to acutely ill children; (4) median birth year after 1990; (5) anthropometry measured at least quarterly. This yielded 83,671 children and 592,030 measurements; for mortality analyses, criteria were relaxed to biannual measurements, adding 4 cohorts for a total of 97,988 children and 662,763 observations. Exposures (30 total) spanned parental/household characteristics (prenatal), at-birth factors, and postnatal variables measured across cohorts and harmonized for pooled analysis. Population Intervention Effects (PIEs) were estimated: the change in population mean z-score (LAZ, WLZ, WAZ) if exposures were shifted from observed levels to the lowest-risk reference level. Targeted maximum-likelihood estimation (TMLE), a doubly robust semi-parametric approach with ensemble machine learning, adjusted for measured confounders (excluding variables on the causal pathway, e.g., not adjusting maternal height-stunting association for birthweight). Cohort-specific parameters were estimated and pooled via random-effects models. Age-varying analyses stratified associations by onset age (0–6 vs 6–24 months). Additional models examined: population-attributable fractions for stunting/wasting; growth trajectories stratified by maternal height/BMI; WLZ quartiles predicting subsequent linear growth velocity; associations between early faltering (0–6 months) and later severe outcomes (persistent wasting 6–24 months; concurrent wasting and stunting at 18 months); and hazard ratios of all-cause mortality by 24 months associated with growth faltering phenotypes in 8 cohorts with death data.

• Shifting population-level prenatal and at-birth factors—particularly maternal anthropometry and child condition at birth—would yield substantial gains by 24 months: LAZ improvements up to +0.40 z and WLZ up to +0.15 z.
• Socioeconomic status indicators (parental education, household wealth, more rooms, clean cooking fuel) predicted higher z-scores; seasonality strongly influenced WLZ (higher in drier periods).
• Exclusive/predominant breastfeeding before 6 months associated with higher WLZ (but not LAZ) at 6 months and was not a major predictor at 24 months.
• Girls consistently had higher LAZ and WLZ than boys across regions.
• Cross-validated R^2 for models with top-10 determinants plus sex: 0.25 for LAZ (n=20 cohorts; 25,647 children) and 0.07 for WLZ (n=18 cohorts; 17,853 children).
• Population-attributable fraction patterns for stunting and wasting mirrored those for continuous z-scores. Example: If all mothers had BMI ≥20, wasting incidence by 24 months would reduce by 8.2% (95% CI 4.4, 12.0).
• Preterm birth had stronger associations with stunting than wasting; rainy season associated with wasting but not stunting.
• Maternal height substantially affected LAZ at birth; postnatal LAZ trajectories were parallel regardless of maternal height/BMI. WLZ trajectories diverged after 3–4 months by maternal BMI/height.
• Age-varying effects: SES factors more strongly associated with incident wasting/stunting after 6 months; higher birth order reduced risk <6 months but increased risk thereafter. First-borns had lower WLZ at birth with rapid postnatal catch-up.
• Most exposures had stronger associations with severe phenotypes (severe stunting/wasting; persistent wasting) than with any stunting/wasting.
• Higher WLZ predicted faster subsequent linear growth velocity (exposure–response across WLZ quartiles).
• Early faltering (0–6 months), especially persistent wasting, strongly predicted later severe faltering (persistent wasting 6–24 months; concurrent wasting and stunting at 18 months).
• Mortality: Growth faltering increased hazard of death by 24 months in 8 cohorts (1,689 deaths; 2.4%). Strongest associations for severe wasting plus stunting (HR 8.7; 95% CI 4.7–16.4) and severe underweight alone (HR 4.2; 95% CI 2.0–8.6).

Findings highlight prenatal, maternal, and at-birth determinants as dominant drivers of growth trajectories in the first 1,000 days, with postnatal insults affecting children similarly regardless of initial status. The limited effects of breastfeeding promotion on LAZ and of reducing diarrhea on LAZ in these analyses suggest that, while beneficial for health, such interventions alone are unlikely to substantially shift population-level linear growth without concurrent broader improvements in living standards and maternal nutrition. Non-modifiable factors (sex, birth order, season) can inform targeting, such as seasonal supplementation and enhanced monitoring for boys. Strong maternal anthropometry–child growth links underscore intergenerational transmission of growth deficits. Age-stratified results indicate that post-weaning environments and cumulative socioeconomic constraints increasingly shape faltering risk, supporting age-tailored strategies. Early ponderal deficits predispose to later linear deficits and to severe, persistent faltering, which carry elevated mortality risks—emphasizing the need for prevention and rapid recovery in early infancy. Overall, policies and programs focusing on preconception and pregnancy, coupled with socioeconomically sensitive, seasonally aware support in infancy and early childhood, are most likely to yield meaningful improvements.

By harmonizing 33 longitudinal cohorts with advanced causal estimation, this study identifies prenatal maternal status and at-birth characteristics as prime levers to reduce child growth faltering, quantifying potential gains in LAZ and WLZ from realistic population exposure shifts. Early faltering substantially elevates risks of persistent/severe faltering and mortality. The work supports prioritizing preconception and pregnancy interventions (maternal nutrition, health care, infection prevention), alongside context-specific socioeconomic improvements and age-targeted child support to prevent and mitigate faltering. Future research should test comprehensive maternal interventions beginning preconception, evaluate seasonally targeted strategies, identify effective packages that combine nutrition with social and environmental improvements, and refine causal mediation pathways from maternal anthropometry to child growth to inform precision targeting.

Mortality analyses were limited to 8 highly monitored cohorts with lower-than-population mortality; deaths were included only if occurring after anthropometry assessments, potentially excluding neonatal deaths, and lacked cause-specific data. Exposure measurement varied across cohorts and harmonization may have residual inconsistencies. Despite TMLE and extensive adjustment, residual confounding and unmeasured variables remain possible in observational cohort data. Some key predictors (e.g., maternal anthropometry) may proxy for broader, unmeasured socioeconomic, genetic, or environmental factors; mediation analyses suggest only partial mediation by birth characteristics. Generalizability across regions is strong in direction but magnitudes vary (e.g., sex differences in South Asia).