Early childhood linear growth faltering in low‐ and middle‐income countries is a whole‐population condition: implications for identifying the causes of stunting

Innumerable epidemiological studies have been conducted to explain the causes of between‐child variation in height‐for‐age z‐scores (HAZ) and the risk of early childhood stunting (HAZ<‐2) in low‐ and middle‐income countries (LMICs). Many studies implicitly assume that between‐child variations in exposure to growth constraining risk factors can explain the overall population‐average decline in HAZ relative to the international norm (World Health Organization child growth standard) that is commonly observed in the first two years of childhood in LMICs. We aimed to demonstrate that postnatal linear growth faltering in LMICs is a whole‐population condition and therefore that sources of between‐child variation in HAZ within populations are unlikely to explain faltering relative to the WHO standard or the high prevalence of stunting in many LMICs. We analyzed the distributions of HAZ (based on World Health Organization child growth standards) in discrete 3‐month age intervals in 168 Demographic and Health Surveys (DHS) from 64 countries (1993 to 2014). For each age band within each survey (average ~275 children per group), we estimated the mean HAZ and the standard deviation (SD), 5 th and 95 th percentiles of the HAZ distribution. Pooling across surveys, we plotted the change in these HAZ distributional properties as a function of age. Using the survey‐specific age groups as the basic unit of analysis, and accounting for clustering of age groups within surveys and surveys within countries, we used multi‐level linear regression models to estimate the associations between mean HAZ (predictor) and the SD, 5 th , and 95 th percentiles of the HAZ distribution. Simulation exercises showed that in order for between‐child variation in stunting risk factors to cause the population‐average decline in mean HAZ with age, the SD of the HAZ distribution must increase with age. From birth to 3 years of age, the age‐related decline in mean HAZ was accompanied by declines in both the 5 th and 95 th percentiles of the HAZ distribution and a narrowing of the distribution represented by a decline in the mean SD ( Figure). For each 1‐unit decline in mean HAZ: SD decreased by 0.12 units (95% CI: −0.14, −0.11), the distance between the mean HAZ and 95 th percentile HAZ changed by −0.16 (−0.19, −0.12), and the distance between the mean HAZ and 5 th percentile HAZ changed by −0.26 (−0.29, −0.22). Overall inferences were unchanged in sub‐analyses of surveys grouped by sample size, year, world region, mean HAZ in the 0–3 month group, or national income. In LMICs where mean HAZ declines in the postnatal period, children across the entire HAZ spectrum experienced slower growth versus the WHO standard. The HAZ distribution narrowed as mean HAZ declined due to convergence of both tails towards the mean. This pattern is inconsistent with the notion that certain vulnerable children within each country are responsible for the population‐average decline in HAZ in the first 2–3 years. Therefore, we conclude that postnatal linear growth faltering in LMICs is a whole‐population condition. Individual‐level risk factors for low HAZ may not explain the phenomenon of population faltering or the high prevalence of stunting. The epidemiologic hunt for causes of stunting in LMICs should shift to a focus on sources of between‐population rather than between‐individual variation in HAZ. Support or Funding Information Bill and Melinda Gates FoundationHeight‐for‐age z‐score (HAZ) distribution parameters (mean, median, standard deviation, 5 th percentile, 95 th percentile) by 3‐month age band, averaged across 168 Demographic and Health Surveys from 1993 to 2014.