Early-childhood linear growth faltering in low- and middle-income countries

Stunting, defined as length-for-age z-score (LAZ) more than 2 standard deviations below the WHO growth standard, affects an estimated 149 million children under 5 (22% globally), with the heaviest burden in South Asia and Africa. Early-life stunting is linked to increased mortality and morbidity and to long-term deficits in cognition and productivity, with large macroeconomic implications. The first 1,000 days, including the prenatal period, are considered a critical window; intrauterine growth restriction and preterm birth are strongly associated with stunting at 24 months. Although some catch-up growth can occur later, its extent depends on the timing and severity of early faltering. Most global estimates rely on cross-sectional surveys (e.g., DHS), which cannot reveal longitudinal patterns, timing of onset, or persistence and may be biased by survivor effects. Few studies have estimated age-specific incidence in the first 2 years. This study aimed to determine the timing of onset, persistence, reversal, and relapse of linear growth faltering, and to quantify linear growth velocity in early life, by pooling longitudinal cohorts in LMICs and focusing on ages 0–24 months.

Prior work using cross-sectional DHS data documented that stunting prevalence increases with age and suggested a whole-population pattern of linear growth deficits, with much of faltering occurring by age 2 and substantial length deficits accrued by age 5. However, cross-sectional analyses cannot identify individual-level onset, persistence, reversal, or relapse dynamics and may omit the most vulnerable children due to survivor bias. Evidence for catch-up growth exists but varies by timing and severity of early faltering. Few longitudinal studies have reported age-specific incidence within the first 2 years. This study responds to these gaps by leveraging multiple longitudinal cohorts to estimate incidence, reversal, relapse, and growth velocity across infancy and toddlerhood, and by comparing cohort growth patterns with contemporary DHS distributions to assess representativeness.

Design and data sources: Pooled analysis of 32 longitudinal cohorts from 14 LMICs (South Asia, sub-Saharan Africa, Latin America, and Eastern Europe) followed between 1987 and 2017, aggregated by the Bill & Melinda Gates Foundation Knowledge Integration (Ki) initiative. Inclusion criteria: LMIC setting; median birth year ≥1990; enrollment between birth and 24 months; repeated length and weight measurements at least every 3 months; no restriction to acutely ill children. Sample: 52,640 children contributing 413,317 measurements; 859 (0.2%) measurements with implausible LAZ (>6 or <−6) were excluded, leaving 412,458 measurements. Measurement frequency: 21 cohorts measured at least monthly; 11 measured quarterly. Cohort sizes ranged from 119 to 14,074 children, with >80% follow-up at scheduled ages in most cohorts. Outcomes and definitions: LAZ computed using WHO 2006 growth standards. Stunting defined as LAZ < −2; severe stunting as LAZ < −3. Incident stunting onset classified within 3-month age bands as the first time a child’s LAZ dropped below −2 in that interval. Stunting reversal defined as LAZ rising from below −2 to above −2; stunting relapse as returning to LAZ < −2 after reversal. Birth LAZ defined as the first LAZ measure before 30 days of age. Linear growth velocity defined as change in recumbent length (cm per month) between consecutive measurements; within-child LAZ change per month was also estimated. Age focus for reversal/relapse analyses was 0–15 months (monthly-measured cohorts), due to sparser measurements thereafter. Statistical analysis: Cohort-specific estimates were pooled using random-effects models with restricted maximum-likelihood (REML) estimation; alternative pooling (fixed-effects or ML) was used if REML did not converge. Age trends in mean LAZ were modeled using cubic splines where appropriate. Heterogeneity was summarized with I statistics and interquartile ranges. Representativeness was assessed by comparing Ki cohort LAZ distributions and mean LAZ-by-age to contemporaneous DHS data from overlapping countries using kernel density plots and spline-smoothed means with simultaneous confidence intervals. Sensitivity analyses included restriction to cohorts with monthly measurements from birth to 24 months and adjusting for gestational age in cohorts where available. Geographic coverage: 17 cohorts in 4 South Asian countries; 7 in 6 African countries; 7 in 3 Latin American countries; and 1 in Eastern Europe (Belarus).

- Linear growth faltering begins very early. Across cohorts, 13% of children were stunted at birth (range 0.3%–42% by cohort). Incident stunting onset between birth and 3 months affected 18% overall (range 7%–57%). Children stunted between birth and 3 months accounted for 23% of all children who ever became stunted by 24 months (69% ever stunted by 24 months). - Regional patterns: Very early stunting was most common in South Asia: 20% stunted at birth and an additional 21% became stunted by 3 months. South Asian cohorts had lower mean LAZ across ages compared with Africa and Latin America. - Whole-population phenomenon: In about half of cohorts, the 95th percentile of LAZ fell below 0 by 15 months; LAZ distributions shifted downward with age while dispersion and skewness remained similar, indicating population-wide growth faltering, not only among those classified as stunted. - Reversal and relapse dynamics: From 0 to 15 months, stunting reversal was uncommon and often not sustained. By 15 months, 34.0% were stunted, 50.5% had ever been stunted, and 16.5% had reversed and were not currently stunted. Relapse after reversal occurred at 2.0%–3.5% per month between 6 and 15 months. - Birth size matters: 86% of children who ever became stunted had LAZ < 0 at birth. Relapse percentages increased with age and were higher among children stunted at birth. Reversal was more frequent at very young ages among those born with LAZ < −2 (consistent with regression to the mean) but remained low after 6 months (<7% per month if birth LAZ < −2; <5% per month if birth LAZ between −2 and 0). - Post-reversal trajectories: Among children who reversed before 6 months, mean LAZ 9 months later declined by −0.69 (95% CI −0.84, −0.55), with larger subsequent declines when reversal occurred at older ages. Improvements after reversal were generally insufficient to eliminate deficits and did not resemble sustained biological recovery. - Growth velocity: From 0–3 months, length velocity varied widely (boys: below WHO 1st to above 50th percentile; girls: up to above 75th percentile). Thereafter, most cohorts’ velocities were between the 15th and 50th WHO percentiles. From 3–24 months, average linear growth ranged from ~0.75 to 1.25 cm per month. - Macro-context: Earlier onset (birth to 3 months) was more prevalent in countries with lower health spending, higher under-5 mortality, and higher poverty. Ki cohort mean LAZ was somewhat lower than DHS in many settings (especially South Asia), but overall distributions were similar, supporting external validity.

The study answers when linear growth faltering begins and how persistent it is in early life by leveraging longitudinal data. Findings show that incidence peaks from birth to 3 months across regions, with especially high at-birth stunting in South Asia. Early stunting strongly shapes subsequent trajectories: reversal is uncommon and often followed by relapse, particularly among those stunted at birth. Even children never classified as stunted experienced declining mean LAZ, underscoring growth faltering as a whole-population issue. These results reinforce and quantify the need for a life course approach focused on maternal and preconception health, prenatal care, and the earliest postnatal period. They further suggest that broader investments in health systems and addressing upstream determinants (women’s nutrition, adolescent pregnancy, family planning, and prevention of prenatal infections) could reduce stunting at birth. The paucity of effective interventions for children under 6 months and modest effects during complementary feeding (6–24 months) call for research to identify new, potentially multisectoral strategies, including nutritional support for lactating parents and addressing environmental and microbiota-related factors. Targeting by age and geography, particularly in South Asia, may yield greater impact than one-size-fits-all approaches.

Linear growth faltering in LMICs commonly begins prenatally or within the first 3 months of life, is infrequently reversed, and often recurs after reversal. Defining stunting targets at earlier ages (for example, by 3 or 6 months) could focus programs on the most impactful window. A life course strategy that strengthens women’s health before and during pregnancy and includes interventions in the first months of life is needed, alongside renewed efforts to develop and deploy effective postnatal interventions and to tailor approaches by regional context (notably South Asia).

- Measurement error in recumbent length could affect threshold-based outcomes (reversal, relapse), though quality assessments indicated high measurement quality across cohorts. - LAZ at birth based on WHO child growth standards can overestimate stunting among preterm infants. Gestational age was unavailable or limited (mostly by last menstrual period or newborn exam; one cohort by ultrasound). In sensitivity analyses adjusting for gestational age where available, stunting at birth was ~1% lower. - Cohort coverage was not exhaustive of all countries in each region, and included African and South Asian cohorts tended to have more growth faltering than contemporary representative surveys; however, attained growth patterns broadly aligned with DHS, supporting external validity. - Measurement schedules varied (1–3 month intervals), leading to differing numbers of contributors by age; analyses restricted to cohorts with monthly measures from birth to 24 months yielded similar results. - Inferences are limited to ages 0–24 months; few cohorts had data beyond 2 years, precluding analysis of later childhood outcomes.