Associations of Suboptimal Growth with All-Cause and Cause-Specific Mortality in Children under Five YearsA Pooled Analysis of Ten Prospective Studies

Nearly twelve times more likely to die. That is the mortality gap between a child with severe wasting and a well-nourished child the same age. Sit with that number for a moment. Then consider that it took ten studies, pooled across three continents and more than fifty thousand children, to pin that figure down with confidence, because for decades, our estimates were all over the place. Child undernutrition at scale is not a secret. About 170 million children under five are moderately or severely stunted, meaning their height lags badly behind where it should be for their age. Roughly 110 million are moderately or severely underweight. But the question of exactly how much that faltering growth raises the risk of death, and whether mild deficits matter as much as severe ones, proved remarkably hard to answer precisely. Earlier pooling efforts disagreed sharply with each other. A 1994 meta-analysis of eight studies found mortality rate ratios ranging from 2.5 to 8.4 for mildly or moderately underweight children. The two thousand four World Health Organization Comparative Risk Assessment extended that work; then the two thousand eight Lancet Series updated it again. However, the Lancet estimates came in as much as 83 percent smaller than the Comparative Risk Assessment figures for some outcomes. Malaria is a striking example of the divergence: the Comparative Risk Assessment found mortality rate ratios rising from 2.1 for mild underweight all the way to 9.5 for severe underweight. The Lancet found no significant association at all, with odds ratios hovering between 0.8 and 1.6. The studies used different cohorts, different growth reference standards, and different statistical approaches, including an arbitrary 15 percent attenuation applied by the Lancet team to account for socioeconomic confounding. What policymakers were left with was a wide range of plausible numbers and no clear way to choose between them. Olofin and colleagues, working under the Nutrition Impact Model Study, set out to resolve that by building a unified framework from the ground up. They pooled individual-level data from ten prospective cohorts in Africa, Asia, and South America: 53,809 children aged one week to 59 months, contributing 55,359 person-years of follow-up, with 1,315 observed deaths. All three standard anthropometric indicators were converted to Z scores: weight-for-age, or underweight; height-for-age, or stunting; and weight-for-height, or wasting — so deficits on different scales could be compared directly. The reference group was a Z score at or above minus one. A mild deficit sat between minus two and minus one. A moderate deficit ran from minus three to minus two. A severe deficit fell below minus three. Cox proportional hazards models, with the child's age as the time scale and stratified by cohort, produced hazard ratios for each band. The same analysis was run twice: once using the two thousand six WHO growth standards, and once using the older nineteen seventy-seven NCHS and WHO reference — to test whether the choice of yardstick changes the answer. The results are striking in their consistency. For underweight, the pooled hazard ratios using WHO two thousand six standards were 1.52 for mild, 2.63 for moderate, and 9.40 for severe — each with tight confidence intervals. The risk doesn't jump abruptly at some clinical threshold. It rises smoothly and steeply with every step down the Z-score ladder. That monotonic dose-response pattern held across all three indicators and across both growth standards. Now, wasting and stunting measure different things. Stunting, or low height-for-age, reflects chronic, cumulative nutritional deprivation over months or years. Wasting, or low weight-for-height, reflects acute, recent nutritional crisis. The data show they carry different mortality profiles. Severe stunting had a pooled hazard ratio of 5.48. Severe wasting came in at 11.63. At the extreme end, children with very severe wasting — Z scores below minus four — faced hazards elevated more than twenty-two fold compared with well-nourished peers. Wasting is the more immediate predictor of death, which makes biological sense: it signals that the body is burning through its reserves right now. Perhaps the most policy-relevant finding is what happens at the mild end. A child with a Z score between minus two and minus one, or mildly underweight, faces roughly a 52 percent higher all-cause mortality hazard than a well-nourished child. That is not a rounding error. That is a child who looks like they're almost fine but is carrying a mortality risk that most screening programs would never flag. There are roughly 144 million children in that mild stunting category and 148 million in mild underweight globally. Mild does not mean safe. When Olofin and colleagues broke the deaths down by cause, the mechanism came into sharper focus. Of the 1,315 deaths in the pooled sample, 371 were from diarrheal diseases, and 187 from respiratory tract infections. Every degree of anthropometric deficit raised the hazard of dying from both. Even mild deficits produced hazard ratios between 1.55 and 1.92 for respiratory deaths, and between 1.60 and 1.73 for diarrheal deaths. Severe wasting pushed the diarrheal mortality hazard ratio to 12.33. The paper points to a cascade of immunological failures associated with severe undernutrition: reduced immunoglobulin A production, impaired T-lymphocyte function, lower cytokine output, and reduced complement protein C3 — all of which compromise the ability to clear a bacterial or viral infection. Add non-immunological complications such as electrolyte imbalances, hypoglycemia, hypothermia, cardiac and respiratory dysfunction, and a respiratory illness that a well-nourished child survives becomes fatal. For malaria, the team was transparent about the limits of their data. Only three cohorts recorded malaria deaths, totaling 51 cases. The study had less than 45 percent power to detect any meaningful effect. Olofin and colleagues stated plainly they could not draw conclusions about undernutrition and malaria mortality, and called for a larger study in malaria-endemic populations. That kind of candor about what a study cannot answer is genuinely useful; it names the next question rather than pretending it has been resolved. The comparison between growth standards deserves a brief moment. Using the nineteen seventy-seven NCHS and WHO reference instead of the two thousand six WHO standards produced hazard ratios that were consistently larger. Weight-for-age hazard ratios were 13 to 46 percent higher, height-for-age hazard ratios were 11 to 22 percent higher, and weight-for-height hazard ratios were 15 to 52 percent higher. The reason is a reclassification effect. The WHO two thousand six standards tend to classify children into lower Z-score categories than the older reference does, shifting who gets counted as deficient and changing the estimated risk within each band. The practical implication is direct: which standard a program uses determines how many children get flagged as malnourished and how severe their measured risk appears. What the pooled analysis ultimately forces into view is an uncomfortable arithmetic. Programs that screen and treat only moderate and severe wasting or underweight leave the vastly larger pool of mildly affected children outside targeted services. Those children are not fine; they carry a 50 percent excess mortality risk and elevated hazards from the two infectious diseases that kill more young children than almost anything else. The evidence supports three broadly validated responses: promoting exclusive breastfeeding, improving complementary feeding practices, and implementing the WHO case-management guidelines for severe acute malnutrition. These are not novel recommendations. But the gradient these ten cohorts document gives them a sharper foundation — the risk doesn't start at the severe end. It starts at mild. If the arithmetic of preventable death is going to shift, interventions need to reach children well before they reach the bottom of the scale. This lecture was created by ennepō. Go to https://ennepo.ai to Discover, Create and Follow the latest research in your field. Read when you can. Listen when you want to.