Causes and consequences of child growth failure in low- and middle-income countries

Child growth failure is associated with a higher risk of illness and mortality, which contributed to the United Nations Sustainable Development Goal 2.2 to end malnutrition by 2030. Current prenatal and postnatal interventions, such as nutritional supplementation, have been insufficient to eliminate growth failure in low resource settings -motivating a search for key age windows and population subgroups in which to focus future preventive efforts. Quantifying the effect of early growth failure on severe outcomes is important to assess burden and longer-term impacts on the child. Here we show through an analysis of 35 longitudinal cohorts (108,336 children) that maternal and child characteristics at birth accounted for the largest attributable differences in growth. Yet, postnatal growth failure was larger than differences at birth, and characteristics of the child's household environment were additional determinants of growth failure after age 6 months. Children who experienced early ponderal or linear growth failure were at much higher risk of persistent growth failure and were 2.0 to 4.8 times more likely to die by age 24 months. High attributable risk from prenatal causes, and severe consequences for children who experienced early growth failure, support a focus on pre-conception and pregnancy as key opportunities for new preventive interventions. Our results suggest that broad improvements in wellbeing will be necessary to eliminate growth failure in low resource settings, but that screening based on weight could help identify children at highest risk of death before age 24 months.

178 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

Introduction 179
Growth failure in the form of stunting, a marker of chronic malnutrition, and wasting, a marker 180 of acute malnutrition, is common among young children in low-resource settings, and may contribute to 181 child mortality and adult morbidity. 1,2 Worldwide, 22% of children under age 5 years are stunted and 7% 182 are wasted, with most of the burden occurring in low-and middle-income counties (LMIC). 3 Current 183 estimates attribute > 250,000 deaths annually to stunting and > 1 million deaths annually to wasting. 2

184
Stunted or wasted children also experience worse cognitive development 4-9 and adult economic Additionally, water, sanitation, and hygiene (WASH) interventions, which aim to reduce childhood 195 infections that may heighten the risk of wasting and stunting in non-emergency settings 19,20 , have had 196 no effect on child growth in several recent large randomized control trials. [21][22][23][24] The small effect sizes of 197 preventative interventions may reflect an incomplete understanding of the optimal way and time to 198 intervene. 25 199 Modest effects of interventions to prevent stunting and wasting in recent decades have spurred 200 renewed interest in efforts to combine rich data sources 26 with advances in statistical methodology 27 to 201 more deeply understand the key causes of child growth failure. 11,22,23,28 Understanding the relationship 202 between the causes and timing of growth failure is also crucial because children who falter early could 203 be at higher risk for more severe growth failure later. In companion articles, we report that the highest 204 rates of incident stunting and wasting occur by age 3 months. 29,30 Behaviours associated with higher risk 205 of stunting or wasting could be targeted by future interventions, and interventions could be optimized 206 to encourage behaviour change before the age at which growth failure occurs. Characteristics associated 207 with higher risk could also be used to identify children at risk of growth failure who might benefit most 208 from preventative interventions. months.

219
Cohorts were assembled as part of the Knowledge Integration (ki) initiative of the Bill & Melinda 220 Gates Foundation which includes a database of millions of participants from studies on childbirth, 221 growth and development. 26 We selected longitudinal cohorts from the database that met five inclusion 222 criteria: 1) conducted in low-or middle-income countries; 2) enrolled children between birth and age 24 223 months and measured their length and weight repeatedly over time; 3) did not restrict enrollment to 224 acutely ill children; 4) enrolled at least 200 children; and 5) collected anthropometric status 225 measurements at least every 3 months (Extended Data Fig 1). Inclusion criteria ensured we could 226 rigorously evaluate the timing and onset of growth failure among children who were broadly 227 representative of populations in low-and middle-income countries. Thirty-one cohorts from 15 228 countries met inclusion criteria, and 94,019 children and 645,869 total measurements were included in 229 this analysis (Fig 1). Child mortality was rare and not reported in many of the ki datasets, so we relaxed

Rank ordered causes of growth failure 245
We selected exposures of interest based on important predictors of stunting and wasting from prior 246 literature that were measured in multiple cohorts and could be harmonized across cohorts for pooled 247 analyses (Fig 1, Extended data table 2). All reported estimates were adjusted for all other measured 248 exposures that we assumed were not on the causal pathway between the exposure of interest and the 249 outcome. For example, the association between maternal height and stunting was not adjusted for a 250 child's birthweight because low maternal height could increase stunting risk through lower child 251 birthweight. 33 Parameters were estimated using targeted maximum likelihood estimation, a doubly-252 robust, semiparametric method that allows for valid inference while adjusting for potential confounders 253 using ensemble machine learning (details in Methods). 27, 34 We estimated cohort-specific parameters, 254 adjusting for measured covariates within each cohort, and then pooled estimates across cohorts using 255 random effects models (Extended data Fig 1). 35,36 When estimating relative risks, Z-score differences, 256 and attributable risk parameters, we chose the reference as the mode of the level of lowest risk across 257 cohorts. We also estimated the effects of optimal dynamic treatment interventions, where no a-priori 258 reference level of low risk was specified, and each child's individual low-risk level of exposure was 259 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20127100 doi: medRxiv preprint estimated from covariates. Timing of exposures varied, from parental and household characteristics 260 present before birth, to fetal or at-birth exposures, and postnatal exposures including breastfeeding and 261 diarrheal disease. We estimated only associations for growth failure occurring after exposure 262 measurements to ensure time-ordering of exposures and outcomes.

263
Longer child birth length, higher maternal weight, earlier child birth order, higher maternal 264 educations, and more rooms in the household were five of the top six population-level predictors of 265 higher LAZ and WLZ at 24 months, as rank-ordered by population attributable difference, the estimated 266 shift in population mean Z-score if the whole population had their exposure shifted from observed levels 267 to the lowest-risk reference level (Fig 2a, 2b). The pooled, cross-validated R2 for models that included 268 these five key determinants, plus child sex and birthweight, was 0.29 for LAZ (N= 15 cohorts, 22,193 269 children) and 0.09 for WLZ (N=15 cohorts, 20,927 children). The dry season of the year was also an 270 important predictor of higher WLZ, and taller mother's height was an important predictor of higher LAZ.

271
Mother's height was a stronger predictor of both LAZ and WLZ than father's height, which may reflect 272 that maternal anthropometric status integrates across multiple distal and proximate causes, such as 273 family socio-economic status (SES), fetal growth environment, and breastmilk quality. Maternal height 274 and weight and child characteristics measured at birth were the strongest predictors of LAZ and WLZ at 275 age 24 months; beyond those, key predictors of higher Z-scores included markers of better household 276 socioeconomic status (e.g., number of rooms in the home, parental education, clean cooking fuel use, 277 household wealth index) and having a cesarean birth, which may reflect healthcare access or larger fetal 278 size. The findings underscore the importance of prenatal exposures for child growth outcomes, and at 279 the population-level growth failure may be difficult to shift without broad improvements in standard of 280 living. 10,37 Exclusive or predominant breastfeeding before 6 months of age, which was not a major 281 predictor of Z-scores at 24 months, was more strongly associated with higher WLZ than with higher LAZ 282 at 6 months of age (Extended Data Figs 2,3,4).

283
Maternal anthropometric status can influence child Z-scores by affecting fetal growth and birth 284 size. 38,39 In a secondary analysis, we estimated the association between parental anthropometric status 285 and child Z-scores controlling for child birth characteristics, which showed that the relationship of 286 maternal anthropometric status to child Z-scores was only partially mediated by child birth 287 characteristics (Extended data Fig 5). Maternal weight and BMI could directly affect postnatal health 288 through breastmilk quality, or reflect family poverty, genetics, undernutrition, or food insecurity, or 289 family lifestyle and diet. 40,41

290
The strongest predictors of stunting and wasting estimated through population attributable 291 fractions closely matched those identified for child LAZ and WLZ at 24 months (Extended Data Fig 6) Fig 6a). Patterns in associations across growth outcomes were broadly consistent, 298 except for preterm birth, which had a stronger association with stunting outcomes than wasting 299 outcomes, and rainy season, which had a stronger association with wasting outcomes than stunting 300 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20127100 doi: medRxiv preprint outcomes (Extended Data Fig 2). Direction of associations did not vary across regions, but magnitude 301 did, notably with male sex less strongly associated with low LAZ in South Asia, and higher parental 302 education and larger household size more strongly associated with higher WLZ in South Asia (Extended 303 Data Figs 7,8).

305
Age-varying effects on growth failure 306 Maternal height and weight consistently arose as key predictors of population attributable 307 differences in child growth failure (Fig 2), so we sought to elucidate the longitudinal relationship 308 between maternal anthropometry and child growth. We estimated trajectories of mean LAZ and WLZ 309 stratified by maternal height, weight, and BMI. We found that maternal height strongly influenced at-310 birth LAZ, but that LAZ progressed along similar trajectories through age 24 months regardless of 311 maternal height (Fig 3a), with similar though slightly less pronounced differences when stratified by 312 maternal weight (Fig 3b). By contrast, children born to taller mothers had similar WLZ at birth and WLZ 313 trajectories until age 3-6 months, when they diverged substantially (Fig 3a); WLZ trajectory differences 314 were even more pronounced when stratified by maternal weight (Fig 3b). Maternal BMI strongly 315 influenced WLZ, but not LAZ, at birth (Fig 3c). The findings illustrate how maternal status strongly 316 influences where child growth trajectories start, but that growth trajectories evolve in parallel, seeming 317 to respond similarly to postnatal insults independent of their starting point.

318
Children who were stunted by age 3 months exhibited a different longitudinal growth trajectory 319 from those who were stunted later. 29 We hypothesized that causes of growth failure could differ by age 320 of growth failure onset. For key exposures identified in the population attributable effect analyses, we 321 conducted analyses stratified by age of onset and in many cases found age-varying effects (Fig 3d). For 322 example, most measures of socioeconomic status were associated with incident wasting or stunting only 323 after age 6 months, and higher birth order lowered growth failure risk under age 6 months, but 324 increased risk thereafter. Stronger relationships between key socio-demographic characteristics and 325 child wasting and stunting as children age likely reflects the accumulation of insults that result from a 326 child's household conditions, particularly as children begin complementary feeding, exploring their 327 environment, and potentially face higher levels of food insecurity in homes with multiple children. 42

328
When viewed across multiple definitions of child growth failure, most causes had stronger associations 329 with severe stunting, severe wasting, or persistent wasting (> 50% of measurements < -2 WLZ), rarer 330 but more serious outcomes, than with incidence of any wasting or stunting (Fig 3e). Additionally, the 331 characteristics strongly associated with lower probability of recovering from a wasting episode in 90 332 days (birth size, small maternal stature, lower maternal education, later birth order, and male sex) were 333 also characteristics associated with higher risk of wasting prevalence and cumulative incidence

Consequences of early growth failure 338
We documented high incidence rates of wasting and stunting from birth to age 6 months. 29,30 339 Individual studies have suggested that early wasting could predispose children to later linear growth 340 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20127100 doi: medRxiv preprint failure. [43][44][45][46] We hypothesized that early wasting could contribute to subsequent linear growth 341 restriction, and early growth failure could be consequential for persistent growth failure and mortality 342 during the first 24 months of life. Among cohorts with monthly measurements, we examined age-343 stratified linear growth velocity by quartiles of WLZ at previous ages. We found a consistent, exposure-344 response relationship between higher mean WLZ and faster linear growth velocity in the following 3 345 months (Fig 4a), with a corresponding inverse relationship between WLZ and incident stunting at older 346 ages (Extended data Fig 9). Persistent wasting from birth to 6 months (defined as > 50% of 347 measurements wasted) was the wasting measure most strongly associated with incident stunting at 348 older ages (Fig 4b).

349
We next examined the relationship between measures of growth failure in the first 6 months 350 and serious growth-related outcomes: persistent wasting from 6-24 months and concurrent wasting and

356
Finally, we estimated the relative risk of mortality across measures of growth failure in the first 357 6 months within eight cohorts that reported mortality endpoints, including 2,510 child deaths by age 24 358 months (4.3% of children in the cohorts). Analyses used all-cause mortality occurring before children 359 turned two years old (Extended data Fig 10). All measures of early growth failure were significantly 360 associated with higher risk of death by age 24 months, and those most strongly associated with death 361 were severely underweight before age 6 months (RR=4.8, 95% CI: 4.1, 5.6), concurrent wasting and 362 stunting (RR=4.8, 95% CI: 3.9, 5.9), and persistent wasting under 6 months (RR=3.4, 95% CI: 3.0, 3.8) ( Fig 11), and an unmeasured confounder would on average need to almost double the 376 risk of both the exposure and the outcome to explain away observed significant associations (median E-377 value: 1.45, Extended Data Fig 12). 54 Finally, included cohorts were highly monitored, so mortality rates 378 were likely lower than in the general population, and without detailed medical histories, growth failure 379 prior to death may have been a sequela of an underlying condition like malaria or a severe respiratory 380 infection that caused death, rather than the cause itself.

381
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. postnatal Z-scores among children born to different maternal phenotypes was much larger than 400 differences at birth, indicating that growth trajectories were not fully "programmed" at birth (Fig 3a-c).

401
Wasting and stunting incidence was highest before age 6 months, but mean LAZ decreased until age 18 402 months, 29 the dangerous concurrence of wasting and stunting peaked at age 18 months, 30

424
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020.                       CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020.

518
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020.              and the optimal intervention attributable difference on the Y-axis, where the level 568 the exposure is shifted to can vary by child. The optimal intervention attributable 569 differences, which are not estimated with an a-priori specified low-risk reference 570 level, were generally close to the static attributable differences, indicating that the 571 chosen reference levels were the lowest risk strata in most or all cohorts. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20127100 doi: medRxiv preprint 585 586 587 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.  Analyses were conducted in R version 3.6.2. All pooled, regional, and cohort-specific results, results for

677
Children were assumed to never recover from stunting episodes, but children were classified as 678 recovered from wasting episodes (and at risk for a new episode of wasting) if their measured WLZ was ≥

679
-2 for at least 60 days (details in Mertens et. al (2020)). 4 Child mortality was all-cause and was restricted 680 to children who died after birth and before age 24 months. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020.

718
Prevalence ratios (PR) between comparison levels of the exposure, compared to the reference 719 level at birth, 6 months, and 24 months. Prevalence was estimated using anthropometry 720 measurements closest to the age of interest and within one month of the age of interest, except 721 for prevalence at birth which only included measures taken on the day of birth.

722
Cumulative incidence ratios (CIR) between comparison levels of the exposure, compared to the 723 reference level, for the incident onset of outcomes between birth and 24 months, 6-24 months, 724 and birth-6 months.

725
Mean Z-scores by continuous age, stratified by levels of exposures, from birth to 24 months 726 were fit within individual cohorts using cubic splines with the bandwidth chosen to minimize the 727 median Akaike information criterion across cohorts. 5 We estimated splines separately for each 728 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20127100 doi: medRxiv preprint exposure category. We pooled spline curves across cohorts into a single prediction, offset by 729 mean Z-scores at one year, using random effects models. 6 730 731

732
We estimated three measures of the population-level effect of exposures on growth failure 733 outcomes:

734
Population attributable difference was defined as the change in population mean Z-score if the 735 entire population's exposure was set to an ideal reference level. For each exposure, we chose 736 reference levels as the category with the highest mean LAZ or WLZ across cohorts.

737
Population attributable fraction (PAF) was defined as the proportional reduction in cumulative 738 incidence if the entire population's exposure was set to an ideal low risk reference level. We 739 estimated the PAF for the prevalence of stunting and wasting at birth, 6, and 24 months and 740 cumulative incidence of stunting and wasting from birth to 24 months, 6-24 months, and from 741 birth to 6 months. For each exposure, we chose the reference level as the category with the 742 lowest risk of stunting or wasting.

743
Optimal individualized intervention impact We employed a variable importance measure (VIM) 744 methodology to estimate the impact of an optimal individualized intervention on an exposure.

745
The optimal intervention on an exposure was determined through estimating individualized

765
Concurrent wasting and stunting prevalence at age 18 months was estimated using the 766 anthropometry measurement taken closest to age 18 months, and within 17-19 months 767 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20127100 doi: medRxiv preprint of age, and a child was considered concurrently stunted and wasted if they had both a 768 LAZ <-2 and a WLZ <-2.

769
Persistent wasting was estimated from child measurements between 6 and 24 months 770 of age, and a child was considered persistently wasted if ≥50% of measurements of WLZ 771 were < -2.

773
We estimated associations between the outcomes listed in 6.1 and the cumulative 774 incidence of wasting, stunting, underweight, and concurrent wasting and stunting from 775 birth to 6 months of age, and persistent wasting before 6 months of age. In the analysis 776 of child mortality, we also estimated associations between mortality and the cumulative 777 incidence of wasting, stunting, underweight, and concurrent wasting and stunting from 778 birth to 24 months of age, and persistent wasting before 24 months of age.

784
We were also interested in determining if low weight-for-length preceded slower linear growth 785 velocity or the onset of stunting. We estimated the difference in linear growth over three-month 786 periods across quartiles of mean WLZ in the prior three-month period and the relative risk of 787 stunting cumulative incidence over three-month periods across quartiles of mean WLZ in the prior 788 three-month period. We calculated linear growth velocity as the change in length in centimeters 789 within 3-month age intervals, including measurements within a two-week window around each age For each exposure, we used directed acyclic graphs (DAGs) to identify potential confounders from the 796 broader set of exposures used in the analysis. 9 We did not condition on characteristics that were

803
To flexibly adjust for potential confounders and reduce the risk of model misspecification, we 804 estimated adjusted PRs, CIRs, and mean differences using targeted maximum likelihood estimation 805 (TMLE), a two-stage estimation strategy that incorporates state-of-the-art machine learning algorithms 806 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20127100 doi: medRxiv preprint (super learner) while still providing valid statistical inference. 11,12 The super learner is an ensemble 807 machine learning method that uses cross-validation to select a weighted combination of predictions 808 from a library of algorithms. 13 We included in the library simple means, generalized linear models, LASSO 809 penalized regressions, 14 generalized additive models, 15 and gradient boosting machines. 16 The super 810 learner was fit to maximize the 10-fold cross-validated area under the receiver operator curve (AUC) for 811 binomial outcomes, and minimize the 10-fold cross-validated mean-squared error (MSE) for continuous 812 outcomes. That is, the super learner was fit using 9/10 of the data, while the AUC/MSE was calculated 813 on the remaining 1/10 of the data. Each fold of the data was held out in turn and the cross-validated 814 performance measure was calculated as the average of the performance measures across the ten folds.

815
This approach is practically appealing and robust in finite samples, since this cross-validation procedure 816 utilizes unseen sample data to measure the estimator's performance. Also, the super learner is 817 asymptotically optimal in the sense that it is guaranteed to outperform the best possible algorithm 818 included in the library as sample size grows. The initial estimator obtained via super learner is 819 subsequently updated to yield an efficient double-robust semi-parametric substitution estimator of the 820 parameter of interest. 11 To estimate the R 2 of models including multiple exposures, we fit super learner 821 models, without the targeted learning step, and within each cohort measuring the exposures. We then 822 pooled cohort-specific R 2 estimates using fixed effects models.

823
Unadjusted PRs and CIRs between the reference level of each exposure and comparison levels 824 were estimated using logistic regressions. 17 Unadjusted mean differences for continuous outcomes were 825 estimated using linear regressions.

826
We estimated influence curve-based, clustered standard errors to account for repeated 827 measures in the analyses of recovery from wasting or progression to severe wasting. We assumed that 828 the children were the independent units of analysis unless the original study had a clustered design, in 829 which case the unit of independence in the original study were used as the unit of clustering. We used 830 clusters as the unit of independence for the iLiNS-Zinc, Jivita-3, Jivita-4, Probit, and SAS Complementary 831 Feeding trials. We estimated 95% confidence intervals for incidence using the normal approximation.

834
We did not estimate relative risks between a higher level of exposure and the reference group if there 835 were 5 or fewer cases in either stratum. In such cases, we still estimated relative risks between other 836 exposure strata and the reference strata if those strata were not sparse.

890
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20127100 doi: medRxiv preprint (c) Estimates compare risk among children with each measure of growth failure before 891 age 6 months with risk among children who did not experience the specific measure 892 of growth failure.

895
We compared estimates pooled using random effects models, which are more conservative in the 896 presence of heterogeneity across studies, with estimates pooled using fixed effects, and we compared 897 adjusted estimates with estimates unadjusted for potential confounders. We estimated associations 898 between growth failure and mortality at different ages, after dropping the trials measuring children less 899 frequently than quarterly, and we plotted Kaplan Meier curves of child mortality, stratified by measures 900 of early growth failure. We also conducted a sensitivity analysis on methods of pooling splines of child 901 growth trajectories, stratified by maternal anthropometry. We re-estimated the attributable differences

932
. CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020.    is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. length-for-age Z-scores. 1046 Exposures, rank ordered by population attributable difference on child LAZ at 24 1047 months, stratified by the age of the child at the time of anthropometry measurement. 1048 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. Exposures, rank ordered by population attributable difference on child WLZ at 24 1062 months, stratified by the age of the child at the time of anthropometry measurement. 1063 The population attributable difference is the expected difference in population mean Z-1064 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20127100 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. The plot panels show the mediating effect of adjusting for child birth anthropometry and 1098 at-birth characteristics on the estimated Z-score differences between levels of parental 1099 anthropometry. There is little to partial mediation of the association between parental 1100 anthropometry and child LAZ and WLZ at 24 months by at-birth characteristics, implying 1101 that the causal pathway between parental stature and child growth failure operates 1102 through both affecting birth size as well as other pathways. Primary estimates were 1103 adjusted for all other measured exposures not on the causal pathway, while the 1104 mediation analysis estimates are adjusted for the same set of measured exposures, 1105 plus birth weight, birth length, gestational age at birth, birth order, vaginal birth vs. C-1106 section, and home vs. hospital delivery. Only estimates from cohorts measuring at least 1107 4 of the 6 at-birth characteristics are used to estimate the pooled Z-score differences. 1108 1109 1110 1111 1112 1113 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. wasting.

1117
(a) Exposures, rank ordered by population attributable difference on the cumulative 1118 incidence of child stunting between birth and 24 months. The population 1119 attributable fraction is the estimated proportion of the observed outcome in the 1120 whole population attributable to the exposure. For postnatal and at-birth 1121 exposures, including birth characteristics, breastfeeding practice, food security, 1122 and diarrheal disease, the cumulative incidence of stunting from 6-24 months is 1123 used. 1124 (b) Exposures, rank ordered by population attributable difference on the cumulative 1125 incidence of child wasting between birth and 24 months. The population 1126 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

(which was not certified by peer review)
The copyright holder for this preprint this version posted June 11, 2020. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20127100 doi: medRxiv preprint

1154
Extended Data Figure 8 | Regionally-stratified population attributable differences 1155 in weight-for-length Z-scores. 1156 Exposures, rank ordered by population attributable difference on child WLZ at 24 1157 months, stratified by region. The population attributable difference is the expected 1158 difference in population mean Z-score if all children had the reference level of the 1159 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20127100 doi: medRxiv preprint . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. Relative risk of stunting onset within 3-month age bands, comparing cohort-specific 1195 quartiles on mean WLZ in the prior 3-month age to the reference level of the first 1196 quartile. Children are assumed to not recover from stunting, so only the first measure of 1197 LAZ < -2 is used to define stunting onset. 1198 1199 1200 Extended Data Figure 10 | Age at death and preceding anthropometry 1201 measurements among recorded deaths within the included ki cohorts 1202 Each row is a child, with the furthest right grey dots marking the age at death, and the 1203 colored points mark anthropometry measurements prior to death. Children are sorted to 1204 show the cumulative distribution of mortality by age 24 months (marked on the right Y-axis), 1205 illustrating that more than half of child deaths occurred before 6 months. The figure is based 1206 on 2,247 recorded child deaths in children under the age of 2 years from 8 cohorts. Deaths 1207 were not plotted for 206 children with missing ages of death. 1208 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20127100 doi: medRxiv preprint

1209
Extended Data Figure 11 | Difference between adjusted and unadjusted Z-score 1210 effects by number of selected adjustment variables. 1211 Points mark the difference in estimates unadjusted and adjusted estimates of the 1212 difference in average Z-scores between exposed and unexposed children across 31 1213 cohorts, 33 exposures and length-for-age and weight-for-length Z-score outcomes 1214 included in the analysis. Different cohorts measured different sets of exposures, and a 1215 different number of adjustment covariates were chosen for each cohort-specific 1216 estimate based on outcome sparsity, so cohort-specific estimates adjust for different 1217 covariates and numbers of covariates. The plot shows no systematic bias between 1218 unadjusted and adjusted estimates based on number of covariates chosen. The blue 1219 line shows the average difference between adjusted estimates from unadjusted 1220 estimates, fitted using a cubic spline. 1221 . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020.

Extended data table 2 1249
All exposures included in the analysis, as well as the categories the exposures were classified into 1250 across all cohorts, categorization rules, and the total number of children and percent of children in each 1251 category. We selected the exposures of interest based on variables present in multiple cohorts that met 1252 our inclusion criteria, were found to be important determinants of stunting and wasting in prior 1253 literature, and could be harmonized across cohorts for pooled analyses.  . CC-BY 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The copyright holder for this preprint this version posted June 11, 2020. . https://doi.org/10.1101/2020.06.09.20127100 doi: medRxiv preprint