BACKGROUND:

Multiple factors constrain the trajectories of child cognitive development, but the drivers that differentiate the trajectories are unknown. We examine how multiple early life experiences differentiate patterns of cognitive development over the first 5 years of life in low-and middle-income settings.

METHODS:

Cognitive development of 835 children from the Etiology, Risk Factors, and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development (MAL-ED) multisite observational cohort study was assessed at 6, 15, 24 (Bayley Scales of Infant and Toddler Development), and 60 months (Wechsler Preschool and Primary Scale of Intelligence). Markers of socioeconomic status, infection, illness, dietary intake and status, anthropometry, and maternal factors were also assessed. Trajectories of development were determined by latent class-mixed models, and factors associated with class membership were examined by discriminant analysis.

RESULTS:

Five trajectory groups of cognitive development are described. The variables that best discriminated between trajectories included presence of stimulating and learning resources in the home, emotional or verbal responsivity of caregiver and the safety of the home environment (especially at 24 and 60 months), proportion of days (0–24 months) for which the child had diarrhea, acute lower respiratory infection, fever or vomiting, maternal reasoning ability, mean nutrient densities of zinc and phytate, and total energy from complementary foods (9–24 months).

CONCLUSIONS:

A supporting and nurturing environment was the variable most strongly differentiating the most and least preferable trajectories of cognitive development. In addition, a higher quality diet promoted cognitive development while prolonged illness was indicative of less favorable patterns of development.

What’s Known on This Subject:

The Nurturing Care Framework highlights the importance of a stimulating and nurturing environment. Most studies of childhood cognitive development are characterized by cross-sectional analyses at key ages (e.g., school readiness or the first 1000 days) to identify drivers.

What This Study Adds:

In our longitudinal approach from 6 to 60 months of age, we identified early life factors including a stimulating environment, maternal education, illness, and infant and young child feeding that differentiate patterns of cognitive development across low- and middle-income settings.

Children living in poverty are disproportionately likely to have constrained development.1  Multiple factors inhibit development including repeated infections and illness, low intake of a nutritious diet, and a less nurturing family environment. Particularly sensitive windows of development have been highlighted;2  however, studies in which researchers examine characteristics of longitudinal patterns of infant development in low- and middle-income countries are needed. Such continued follow-up allows for better interpretation of infant assessments, which have low-to-moderate predictive power.3 

Considerable attention has been devoted to the first 1000 days of life as a critical period in child development. However, more recent work has argued that the second 1000 days may be more influential in establishing a platform for cognitive development.4  Along with others, we have previously shown that a supportive and stimulating home environment is positively associated with development at 5 years of life, whereas enteropathogen exposure from birth to 2 years, mediated by symptomatic illness, negatively impacted cognitive scores at 2 years but not at 5 years.5,6  This supports the argument that fostering a stimulating and nurturing environment, as laid out in the Nurturing Care Framework,7  is beneficial throughout the pre-school years. In contrast, reducing exposure to enteropathogens is likely to be beneficial in the short-term but may have a lesser impact on shaping the overall pattern of development.

In this study, we examine groups of trajectories of early cognitive development using 4 time points from 6 to 60 months of age in the Etiology, Risk Factors, and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development (MAL-ED) cohort of children who were sampled intensely over the first 2 years of life. Building on our earlier work, which described the effects of variables on the mean of the study population, we now evaluate if there are distinct trajectories of cognitive development that can be differentially identified by early life factors. With this work, therefore, we leverage longitudinal observations of cognitive trajectories to characterize subpopulations who are most at risk and identify early life adversity factors on which to intervene.

The MAL-ED study was a multidisciplinary prospective community-based observational birth cohort study at 8 low- and middle-income sites8 : Dhaka, Bangladesh; Fortaleza, Brazil; Vellore, India; Bhaktapur, Nepal; Loreto, Peru; Naushero Feroze, Pakistan; Venda, South Africa; and Haydom, Tanzania. Sites were selected because of historically high rates of malnutrition and diarrheal disease. Each site recruited ≥200 singleton children within the first 17 days of life who were managed until 24 months of age. Additional funding allowed follow-up at 5 years. Children were eligible for enrollment if they had no health problems at birth, weighed >1500 g, their mother was >16 years old, and the family planned to stay in the study area for >6 months. The study design has been published,8  and brief details are provided below. Each site obtained institutional ethical approval and written consent from caregivers for the original and follow-up studies.

A modified version of the Bayley Scales of Infant and Toddler Development, Third Edition (BSID-III) was administered at 6, 15, and 24 months. Adaptations included changes to enhance cultural relevance following best practices9 : redrawing pictures (or replacing photographs) to locally relevant equivalent images and dropping culturally inappropriate items to ensure that the test was appropriate in each population. Pooling data between sites was supported by detailed psychometric analysis (eg, exploratory and confirmatory factor analyses, multiple indicator multiple cause modeling), refining items to just those that were consistent and reliable across all sites.10  In this article, the cognitive subscale is used.11 

At 60 months, the Wechsler Preschool and Primary Scale of Intelligence, Third Edition (WPPSI-III) was used to assess cognitive function, again with culturally appropriate adaptations. Detailed psychometric analyses were conducted as before and yielded a fluid reasoning factor composed of items from the block design, matrix reasoning, and picture completion subscales that was consistent and reliable across all MAL-ED sites.12 

For analyses, we required three assessments per child, including the WPPSI-III at 60 ± 2 months. Rigorous quality control and assurance meant eliminating data from the Tanzanian (issues with standardized administration) and Peruvian (WPPSI-III data unavailable) sites. Additionally, the psychometric analysis of the 24-month BSID-III showed that the Nepalese data were inconsistent with the data from other sites and could not be pooled (although valid within-site),10  hence these data were also excluded because dropping the 24-month observation for all Nepalese children would introduce a systematic bias.

Our aim with these analyses was to identify factors from early life that distinguished membership in the cognitive trajectory groups. Candidate variables were drawn from an underlying conceptual model that was used to identify factors influencing development at the 24- and 60-month assessments.5,6  Briefly, 58 variables describing illness (personal prevalence of diarrhea, fever, vomiting, and acute lower respiratory infection [ALRI] between 0 and 24 months)13  and enteropathogen exposure (average number of pathogens detected per monthly [first year] or quarterly [second year] nondiarrheal stool samples and additionally by category of pathogen),14  complementary feeding intake (nutrient densities [per 1000 kcal] from monthly 24-hour recalls, 9–24 months),15  nutrient status (mean plasma concentration of hemoglobin, zinc, retinol, ferritin, and transferrin receptor from 3 blood draws: 7, 15, 24 months), biomarkers of gut function (mean monthly or quarterly concentration in the first and second year respectively of myeloperoxidase, neopterin, and α-1-antitrypsin),16  size at birth (weight-for-age z score at enrollment), weight gain velocity (from enrollment to 2 months old), socioeconomic status (a count of specified household assets, mean monthly household income, years of maternal education and whether the child had schooling),17  and the Home Observation for Measurement of the Environment (HOME) Inventory18  at 6, 24, and 60 months were subjected to psychometric analyses (see Supplemental Information for details).19  At the 6- and 24-month assessments, 3 factors were identified that described the environmental safety of the home, child cleanliness, and the emotional and verbal responsivity of the caregiver. The 60-month HOME psychometric assessment resulted in a single factor characterizing the support for learning that was present in the household. Maternal reasoning was assessed using the Raven Progressive Matrices20  when the child was between 6 and 8 months old, and a single factor was supported in psychometric analyses. Maternal depressive symptoms were evaluated at 1, 6, 15, 24, and 60 months by using the Self-Reporting Questionnaire-2021 ; a single factor was reported in the psychometric analyses,22  but data from Brazil could not be pooled with the other sites for the 1- to 24-month time points, whereas all sites could be pooled at the 60-month time point. For these analyses, we chose to include the 60-month assessment because it was correlated (0.4–0.6) with the earlier time points and was useful across all sites.

Cognitive assessments at each of the 4 ages were converted into T scores to eliminate differences in the scale of each assessment.

Groups of children with similar trajectories were identified using latent class-mixed models. Latent class models identify heterogeneous subgroups23  rather than fit a mean curve to the population. Each group can therefore respond differently to the same suite of exposures and consequently follow a different path. Cognitive scores were modeled as a function of age, and these parameters were allowed to vary by trajectory group. A random intercept was included for the site of each child, which was considered a structural rather than explanatory factor. The optimum number of classes was determined statistically by using the Akaike information criterion. Trajectories were then reviewed visually, and those with few children that were similar (in shape and mean T score values) were combined to simplify the interpretation.

Discriminant analysis was then used to identify variables that differentiate between the latent classes. Linear discriminant analysis with shrinkage and correlation-adjusted Student t (CAT) scores was used to account for correlation between independent variables.24  Each variable was used individually to discriminate between trajectory groups and the agreement between the latent classes and those predicted by the discriminant analysis were compared with Cohen κ. Variables were then filtered using the univariate κ statistic, retaining those with κ ≥ 0.1. Retained variables were entered into a multivariable discriminant analysis (both altogether and sequentially), and the sum of the squared CAT scores gives a summary of the relative importance for each variable.25  Missing values for 0.5% of the exposure data were imputed (using predictive mean matching from multiple imputation chain equations and 10 imputations, see Supplemental Information). Analyses were performed in R 3.6.0 (R Foundation for Statistical Computing, Vienna, Austria).

Cognitive data were included from 835 children from 5 sites (Bangladesh, India, Pakistan, Brazil, and South Africa). The children represented 75% to 98% of those re-enrolled at 5 years. Descriptive child and family characteristics are provided in Table 1. There were considerable differences across sites in maternal education that ranged from a median of 11 years (Brazil and South Africa) to 2 years (Pakistan), mean monthly household income (US$349 in Brazil and US$72 in India), enrollment weight-for-age z score (0.04 in Brazil and −1.29 in Pakistan), duration of exclusive breastfeeding (110 days in Bangladesh but only 12.5 days in Pakistan), and proportion of the first 2 years of life when the mother reported symptoms of child illness (most respiratory) that was much higher in Pakistan (32%) than other sites (6% averaged across all other sites).

TABLE 1

Selected Characteristics of the Analytic Population

VariableDhaka, BangladeshFortaleza, BrazilVellore, IndiaNaushero Feroze, PakistanVenda, South Africa
Sex, n (%), male 89 (51.6) 36 (58.1) 111 (46.4) 93 (51.6) 81 (51.2) 
First child, n (%) 72 (39.1) 27 (31.4) 68 (33.2) 42 (21.9) 60 (36.1) 
Enrollment wt-for-age, z score, median (IQR) −1.09 (−1.79 to −0.54) 0.04 (−0.57 to 0.68) −1.02 (−1.75 to −0.48) −1.29 (−1.94 to −0.68) −0.24 (−0.9 to 0.33) 
Exclusive breastfeeding, d, median (IQR) 110 (66 to 157) 78 (29 to 132) 81 (43 to 113) 12.5 (7 to 15) 26 (15 to 46) 
Reported child illness,% d, 0–24 mo, median (IQR) 13 (9 to 18) 1 (0 to 2) 9 (6 to 13) 32 (22 to 48) 1 (1 to 2) 
Number of household assets, 0–8, median (IQR) 3.3 (2.0 to 4.4) 7.3 (6.8 to 7.5) 3.5 (2.3 to 5.3) 2.5 (1.0 to 4.3) 7.3 (6.3 to 7.8) 
Household monthly income, $, median (IQR) 116 (85.9 to 154) 349 (315 to 419) 71.7 (57.5 to 97.9) 143 (88.5 to 219) 232 (164 to 375) 
Maternal education, y, median (IQR) 5 (2 to 7) 11 (8 to 12) 8 (4 to 9) 2 (0 to 5) 11 (9 to 12) 
VariableDhaka, BangladeshFortaleza, BrazilVellore, IndiaNaushero Feroze, PakistanVenda, South Africa
Sex, n (%), male 89 (51.6) 36 (58.1) 111 (46.4) 93 (51.6) 81 (51.2) 
First child, n (%) 72 (39.1) 27 (31.4) 68 (33.2) 42 (21.9) 60 (36.1) 
Enrollment wt-for-age, z score, median (IQR) −1.09 (−1.79 to −0.54) 0.04 (−0.57 to 0.68) −1.02 (−1.75 to −0.48) −1.29 (−1.94 to −0.68) −0.24 (−0.9 to 0.33) 
Exclusive breastfeeding, d, median (IQR) 110 (66 to 157) 78 (29 to 132) 81 (43 to 113) 12.5 (7 to 15) 26 (15 to 46) 
Reported child illness,% d, 0–24 mo, median (IQR) 13 (9 to 18) 1 (0 to 2) 9 (6 to 13) 32 (22 to 48) 1 (1 to 2) 
Number of household assets, 0–8, median (IQR) 3.3 (2.0 to 4.4) 7.3 (6.8 to 7.5) 3.5 (2.3 to 5.3) 2.5 (1.0 to 4.3) 7.3 (6.3 to 7.8) 
Household monthly income, $, median (IQR) 116 (85.9 to 154) 349 (315 to 419) 71.7 (57.5 to 97.9) 143 (88.5 to 219) 232 (164 to 375) 
Maternal education, y, median (IQR) 5 (2 to 7) 11 (8 to 12) 8 (4 to 9) 2 (0 to 5) 11 (9 to 12) 

Six latent trajectory groups were found; however, 2 had visually similar profiles and few children in each (n = 61 and 80); they both increased from 6 months, then declined from 24 to 60 months, and differed only in the mean 15-month BSID-III. Therefore, we combined these 2 groups, leaving 5 trajectories (Fig 1): children with consistently “high” scores throughout; children with “increasing” scores; children with intermediate scores that had an “early” decline or “late” decline; and children with consistently “low” scores. The sample size differed considerably between groups, with 2 groups (high and low) accounting for 61% (506 of 835) of the sample.

FIGURE 1

Profiles of the five latent classes; the bold line shows the mean of each trajectory group. Six classes were identified statistically, but two sparse and similar groups were combined into the late group. The number of children in each class is shown above each plot.

FIGURE 1

Profiles of the five latent classes; the bold line shows the mean of each trajectory group. Six classes were identified statistically, but two sparse and similar groups were combined into the late group. The number of children in each class is shown above each plot.

Close modal

Discriminant analysis scores for the 58 variables are shown in the Supplemental Information and those with some discriminatory power (κ ≥ 0.1) are shown in Fig 2. The cumulative κ statistic, indicating the accuracy of the discriminant analysis, plateaued at 8 variables: 3 aspects of the home environment (environmental safety, support for learning, and the caregiver’s emotional and verbal responsivity), maternal reasoning ability, the proportion of days (0–24 months) for which the child had symptoms of illness, the mean nutrient densities of zinc and phytate, and the total energy from complementary foods (all from 9 to 24 months).

FIGURE 2

Left: the κ agreement between univariate linear discriminant analysis and the latent class model, only showing variables with κ ≥ 0.1. Middle: the multivariable sum of the CAT gives a rank score of the contribution of each variable to the overall discriminant analysis. Right: the cumulative (included from top to bottom) contribution of each variable to the κ agreement of the multivariable discriminant analysis model. a Factors derived from HOME Inventory.

FIGURE 2

Left: the κ agreement between univariate linear discriminant analysis and the latent class model, only showing variables with κ ≥ 0.1. Middle: the multivariable sum of the CAT gives a rank score of the contribution of each variable to the overall discriminant analysis. Right: the cumulative (included from top to bottom) contribution of each variable to the κ agreement of the multivariable discriminant analysis model. a Factors derived from HOME Inventory.

Close modal

Additional variables had diminishing discriminating power. Notably, the next ranking variables also described the home environment from earlier times points, maternal characteristics (both education and depressive scores), and additional nutrient densities from non–breast milk foods. The mean subclinical pathogen detection rate was also relatively highly ranked. Factors such as child sex, anemia, weight-for-age at enrollment, and household assets and income were found to have little discriminatory power (Supplemental Fig 4).

The 17 factors with univariate discriminatory power (Fig 2) were “fairly accurate” (Cohen κ, 0.34; 95% confidence interval, 0.29–0.39) (Table 2) in determining membership in the 5 trajectories from the latent class-mixed model. Their individual contribution to overall classification accuracy is shown in Fig 2. The 2 largest trajectory groups were most accurately classified (>60% correct in the case of both high and low), and the smaller clusters were less accurate (39%–47% correct). The distributions of each of these factors is compared in Table 3.

TABLE 2

Matrix of Classification Errors by the Latent Class Model and Discriminant Analysis Based on the 17 Variables Shown in Fig 2 

Discriminant AnalysisLatent Class-Mixed Model
HighIncreasingEarlyLateLow
High 198 34 28 65 33 
Increasing 18 24 13 10 
Early 11 
Late 19 16 27 10 
Low 35 12 57 29 173 
Discriminant AnalysisLatent Class-Mixed Model
HighIncreasingEarlyLateLow
High 198 34 28 65 33 
Increasing 18 24 13 10 
Early 11 
Late 19 16 27 10 
Low 35 12 57 29 173 

Values give the number of children in each trajectory group.

TABLE 3

Mean (Interquartile Range) of the Observed Data From the 17 Variables That Most Discriminated Between Trajectory Groups, Shown in Order of Most to Least Discriminatory

Trajectory GroupTime, moUnitsHighIncreasingEarlyLateLow
Environmental safety* 24 z score 0.4 (0.7 to 0.7)a 0.4 (−0.1 to 0.7)a −0.3 (−1 to 0.7)b 0.4 (−0.1 to 0.7)a −0.7 (−2 to −0.1)b 
Behaviors and environment to support learning* 60 z score 0.40 (0.01 to 0.82)a 0.28 (−0.1 to 0.71)a −0.15 (−0.6 to 0.21)b 0.11 (−0.2 to 0.21)c −0.18 (−0.4 to 0.21)b 
Maternal reasoning 6–8 Score (0–45) 31 (24 to 40)a 32 (26 to 41)a 25 (12 to 36)b 28 (18 to 38)a,b 19 (8 to 28)c 
Zinc density 9–24 sqrt, g per 1000 kcal 2.2 (1.9 to 2.5)a 2.3 (1.9 to 2.9)a 2.1 (1.8 to 2.5)b,d 2.2 (1.9 to 2.6)a,d 1.9 (1.7 to 1.9)c 
Phytate density 9–24 sqrt, g per 1000 kcal 24 (18 to 26)a 19 (17 to 22)b 23 (16 to 34)c 24 (17 to 34)a,c 18 (15 to 19)d 
Illness** 0–24 % d 9.3 (1.9 to 12)a 8.5 (1.7 to 13)a 17 (1.2 to 27)b 9 (1.2 to 13)a 25 (9.2 to 37)c 
Total energy from complementary foods 9–24 kcal/d 790 (600 to 970)a 720 (540 to 890)a,c 700 (490 to 880)a 650 (380 to 880)b,c 580 (360 to 770)b 
Emotional–verbal responsiveness* 24 z score 0.3 (0.3 to 0.8)a 0.1 (−0.1 to 0.8)b,c −0.3 (−0.6 to 0.3)b,e 0.2 (−0.1 to 0.8)a,c −0.4 (−1 to 0.3)d,e 
Maternal education — 7.6 (5 to 11)a,c 8.7 (6 to 12)a 6.2 (0.75 to 10)c 7.4 (5 to 10)a,c 4.1 (0 to 8)b 
Environmental safety z score 0.3 (0.6 to 0.6)a 0.2 (−0.5 to 0.6)a,b −0.1 (−0.5 to 0.6)b,c 0.2 (−0.5 to 0.6)a,c −0.5 (−2 to 0.6)b 
Emotional–verbal responsiveness* z score 0.4 (−0.1 to 1)a −0.1 (−0.9 to 1)b,c 0 (−0.9 to 1)b,c 0.1 (−0.5 to 1)b −0.4 (−1 to 0.2)c 
Vitamin A density 9–24 sqrt, g per 1000 kcal) 18 (14 to 20)a 20 (14 to 29)b 17 (13 to 20)a,d 17 (12 to 21)b,d 15 (11 to 17)c 
Folate density 9–24 sqrt, g per 1000 kcal 12 (10 to 14)a,b 12 (10 to 13)a 13 (9.5 to 16)b 13 (9.9 to 16)b 11 (9.5 to 11)c 
Nondiarrheal pathogen exposure 0–24 Count 1.1 (0.69 to 1.3)a 1.1 (0.82 to 1.3)a 1.2 (0.81 to 1.6)a,b 1 (0.73 to 1.3)a 1.4 (1 to 1.7)b 
Vitamin E density 9–24 sqrt, g per 1000 kcal 1.6 (1.3 to 1.9)a 1.6 (1.3 to 1.8)a 1.6 (1.2 to 2)a 1.6 (1.3 to 2)a 1.4 (1.2 to 1.6)b 
Maternal depressive symptoms 60 Count, 0–18 2.7 (1 to 4)a 3.4 (1 to 5)a,c 4.4 (1 to 7)b,c,d 3.7 (1 to 5)a,d 5 (2 to 7)b,c 
Nondiarrheal bacterial pathogen exposure 0–24 Count 0.79 (0.54 to 1)a 0.83 (0.59 to 1)a 0.85 (0.6 to 1.1)a,b 0.78 (0.5 to 1)a 0.98 (0.75 to 1.2)b 
Trajectory GroupTime, moUnitsHighIncreasingEarlyLateLow
Environmental safety* 24 z score 0.4 (0.7 to 0.7)a 0.4 (−0.1 to 0.7)a −0.3 (−1 to 0.7)b 0.4 (−0.1 to 0.7)a −0.7 (−2 to −0.1)b 
Behaviors and environment to support learning* 60 z score 0.40 (0.01 to 0.82)a 0.28 (−0.1 to 0.71)a −0.15 (−0.6 to 0.21)b 0.11 (−0.2 to 0.21)c −0.18 (−0.4 to 0.21)b 
Maternal reasoning 6–8 Score (0–45) 31 (24 to 40)a 32 (26 to 41)a 25 (12 to 36)b 28 (18 to 38)a,b 19 (8 to 28)c 
Zinc density 9–24 sqrt, g per 1000 kcal 2.2 (1.9 to 2.5)a 2.3 (1.9 to 2.9)a 2.1 (1.8 to 2.5)b,d 2.2 (1.9 to 2.6)a,d 1.9 (1.7 to 1.9)c 
Phytate density 9–24 sqrt, g per 1000 kcal 24 (18 to 26)a 19 (17 to 22)b 23 (16 to 34)c 24 (17 to 34)a,c 18 (15 to 19)d 
Illness** 0–24 % d 9.3 (1.9 to 12)a 8.5 (1.7 to 13)a 17 (1.2 to 27)b 9 (1.2 to 13)a 25 (9.2 to 37)c 
Total energy from complementary foods 9–24 kcal/d 790 (600 to 970)a 720 (540 to 890)a,c 700 (490 to 880)a 650 (380 to 880)b,c 580 (360 to 770)b 
Emotional–verbal responsiveness* 24 z score 0.3 (0.3 to 0.8)a 0.1 (−0.1 to 0.8)b,c −0.3 (−0.6 to 0.3)b,e 0.2 (−0.1 to 0.8)a,c −0.4 (−1 to 0.3)d,e 
Maternal education — 7.6 (5 to 11)a,c 8.7 (6 to 12)a 6.2 (0.75 to 10)c 7.4 (5 to 10)a,c 4.1 (0 to 8)b 
Environmental safety z score 0.3 (0.6 to 0.6)a 0.2 (−0.5 to 0.6)a,b −0.1 (−0.5 to 0.6)b,c 0.2 (−0.5 to 0.6)a,c −0.5 (−2 to 0.6)b 
Emotional–verbal responsiveness* z score 0.4 (−0.1 to 1)a −0.1 (−0.9 to 1)b,c 0 (−0.9 to 1)b,c 0.1 (−0.5 to 1)b −0.4 (−1 to 0.2)c 
Vitamin A density 9–24 sqrt, g per 1000 kcal) 18 (14 to 20)a 20 (14 to 29)b 17 (13 to 20)a,d 17 (12 to 21)b,d 15 (11 to 17)c 
Folate density 9–24 sqrt, g per 1000 kcal 12 (10 to 14)a,b 12 (10 to 13)a 13 (9.5 to 16)b 13 (9.9 to 16)b 11 (9.5 to 11)c 
Nondiarrheal pathogen exposure 0–24 Count 1.1 (0.69 to 1.3)a 1.1 (0.82 to 1.3)a 1.2 (0.81 to 1.6)a,b 1 (0.73 to 1.3)a 1.4 (1 to 1.7)b 
Vitamin E density 9–24 sqrt, g per 1000 kcal 1.6 (1.3 to 1.9)a 1.6 (1.3 to 1.8)a 1.6 (1.2 to 2)a 1.6 (1.3 to 2)a 1.4 (1.2 to 1.6)b 
Maternal depressive symptoms 60 Count, 0–18 2.7 (1 to 4)a 3.4 (1 to 5)a,c 4.4 (1 to 7)b,c,d 3.7 (1 to 5)a,d 5 (2 to 7)b,c 
Nondiarrheal bacterial pathogen exposure 0–24 Count 0.79 (0.54 to 1)a 0.83 (0.59 to 1)a 0.85 (0.6 to 1.1)a,b 0.78 (0.5 to 1)a 0.98 (0.75 to 1.2)b 

Values within a row with different superscript letters are significantly different (P ≤ .05 with Bonferroni correction tested using Kolmogorov–Smirnov tests). sqrt, square root; —, not applicable.

*

HOME z score with higher scores indicating a more stimulating and responsive environment.

**

Illness defined as the percent of days between 0 and 24 mo with reported diarrhea, fever, vomiting, or ALRI.

Radar plots contrast the profiles of the 8 most-discriminatory factors influencing membership in each group (Fig 3). Stimulating and nurturing aspects of the home environment and maternal reasoning ability were strongly positively associated with the consistently high or increasing trajectories and strongly negatively correlated with the early and low groups. The late group had positive correlations with earlier (24 month) descriptions of a safe environment and maternal reasoning ability but a null correlation with the later (60 month) assessment of support for learning.

FIGURE 3

Profiles for each trajectory group showing the CAT comparing each group to the pooled mean for the top 8 discriminating variables. Scores that are closer to the center indicate a negative association between a given variable and the probability of belonging to a class; scores that are further to the outer ring have a positive association with the probability of belonging to a class. The solid circle indicates a t score of zero, hence no association between the variable and a trajectory class.

FIGURE 3

Profiles for each trajectory group showing the CAT comparing each group to the pooled mean for the top 8 discriminating variables. Scores that are closer to the center indicate a negative association between a given variable and the probability of belonging to a class; scores that are further to the outer ring have a positive association with the probability of belonging to a class. The solid circle indicates a t score of zero, hence no association between the variable and a trajectory class.

Close modal

Aspects of complementary feeding also influenced membership across trajectories. Greater energy intake was positively associated with the high group and negatively associated with the late and low groups. The combination of zinc and phytate densities is informative for differentiating the various groups. Higher zinc density and lower phytate density are associated with the increasing group, but the reverse is true for the early group. Greater zinc and phytate densities are associated with the late group but are negatively associated with the low group.

As shown, illness prevalence was strongly positively associated with membership in the low group and more weakly with membership in the early group. In contrast, greater time spent ill was negatively associated with the high, increasing, and late groups.

We describe 5 clusters depicting differing trajectories of development in resource-limited settings. A harmonized protocol with rigorous quality control and extensive psychometric analyses to support pooling of cognitive development assessments across settings permitted this novel analysis. Because of the rich multidisciplinary focus in the study, we were able to evaluate a wide range of influences on child development over time. This study represents a unique opportunity to consider the developmental paths of pre-school aged children and the role of early environmental factors influencing those paths.

The contrasts between life experiences of children in the positive trajectories of the high and increasing groups and children in the low group, a concerning pattern of development, are most striking. The most discriminating factors were aspects of the home environment with higher nurturing associated with the high and increasing groups and less nurturing associated with the low group. Our assessment of the home environment captured the opportunities, constraints, and demands afforded to these children within the context of their home environments. Although there are societal differences in parenting behavior, there is strong evidence that provision of certain experiences to children within their home context is universally related to child adaptive functioning and development. These experiences include (1) warmth and responsiveness, (2) discipline and control, and (3) stimulation and teaching.2628  Indeed, accumulating evidence reveals that parental participation in cognitively stimulating activities like reading, provision of child-appropriate learning materials, and sensitive and responsive interactions between caregiver and child are significant contributors to optimal child development.2931  The MAL-ED children live in settings characterized by poverty, and yet those living in a more stimulating and responsive home environment were more likely to be in groups with higher scoring trajectories. Greater maternal reasoning ability may also lead to a more responsive or nurturing environment; here, it is independently associated with cognitive trajectories in the same pattern as aspects of the nurturing home environment. As evidenced in the recent study by Zhu et al,32  the factors identified in our study as influencing early life trajectories can also differentiate functioning in midchildhood and adolescence.

Higher socioeconomic status is a frequently reported driver of cognitive development.33  However, in these analyses, aspects of wealth (household income or assets) did not differentiate trajectories despite the fact that both were associated with cognitive development at specific ages.5,6  We posit that our consideration of multiple aspects of the home environment, more closely associated with behavior and engagement over time, displaced influences associated with wealth.34,35 

Other key factors differentiating these contrasting groups related to child health and variation in infant feeding practices. The group with consistently low scores was characterized by high rates of illness in the first 2 years of life, higher rates of enteropathogen detection, and lower intakes from complementary foods in terms of energy and densities of zinc and phytate. Chronic illness during childhood is negatively associated with cognitive development and consequently delays school readiness.36  In our study, we examined the prevalence of symptoms of illness in early life that were related to cognition in our earlier analyses at 245  but not 60 months.6  Infants were excluded from the study if they weighed <1500 g or had serious health problems at birth, which ought to have excluded the most vulnerable children who might be more prone to illnesses; however, despite this, children in the low group had ≥1 illness symptom for an average 25% of their first 2 years of life. The most common symptoms were respiratory.13  Combined, the negative roles of illness, infection, and poor infant feeding practices in settings with less promotion of development suggest an identifiable subgroup of vulnerable infants who are at risk for constrained cognitive development.

Higher mean energy from non–breast milk foods was positively associated with the high group, and greater zinc and lower phytate densities were associated with the increasing group. The zinc and phytate densities also influenced membership in the early and late groups. Our earlier analyses found that diets richer in B vitamins were positively associated with cognitive scores at 24 months5  and that greater intakes of nutrients consistent with animal source foods were positively associated with cognitive scores at 60 months.6  Analyses at 24 months were determined by specific hypotheses regarding dietary factors, hemoglobin, and cognitive development, whereas those at 5 years broadly considered dietary exposures, as we have done here. In this data set, the zinc density of the diet is associated with iron, folate, and protein densities, and a lower phytate density is associated with a more bioavailable trace mineral diet or a diet with more animal food sources. In this way, our current findings are consistent with our earlier findings.

There were unexpected findings from this analysis, the most notable of which was the relatively weak discriminatory power of maternal depressive symptoms, which is a known correlate with child development, especially through mother–child interactions.37  In this study, we included maternal depressive symptoms at 60 months in an effort to capture underlying (not postpartum) depression, although considerable development has already occurred. As noted earlier, we chose the latter time point because of its psychometric properties and correlations with earlier assessments. The high univariate association but lower multivariate association between maternal depressive symptoms and the trajectory groups suggests that maternal depression covaries with other aspects of the nurturing environment, which were captured by other included variables. Anemia and micronutrient deficiencies,38  and gut inflammation39  in the first 2 years of life, were weakly associated with trajectory groups. If data were available on micronutrient deficiencies or gut inflammation after 24 months, it is possible that we may have found an association. This temporal resolution (high in the first 2 years but low thereafter) is a limitation of the study, as was the loss of children to follow-up after the original study had ended. Additionally, the rigorous requirements for inclusion meant that we necessarily excluded some children and some sites from analyses. As children get older, the importance of their microenvironment diminishes as the mesoenvironment becomes increasingly influential.40  Although the original study design did not capture this expanding web, the breadth of domains that were measured is unique, and with these results, we identify early drivers of cognitive development across markedly different populations using common (and validated) methods.

Even with the breadth of variables collected, the classification agreement between the discriminant analysis and latent classes was at best “fair” (including all variables), and only then for the more populous trajectory groups. The sparser groups (notably the early and late groups) were poorly cross-classified, and some 16% (12 of 76) from the increasing group were assigned to the low group. Figure 1 reveals considerable variability and overlap between groups. It is likely that additional or combined factors were missing from these analyses, but the variables that did emerge to either constrain or support cognitive development fit within the existing evidence of a pattern of vulnerability and the potential of a nurturing environment to make a positive difference. Importantly, these results suggest that positive impacts can be made in the postnatal period, for example, teaching optimal caregiver behaviors through educational sessions and learning opportunities with children including semistructured play and access to materials.4144  With our analyses, we point to patterns that may help identify subpopulations of children who would benefit the most from interventions and also indicate aspects of the environment on which to intervene.

Despite their impoverished environments, some children had preferable trajectories of cognitive development through age 60 months. Such patterns were characterized by fewer illnesses, a more nutrient dense diet from complementary foods, having a mother with greater reasoning ability, and, crucially, living in a home with a more nurturing and stimulating environment over the pre-school period. Although we cannot provide causal evidence, with our findings, we support the need for strengthening comprehensive services to improve child health, nutrition, and the responsivity of the home environment to support cognitive development45  and suggest that specific early life characteristics can identify a group of children at high risk who would benefit from additional inputs to support their development through the pre-school period.

We thank the staff and participants of the MAL-ED Network for their important contributions. The findings and conclusions in this report are those of the authors and do not necessarily represent the official position of the US National Institutes of Health or Department of Health and Human Services. MAL-ED Network Investigators are as follows: Angel Mendez Acosta, Rosa Rios de Burga, Cesar Banda Chavez, Julian Torres Flores, Maribel Paredes Olotegui, Silvia Rengifo Pinedo, Mery Siguas Salas, Dixner Rengifo Trigoso, and Angel Orbe Vasquez (A.B. PRISMA, Iquitos, Peru); Imran Ahmed, Didar Alam, Asad Ali, Zulfiqar A Bhutta, Shahida Qureshi, Muneera Rasheed, Sajid Soofi, Ali Turab, and Anita KM Zaidi (Aga Khan University, Karachi, Pakistan); Ladaporn Bodhidatta and Carl J Mason (Armed Forces Research Institute of Medical Sciences, Bangkok, Thailand); Sudhir Babji, Anuradha Bose, Ajila T George, Dinesh Hariraju, M. Steffi Jennifer, Sushil John, Shiny Kaki, Gagandeep Kang, Priyadarshani Karunakaran, Beena Koshy, Robin P Lazarus, Jayaprakash Muliyil, Mohan Venkata Raghava, Sophy Raju, Anup Ramachandran, Rakhi Ramadas, Karthikeyan Ramanujam, Anuradha Bose, Reeba Roshan, Srujan L Sharma, Shanmuga Sundaram E, and Rahul J Thomas (Christian Medical College, Vellore, India); William K Pan (Duke University, Durham, NC; Fogarty International Center, National Institutes of Health, Bethesda, MD); Ramya Ambikapathi, J Daniel Carreon, Vivek Charu, Viyada Doan, Jhanelle Graham, Christel Hoest, and Stacey Knobler (Fogarty International Center, National Institutes of Health, Bethesda, MD); Dennis R Lang (Fogarty International Center, National Institutes of Health, Bethesda, MD; Foundation for the National Institutes of Health, Bethesda, MD); Benjamin JJ McCormick, Monica McGrath, Mark A Miller, Archana Mohale, Gaurvika Nayyar, Stephanie Psaki, Zeba Rasmussen, Stephanie A Richard, Jessica C Seidman, and Vivian Wang (Fogarty International Center, National Institutes of Health, Bethesda, MD); Rebecca Blank, Michael Gottlieb, and Karen H Tountas (Foundation for the National Institutes of Health, Bethesda, MD); Caroline Amour, Eliwaza Bayyo, Estomih R Mduma, Regisiana Mvungi, Rosemary Nshama, John Pascal, Buliga Mujaga Swema, and Ladislaus Yarrot (Haydom Lutheran Hospital, Haydom, Tanzania); Tahmeed Ahmed, AM Shamsir Ahmed, Rashidul Haque, Iqbal Hossain, Munirul Islam, Mustafa Mahfuz, Dinesh Mondal, and Fahmida Tofail (icddr,b, Dhaka, Bangladesh); Ram Krishna Chandyo, Prakash Sunder Shrestha, Rita Shrestha, and Manjeswori Ulak (Institute of Medicine, Tribhuvan University, Kathmandu, Nepal); Aubrey Bauck, Robert E Black, and Laura E Caulfield (Johns Hopkins University, Baltimore, MD); William Checkley (Johns Hopkins University, Baltimore, MD; Fogarty International Center, National Institutes of Health, Bethesda, MD); Margaret N Kosek, Gwenyth Lee, Kerry Schulze, and Pablo Peñataro Yori (Johns Hopkins University, Baltimore, MD); Laura E. Murray-Kolb (The Pennsylvania State University, University Park, PA); A Catharine Ross (The Pennsylvania State University, University Park, PA); Barbara Schaefer (The Pennsylvania State University, University Park, PA; Fogarty International Center, National Institutes of Health, Bethesda, MD); Suzanne Simons (The Pennsylvania State University, University Park, PA); Laura Pendergast (Temple University, Philadelphia, PA); Cláudia B Abreu, Hilda Costa, and Alessandra Di Moura (Universidade Federal do Ceara, Fortaleza, Brazil); José Quirino Filho (Universidade Federal do Ceara, Fortaleza, Brazil; Fogarty International Center, National Institutes of Health, Bethesda, MD); Alexandre Havt, Álvaro M Leite, Aldo AM Lima, Noélia L Lima, Ila F Lima, Bruna LL Maciel, Pedro HQS Medeiros, Milena Moraes, Francisco S Mota, and Reinaldo B Oriá (Universidade Federal do Ceara, Fortaleza, Brazil); Josiane Quetz, Alberto M Soares, and Rosa MS Mota (Universidade Federal do Ceara, Fortaleza, Brazil); Crystal L Patil (University of Illinois, Chicago, IL); Pascal Bessong, Cloupas Mahopo, Angelina Maphula, Emanuel Nyathi, and Amidou Samie (University of Venda, Thohoyandou, South Africa); Leah Barrett, Rebecca Dillingham, Jean Gratz, Richard L Guerrant, Eric Houpt, William A Petri, Jr, James Platts-Mills, and Rebecca Scharf (University of Virginia, Charlottesville, VA); Binob Shrestha and Sanjaya Kumar Shrestha (Walter Reed, Armed Forces Research Institute of Medical Sciences Research Unit, Kathmandu, Nepal); Tor Strand (Walter Reed, AFRIMS Research Unit, Kathmandu, Nepal; University of Bergen, Norway); and Erling Svensen (Haukeland University Hospital, Bergen, Norway; Haydom Lutheran Hospital, Haydom, Tanzania).

Data (with the exception of the cognitive development variables) are available from https://clinepidb.org/ce/app/after registration and approval of proposed use.

Drs McCormick, Richard, and Seidman along with Profs Caulfield and Murray-Kolb designed and ran the analyses, then led the manuscript drafting; Prof Pendergast and Drs Maphula, Koshy, Roshan, Nahar, Shrestha, Haque, Oria, and Svensen, Mr Blacy, and Mrs Rasheed collected, curated, and processed cognitive data; Drs Rasmussen and Scharf contributed to the interpretation of results; and all authors contributed to the editing and approval of the manuscript.

FUNDING: The Etiology, Risk Factors, and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development (MAL-ED) Project was conducted as a collaborative project supported by the Bill and Melinda Gates Foundation (OPP47075), the Foundation for the National Institutes of Health (AI130326), and the Fogarty International Center. Funded by the National Institutes of Health (NIH).

ALRI

acute lower respiratory infection

BSID-III

Bayley Scales of Infant and Toddler Development, Third Edition

CAT

correlation-adjusted Student t

HOME

Home Observation for Measurement of the Environment

MAL-ED

The Etiology Risk Factors and Interactions of Enteric Infections and Malnutrition and the Consequences for Child Health and Development

WPPSI-III

Wechsler Preschool and Primary Scale of Intelligence, Third Edition

1
Lu
C
,
Black
MM
,
Richter
LM
.
Risk of poor development in young children in low-income and middle-income countries: an estimation and analysis at the global, regional, and country level
.
Lancet Glob Health
.
2016
;
4
(
12
):
e916
e922
2
Grantham-McGregor
S
,
Cheung
YB
,
Cueto
S
,
Glewwe
P
,
Richter
L
,
Strupp
B
;
International Child Development Steering Group
.
Developmental potential in the first 5 years for children in developing countries
.
Lancet
.
2007
;
369
(
9555
):
60
70
3
Pollitt
E
,
Triana
N
.
Stability, Predictive validity, and sensitivity of mental and motor development scales and pre-school cognitive tests among low-income children in developing countries
.
Food Nutr Bull
.
1999
;
20
(
1
):
45
52
4
Black
MM
,
Pérez-Escamilla
R
,
Rao
SF
.
Integrating nutrition and child development interventions: scientific basis, evidence of impact, and implementation considerations
.
Adv Nutr
.
2015
;
6
(
6
):
852
859
5
MAL-ED Network Investigators
.
Early childhood cognitive development is affected by interactions among illness, diet, enteropathogens and the home environment: findings from the MAL-ED birth cohort study
.
BMJ Glob Health
.
2018
;
3
(
4
):
e000752
6
McCormick
BJJ
,
Richard
SA
,
Caulfield
LE
, et al;
MAL-ED Network Investigators
.
Early life child micronutrient status, maternal reasoning, and a nurturing household environment have persistent influences on child cognitive development at age 5 years: results from MAL-ED
.
J Nutr
.
2019
;
149
(
8
):
1460
1469
7
World Health Organization, United Nations Children’s Fund, World Bank Group
.
Nurturing Care for Early Childhood Development: A Framework for Helping Children Survive and Thrive to Transform Health and Human Potential
.
Geneva, Switzerland
:
World Health Organization
;
2018
8
MAL-ED Network Investigators
.
The malnutrition and enteric disease study (MAL-ED): understanding the consequences for child health and development
.
Clin Infect Dis
.
2014
;
59
(
suppl 4
):
S193
S330
9
Hambleton
RK
,
Kanjee
A
.
Increasing the validity of cross-cultural assessments: use of improved methods for test adaptations
.
Eur Psycholl Assess
.
1995
;
11
(
3
):
147
157
10
Pendergast
LL
,
Schaefer
BA
,
Murray-Kolb
LE
, et al;
MAL-ED Network Investigators
.
Assessing development across cultures: invariance of the bayley-III scales across seven international MAL-ED sites
.
Sch Psychol Q
.
2018
;
33
(
4
):
604
614
11
Murray-Kolb
LE
,
Rasmussen
ZA
,
Scharf
RJ
, et al;
MAL-ED Network Investigators
.
The MAL-ED cohort study: methods and lessons learned when assessing early child development and caregiving mediators in infants and young children in 8 low- and middle-income countries
.
Clin Infect Dis
.
2014
;
59
(
suppl 4
):
S261
S272
12
Ruan-Iu
L
,
Pendergast
LL
,
Rasheed
M
, et al
.
Assessing early childhood fluid reasoning in low- and middle-income nations: validity of the Wechsler Preschool and Primary Scale of Intelligence across seven MAL-ED sites
.
J Psychoeduc Assess
.
2020
;
38
(
2
):
256
262
13
Richard
SA
,
McCormick
BJJ
,
Seidman
JC
, et al;
On Behalf Of The Mal-Ed Network Investigators
.
Relationships among common illness symptoms and the protective effect of breastfeeding in early childhood in MAL-ED: an eight-country cohort study
.
Am J Trop Med Hyg
.
2018
;
98
(
3
):
904
912
14
Platts-Mills
JA
,
Babji
S
,
Bodhidatta
L
, et al;
MAL-ED Network Investigators
.
Pathogen-specific burdens of community diarrhoea in developing countries: a multisite birth cohort study (MAL-ED)
.
Lancet Glob Health
.
2015
;
3
(
9
):
e564
e575
15
Caulfield
LE
,
Bose
A
,
Chandyo
RK
, et al;
MAL-ED Network Investigators
.
Infant feeding practices, dietary adequacy, and micronutrient status measures in the MAL-ED study
.
Clin Infect Dis
.
2014
;
59
(
suppl 4
):
S248
S254
16
McCormick
BJJ
,
Lee
GO
,
Seidman
JC
, et al
.
Dynamics and trends in fecal biomarkers of gut function in children from 1–24 months in the MAL-ED study
.
Am J Trop Med Hyg
.
2017
;
96
(
2
):
465
472
17
Psaki
SR
,
Seidman
JC
,
Miller
M
, et al;
MAL-ED Network Investigators
.
Measuring socioeconomic status in multicountry studies: results from the eight-country MAL-ED study
.
Popul Health Metr
.
2014
;
12
(
1
):
8
18
Caldwell
BM
,
Bradley
RH
.
Home Observation for Measurement of the Environment
.
Little Rock, AR
:
University of Arkansas
;
1984
19
Jones
PC
,
Pendergast
LL
,
Schaefer
BA
, et al;
MAL-ED Network Investigators
.
Measuring home environments across cultures: invariance of the HOME scale across eight international sites from the MAL-ED study
.
J Sch Psychol
.
2017
;
64
:
109
127
20
Raven
JC
,
Court
JH
.
Manual for Raven’s Progressive Matrices and Vocabulary Scales: Standard Progressive Matrices
.
San Antonio, TX
:
Harcourt Assessment
;
2004
21
Beusenberg
M
,
Orley
JH
.
Users Guide to the Self Reporting Questionnaire (SRQ)
.
Geneva, Switzerland
:
WHO
;
1994
. Available at: www.who.int/iris/handle/10665/61113. Accessed November 1, 2019
22
Pendergast
LL
,
Scharf
RJ
,
Rasmussen
ZA
, et al;
MAL-ED Network Investigators
.
Postpartum depressive symptoms across time and place: structural invariance of the Self-Reporting Questionnaire among women from the international, multi-site MAL-ED study
.
J Affect Disord
.
2014
;
167
:
178
186
23
Berlin
KS
,
Parra
GR
,
Williams
NA
.
An introduction to latent variable mixture modeling (part 2): longitudinal latent class growth analysis and growth mixture models
.
J Pediatr Psychol
.
2014
;
39
(
2
):
188
203
24
Zuber
V
,
Strimmer
K
.
Gene ranking and biomarker discovery under correlation
.
Bioinformatics
.
2009
;
25
(
20
):
2700
2707
25
Ahdesmäki
M
,
Strimmer
K
.
Feature selection in omics prediction problems using cat scores and false nondiscovery rate control
.
Ann Appl Stat
.
2010
;
4
(
1
):
503
519
26
Whiting
BB
,
Edwards
CP
.
Children of Different Worlds: The Formation of Social Behavior
, 3rd ed.
Cambridge, MA
:
Harvard Univ Press
;
1994
27
Bradley
R
,
Corwyn
R
.
Caring for children around the world: a view from HOME
.
Int J Behav Dev
.
2005
;
29
(
6
):
468
478
28
Bradley
RH
.
Constructing and adapting causal and formative measures of family settings: the HOME inventory as illustration
.
J Fam Theory Rev
.
2015
;
7
(
4
):
381
414
29
Hair
NL
,
Hanson
JL
,
Wolfe
BL
,
Pollak
SD
.
Association of child poverty, brain development, and academic achievement
.
JAMA Pediatr
.
2015
;
169
(
9
):
822
829
30
Luby
J
,
Belden
A
,
Botteron
K
, et al
.
The effects of poverty on childhood brain development: the mediating effect of caregiving and stressful life events
.
JAMA Pediatr
.
2013
;
167
(
12
):
1135
1142
31
Shonkoff
JP
,
Garner
AS
;
Committee on Psychosocial Aspects of Child and Family Health
;
Committee on Early Childhood, Adoption, and Dependent Care
;
Section on Developmental and Behavioral Pediatrics
.
The lifelong effects of early childhood adversity and toxic stress
.
Pediatrics
.
2012
;
129
(
1
). Available at: www.pediatrics.org/cgi/content/full/129/1/e232
32
Zhu
Z
,
Chang
S
,
Cheng
Y
, et al
.
Early life cognitive development trajectories and intelligence quotient in middle childhood and early adolescence in rural western China
.
Sci Rep
.
2019
;
9
(
1
):
18315
33
Lawson
GM
,
Hook
CJ
,
Farah
MJ
.
A meta-analysis of the relationship between socioeconomic status and executive function performance among children
.
Dev Sci
.
2018
;
21
(
2
):
e12529
34
Christensen
DL
,
Schieve
LA
,
Devine
O
,
Drews-Botsch
C
.
Socioeconomic status, child enrichment factors, and cognitive performance among preschool-age children: results from the Follow-Up of Growth and Development Experiences study
.
Res Dev Disabil
.
2014
;
35
(
7
):
1789
1801
35
Johnson
DL
,
Swank
P
,
Howie
VM
,
Baldwin
CD
,
Owen
M
,
Luttman
D
.
Does HOME add to the prediction of child intelligence over and above SES?
J Genet Psychol
.
1993
;
154
(
1
):
33
40
36
Bell
MF
,
Bayliss
DM
,
Glauert
R
,
Harrison
A
,
Ohan
JL
.
Chronic illness and developmental vulnerability at school entry
.
Pediatrics
.
2016
;
137
(
5
):
e20152475
37
Herba
CM
,
Glover
V
,
Ramchandani
PG
,
Rondon
MB
.
Maternal depression and mental health in early childhood: an examination of underlying mechanisms in low-income and middle-income countries
.
Lancet Psychiatry
.
2016
;
3
(
10
):
983
992
38
Lozoff
B
.
Iron deficiency and child development
.
Food Nutr Bull
.
2007
;
28
(
4 suppl
):
S560
S571
39
Vaivada
T
,
Gaffey
MF
,
Bhutta
ZA
.
Promoting early child development with interventions in health and nutrition: a systematic review
.
Pediatrics
.
2017
;
140
(
2
):
e20164308
40
Bronfenbrenner
U
.
The Ecology of Human Development
.
Cambridge, MA
:
Harvard University Press
;
1979
41
Hamadani
JD
,
Tofail
F
,
Huda
SN
, et al
.
Cognitive deficit and poverty in the first 5 years of childhood in Bangladesh
.
Pediatrics
.
2014
;
134
(
4
). Available at: www.pediatrics.org/cgi/content/full/134/4/e1001
42
Black
MM
,
Dubowitz
H
,
Hutcheson
J
,
Berenson-Howard
J
,
Starr
RH
 Jr.
.
A randomized clinical trial of home intervention for children with failure to thrive
.
Pediatrics
.
1995
;
95
(
6
):
807
814
43
Chang
SM
,
Grantham-McGregor
SM
,
Powell
CA
, et al
.
Integrating a parenting intervention with routine primary health care: a cluster randomized trial
.
Pediatrics
.
2015
;
136
(
2
):
272
280
44
Singla
DR
,
Kumbakumba
E
,
Aboud
FE
.
Effects of a parenting intervention to address maternal psychological wellbeing and child development and growth in rural Uganda: a community-based, cluster randomised trial
.
Lancet Glob Health
.
2015
;
3
(
8
):
e458
e469
45
Britto
PR
,
Lye
SJ
,
Proulx
K
, et al;
Early Childhood Development Interventions Review Group, for the Lancet Early Childhood Development Series Steering Committee
.
Nurturing care: promoting early childhood development
.
Lancet
.
2017
;
389
(
10064
):
91
102

Competing Interests

POTENTIAL CONFLICT OF INTEREST: The authors have indicated they have no potential conflicts of interest to disclose.

FINANCIAL DISCLOSURE: The authors have indicated they have no financial relationships relevant to this article to disclose.

Supplementary data