Skip to main content
Download PDF
Download ePub

Impact of air pollution on stunting among children in Africa

Environmental Health volume 21, Article number: 128 (2022) Cite this article

Abstract

Background

Undernutrition is a global public health crisis, causing nearly half of deaths for children under age 5 years. Little is known regarding the impact of air pollution in-utero and early childhood on health outcomes related to undernutrition. The aim of our study is to evaluate the association of prenatal and early-life exposure to PM2.5 and child malnutrition as captured by the height-for-age z-score (HAZ), and stunting in 32 countries in Africa. We also evaluated critical windows of susceptibility during pregnancy to each environmental risk.

Methods

We linked nationally representative anthropometric data from 58 Demographic and Health Surveys (DHS) (n = 264,207 children < 5 years of age) with the average in-utero PM2.5 concentrations derived from satellite imagery. We then estimated associations between PM2.5 and stunting and HAZ after controlling for child, mother and household factors, and trends in time and seasonality.

Results

We observed lower HAZ and increased stunting with higher in-utero PM2.5 exposure, with statistically significant associations observed for stunting (OR: 1.016 (95% CI: 1.002, 1.030), for a 10 μg/m3 increase). The associations observed were robust to various model specifications. Wald tests revealed that sex, wealth quintile and urban/rural were not significant effect modifiers of these associations. When evaluating associations between trimester-specific PM2.5 levels, we observed that associations between PM2.5 and stunting was the largest.

Conclusions

This is one of the first studies for the African continent to investigate in-utero and early-life exposure to PM2.5 is an important marker of childhood undernutrition. Our results highlight that PM2.5 concentrations need to be urgently mitigated to help address undernutrition in children on the continent.

Peer Review reports

Introduction

Child undernutrition is a major public health crisis. Globally, nearly half of the deaths of children under 5 years of age are caused by poor nutrition. Undernourishment in the first few years of life has been shown to be associated with adverse cognitive health that manifests in lower educational achievement and lower economic productivity later in life, and physical health effects such as lower adult height and higher morbidity and mortality during childhood [2, 3, 7, 17, 24, 25].

The global burden of childhood undernutrition is uneven. According to a 2015 Millennium Development Goals (MDG) report sub-Saharan African accounts for one-third of all undernourished children globally with about 39% stunted, 10% wasted and 25% underweight children under-5 years of age [37]. In light of this, the United Nations Sustainable Development Goal (SDG 2) is to end all forms of malnutrition by 2030 (https://sdgs.un.org/goals/goal2, Last accessed December 16, 2021).

Yet, worsening air pollution can undermine future reductions in undernutrition through direct and indirect effects of health and livelihoods. Population-weighted annual average exposure to PM2.5 has increased over the last decade in many African countries [20]. This paper examines associations between exposure to concentrations of fine particulate matter (PM2.5), experienced in-utero in 32 African countries and malnutrition in early childhood. Stunting, derived from height-for-age measures, is widely accepted as an indicator of growth and nutrition in children < 5 years of age [8]. We also consider whether factors such as household wealth, urban or rural residence, or sex modify the relationship between PM2.5 and malnutrition. Finally, we attempt to identify the critical windows of susceptibility during pregnancy to each air pollution.

Review of existing evidence that has evaluated associations between PM2.5 and post-natal growth

Epidemiologic studies have consistently observed that high concentrations of fine particulate matter (PM2.5) has been linked with a wide range of health effects including premature mortality [4, 12], cardiovascular disease [11, 13], respiratory illnesses [14], and cognitive and developmental disorders [5], among others.

Most of the evidence of a link between air pollution and child growth has focused on adverse health outcomes such as early foetal loss [21], and preterm birth, small for gestational age, and low birthweight [9, 15, 30]. Very few studies have evaluated association between ambient air pollution and postnatal growth. One study that used DHS estimates for Bangladesh found that after controlling for other factors that contribute to child anthropometric failure, the relative risks of stunting in the second, third and fourth quartiles of exposure to in-utero PM2.5 were 1.074 (95% CI: 1.014, 1.138), 1.150 (95% CI: 1.069, 1.237) and 1.132 (95% CI: 1.031, 1.243), respectively [16]. Another DHS study for India found that after controlling for potential confounders, a 100 μg/m3 increase in PM2.5 in the month of birth was associated with a 0.05 (95% CI: 0.01–0.09) standard deviation reduction in child height [29]. A final study found that mean in-utero exposure to PM2.5 in Indonesia during the 1997 forest fires was associated with a half standard-deviation (0.41) decrease in HAZ at age 17 [31].

Potential mechanisms underlying the relationship between in-utero and early-life exposure to PM2.5 and anthropometric status

Sinharoy et al. [27], found that studies investigating air pollution and intrauterine growth impairment focus on effects at the cellular level. Specifically, these studies hypothesize that exposure to air pollution during pregnancy can cause oxidative stress, which in turn causes inflammation, and potentially poor foetal growth. In addition, they suggest that exposure to particulate matter in-utero modulates DNA methylation, affecting foetal growth. Further, exposure to poor air quality in-utero and in early-life can affect immune ontogeny, which can lead to growth failure in multiple ways. Moreover, children’s lungs are not fully formed until approximately 6 years of age. Exposure to air pollution in young children could affect the formation of the respiratory system, which could in turn affect growth. Finally, the authors posit that air pollution might be responsible for some prenatal vitamin D deficiency, with implications for immune function and bone metabolism. More research is needed to explore these relationships and the critical window of in-utero and early-life exposure to PM2.5 that most impacts postnatal growth.

Methods

Study population

We drew data from 58 Standard DHSs conducted in 32 countries (Burkina Faso (BF), Benin (BJ), Burundi (BU), Cameroon (CM), Chad (TD), Comoros (KM), Democratic Republic of Congo (CD), Côte d’Ivoire (CI), Eswatini (SZ), Ethiopia (ET), Gabon (GA), Ghana (GH), Guinea (GN), Kenya (KE), Lesotho (LS), Liberia (LB), Madagascar (MD), Malawi (MW), Mali (ML), Mozambique (MZ), Niger (NI), Namibia (NM), Nigeria (NG), Rwanda (RW), Senegal (SN), Sierra Leone (SL), South Africa (ZA), Tanzania (TZ), Togo (TG), Uganda (UG), Zambia (ZM), Zimbabwe (ZW) between 2005 and 2019, where the GPS data were available. We restricted our analysis to DHS surveys for African countries from after 2005, as earlier surveys tend to avoid inconsistencies in measurements, data collection and data reporting.

DHS (https://www.dhsprogram.com/, Last accessed October 18, 2022) are nationally representative household surveys that collect detailed nutrition and health information on children, their parents, and households using a multistage, stratified sampling design. The first stage involves the division of each country into geographic areas. Within these subnational regions, populations are divided into urban and rural areas. These primary sampling units or clusters are selected with probability proportional to the contribution of that cluster’s population to the total population. In the second stage of sampling, all households within the cluster are listed, and an average of 25 houses are randomly selected for an interview with equal-probability systematic sampling.

The eligibility criteria for our analytic sample included survey respondents with:

1) geographic coordinates for each cluster, and 2) complete measures on height-for-age, and 3) same locations for conception and survey interview (when data was collected), 4) estimates for PM2.5, and climate data (temperature and precipitation) during the prenatal period and early-life, 5) non-missing information on covariates. Table S1 in Supplementary Information reports the number of children who met these criteria in each of the DHS surveys considered in this study. Our final sample had 264,207 children from 32 countries in Africa (Table 1). Due to the lack of data on important covariates (such as mother’s height, or the outcome variables themselves, we did not include data from other DHS surveys conducted during this time period. For example, the 2003 DHS survey in Ghana did not have information on the anthropometric outcomes considered in this analysis and was therefore, excluded.

Table 1 Descriptive Statistics of the DHS variables used as controls in our analysis
Full size table

Our study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guidelines.

Stunting

Environmental factors explain more variation in height for children under 5 y of age than ethnic differences. Consequently, child height is a widely accepted indicator of child nutrition. We built our model using the HAZ for children under 5 y of age as the outcome variable, a standardized measure of child heights and a common indicator of stunting. The height-for-age z-score (HAZ) was calculated by comparing the child’s measurements with the median value in the reference population of the National Centre for Health Statistics International Growth Reference. Stunting was defined as a HAZ < -2 standard deviations from the median (WHO Expert Committee on Physical Status: the Use and Interpretation of Anthropometry [34]. Stunting is affected by chronic long-term undernutrition and reflects exposures in-utero, unlike other anthropometric outcomes like underweight and wasting.

Environmental exposures

PM2.5 concentrations

The main exposure variable in this study was long-term ambient PM2.5 concentrations. Because most of Africa lacks surface PM2.5 monitoring sites at the spatial resolution required for the study, we used satellite-derived monthly PM2.5 estimates at a 0.01° × 0.01° (~ 1 km × 1 km at the equator) resolution derived for Africa (Fig. 1) [32]. Satellite aerosol optical depths (AODs) were combined from multiple satellite products: MISR, MODIS Dark Target, MODIS and SeaWiFS Deep Blue, and MODIS MAIAC with simulation-based results based on their relative uncertainties. These AODs were related to near-surface monthly PM2.5 concentrations using the ratio of simulated AOD and PM2.5 from the GEOS-Chem model. The modeled estimates of PM2.5 used in this study agree well with concentrations from global ground-based monitors (R2 = 0.81). However, we note that due to the lack of ground-based monitors on the African continent, more validation of the modeled estimates in Africa is needed.

Fig. 1
figure 1

Monthly averaged PM2.5 (μg/m3) for each cluster in our study for January, 2018 classified into quantiles. The countries we focus on in this study are highlighted

Full size image

Each child was assigned the average ambient PM2.5 exposure estimates of members of his/her household cluster, based on GPS coordinates of different household clusters available from the DHS. To maintain the privacy of these respondents, the clusters are randomly displaced by a maximum of 5 km from the true location for rural areas and 2 km in urban areas, with a further 5% of all clusters displaced by 10 km. We extracted the mean PM2.5 values from a 2 km buffer around urban clusters and a 5 km buffer around rural clusters (Fig. 2).

Fig. 2
figure 2

This figure conceptually demonstrates how environmental variables such as PM2.5 were joined with DHS sampling clusters. The blue point in the middle represents the sampling cluster. The red circle represents a 2 km urban area buffer around the cluster (note this is 5 km in rural areas). All environmental variables were averaged over the grid cells whose centroids fall within the red circle, which in this figure include cells E and F

Full size image

We calculated the mean PM2.5 exposure corresponding to the 9 months in-utero period (not including month of birth) for the respondent as our main exposure of interest. We also calculated PM2.5 exposures for each trimester for each child in the dataset, as well as average PM2.5 concentrations experienced in early-life following birth from the month of birth to the month of the DHS interview, which we term as early-life PM2.5. The mean age in children in months at the time of the DHS interview was 28.4 (minimum = 0, first quartile = 13, median = 28, third quartile = 43, maximum = 59).

Temperature and precipitation

Monthly average temperature and precipitation data during the study period were obtained from the University of East Anglia Climate Research Unit’s Time Series (CRUTS) version 4.05 dataset [19]. CRUTS is a global dataset of monthly weather conditions. The CRUTS data are constructed at a 0.5o resolution and are based on the statistical interpolation of data from over 4000 weather stations, and are considered highly accurate [6, 18, 26, 36]. Mean cluster-level temperature and precipitation were estimated using a spatially-weighted average of the grid cells in the respective buffer around each cluster, in a similar manner to assigning cluster-specific PM2.5 exposures. As with PM2.5, we also evaluated the mean temperature and precipitation corresponding to the 9 months in-utero, as well as for each trimester, and early-life for every child in our dataset.

Statistical approach

Characteristics of the 264,207 children in our study were explored using descriptive statistics. We examined correlations between each of the environmental exposures analyzed using Pearson correlations.

We applied a linear model with HAZ as a continuous variable and a logistic model comparing stunted and non-stunted children to evaluate associations with in-utero PM2.5. These methodologies have been applied in previous air pollution research [16, 29].

We included the following factors as controls in our model:

  1. 1)

    Child-level characteristics: sex fixed effects, linear and squared terms for the child’s age in months to account for the nonlinear relationship between age and child growth in early childhood [33], birth order, whether the birth was singleton.

  2. 2)

    Maternal characteristics: education level with categories: 1) no education, 2) primary school education, 3) secondary school education, 4) higher); height (cm) and BMI (kg/m2), marital status < 18 years of age.

  3. 3)

    Household-level variables: if an improved sanitation facility was present, if a source of safe water was available, if solid fuels were used in the household for cooking, if the household was in an urban or rural environment, and household wealth quintile. The wealth quintile for each household is provided by DHS based on a composite measure of a household’s living standard.

  4. 4)

    Temporal and country fixed effects: birth-year and country-month fixed effects to control for time-varying trends

  5. 5)

    Spatial fixed effects: cluster fixed effects (clusters correspond to villages in rural areas and census enumeration blocks in urban areas) to control for time-invariant cross-region differences (for example higher or lower stunting rates) (Table 1).

  6. 6)

    Meteorological exposures: In-utero and early-life exposure to temperature and precipitation.

This statistical approach was chosen to address sources of confounding. We added the potential confounders in stages to evaluate the robustness of the association between air pollution and HAZ to model specification. We first estimated the model without cluster-level fixed effects. We then added child, mother and household characteristics and meteorological exposures. We report the change in the different outcome parameters for a 10 μg/m3 increase in in-utero PM2.5 concentrations.

The Pearson correlation coefficient between early-life PM2.5 concentrations and in-utero PM2.5 concentrations are 0.84 (Fig. 3). In sensitivity analyses, we used early-life PM2.5 instead of in-utero PM2.5 levels as our main exposure of interest. We also applied a model that mutually adjusted for both in-utero PM2.5 exposure as well as early-life PM2.5 concentrations.

Fig. 3
figure 3

Correlations between in-utero, trimester-specific, and early-life PM2.5, temperature and precipitation for the children in our study

Full size image

In further sensitivity analyses, we used fixed effects at the household-level instead of at the cluster-level in our fully-adjusted models. In this design, the effects of in-utero PM2.5 on HAZ and stunting derive from determining if a child born in a given household during a high period of PM2.5 levels is less likely to survive compared to a child born in the same household during a period of low PM2.5 concentrations. This design addresses the concern that households with different levels of pollution exposure could be inherently different in unobservable ways.

We tested univariate effect modification of the association by wealth quintile, urban/rural and sex by adding into the model a cross-product term between the modifier and the in-utero PM2.5 using Wald tests. We evaluate subgroup-specific associations between in-utero PM2.5 and the child growth metrics by including interaction terms between: wealth quintile, urban/rural and sex and in-utero PM2.5.

As an additional analysis, we estimated the country-specific relationships between exposure to in-utero PM2.5 and HAZ and stunting for each country (except for South Africa where we only have 334 children) in our sample. We then conducted pooled-analyses to combine the country-specific estimates. Estimates are presented for a 10 μg/m3 increase in PM2.5 to the pooled global estimate and we present these results using a forest plot. In our meta-analysis, the relative contribution (weight) of each country-specific estimate to the pooled estimate is calculated along with the 95% CI from each country-specific analysis. We included random effects in accordance with the DerSimonian and Laird method [10], and we conducted a test of whether the overall pooled estimate is equal to the null as well as a test for heterogeneity, i.e. whether the country-specific estimates are the same. This heterogeneity is quantified using the I-squared measure.

Finally, we developed separate models considering trimester exposures, instead of in-utero PM2.5 levels. We controlled for the corresponding trimester temperature and precipitation, and early-life temperature and precipitation in these models. However, because of seasonality bias described in Wilson et al. [35], ,the estimates derived for trimester-specific associations could be biased. Pronounced seasonal variations result in correlations of PM2.5 levels between exposure windows (Pearson correlation coefficients range between 0.25 to 0.61) (Fig. 3).

Exposures in other time windows are analogous to confounders, with the possibility that PM2.5 in one time period may have an independent association with the outcome, and through its correlations with exposures in other time periods, may confound the observed association. Therefore, we also used the residuals of regressing exposure during each trimester against the other two remaining trimester-specific exposures, as our main exposure of interest. The residuals obtained represent exposure to PM2.5 in the given trimester of interest after adjusting for other trimester-specific exposures. This approach avoids covariance among variables representing trimester exposures [1].

We used cluster-adjusted standard errors when reporting results from our models to reflect the assumption that metrics of child growth in the same cluster are not independent. DHS data include sampling weights, to enable generation of representative estimates of the population of the respective country for children under five. Sampling weights are not appropriate for estimating associations [28] as the impact of each environmental exposure on the different administrative units are likely not homogeneous. Sample weights that do not account for such interactions may yield biased associations and so we do not include them in our main analysis.

All statistical tests were 2-tailed, and p-values < .05 were considered statistically significant. All analyses were done in R. The main analysis was carried out using the fixest package.

Results

Information about stunting, PM2.5, as well as other covariates were available for 264,207 children across the 58 surveys considered in the study conducted in 32 African countries (Table 1).

Across the entire sample, the average in-utero PM2.5 (min = 4.4 μg/m3, mean = 36.4 μg/m3, median = 21.1 μg/m3, max values = 259.4 μg/m3), and early-life PM2.5 (min = 3.1 μg/m3, mean = 35.7 μg/m3, median = 31.0 μg/m3, max = 233.7 μg/m3) varied considerably. Our initial descriptive analysis suggests that stunted children were generally poorer, had mothers of shorter stature, and mainly belonged to rural households, and experienced higher PM2.5 concentrations than children who were not stunted (Table 1). The PM2.5 exposures considered in this analysis in different windows ang from moderately to highly correlated (Fig. 3).

Table S2 shows results for selected variables from running the linear and logistic models including all variables except for air pollution. Healthier indicators, higher HAZ scores and decreased stunting, were observed among females in comparison to males, for singleton births, for subsequent children after the first birth, for higher educated mothers, for taller mothers, for mothers with a higher BMI, in wealthier households, and in houses with access to improved sanitation.

In Table S3, we present the association and 95% Confidence Interval (CI) between in-utero PM2.5 and HAZ, as well as the Odds Ratios (ORs) and 95% CI corresponding to the association between in-utero PM2.5 and stunting, adjusting for different sets of covariates. The associations are reported for a 10 μg/m3 increase in PM2.5. In the simplest models that only adjust for sex and age, we observed significant associations between in-utero PM2.5 and the outcomes (Model 1). Further, adjusting for temporal fixed effects (Model 2), strengthened the associations between in-utero PM2.5 and stunting and HAZ. However, including DHS cluster fixed effects (Model 3) substantially attenuated the estimated associations. This indicates that seasonality and spatial factors are important confounders of the impact of the different environmental risks on nutrition. By including country-month, year and cluster fixed effects, we adopt a conservative approach to control for residual spatial and temporal confounding in our dataset. The estimates appear to be robust (the change in associations was < 5%) to the inclusion of child-level and maternal (Model 4), household (Model 5) and meteorological covariates (Model 6) (Table S3).

Although we did not observe a significant association between in-utero exposure to PM2.5 and HAZ in the fully-adjusted Model 6 displayed in Table 2 (Main Model), the association demonstrated the same general trend of a negative association between in-utero PM2.5 and HAZ: -0.003 (95% CI: − 0.012, 0.006). We also observed that in-utero PM2.5 was significantly associated with increased stunting OR: 1.017 (95% CI: 1.003, 1.032). The associations observed between in-utero PM2.5 and HAZ: -0.005 (95% CI: − 0.014, 0.004) and stunting: OR: 1.017 (95% CI: 1.003, 1.032) remained robust to the inclusion of early-life PM2.5 levels in a sensitivity analysis (Table S4).

Table 2 Associations (95% Confidence Intervals) between in-utero PM2.5 and HAZ, and the Odds Ratios (95% Confidence Intervals) corresponding to associations between in-utero PM2.5 exposure and stunting for an increase of 10 μg/m3 in PM2.5 . Standard errors presented are clustered at the cluster-level
Full size table

In supplementary analyses, we observed that when we considered early-life PM2.5 exposure instead of in-utero exposure as our main exposure of interest, we observed significant associations with HAZ: -0.033 (95% CI: − 0.059, − 0.008), and indications of a general trend of a positive association with stunting: 1.024 (95% CI: 0.991, 1.059) (Table S5). However, given the strong correlation (0.84, Fig. 3) between in-utero and early-life PM2.5 levels, we retain in-utero PM2.5 concentrations as our main exposure of interest in this analysis.

In further sensitivity analyses, using household-level fixed effects instead of cluster-level fixed effects (to remove potential omitted variable biases), we observed similar results. The association between in-utero exposure to PM2.5 and HAZ was − 0.006 (95% CI: − 0.018, 0.007), and with stunting was: OR: 1.049 (95% CI: 1.009, 1.092) (Table S6). These associations were also robust to the inclusion of early-life PM2.5 exposure in models using household-level fixed effects (Table S7). However, only 69,226 households (with a total of 155,317 children) had more than one child. The associations reported in Tables S6 and S7 were for this subset, alone.

Wald tests revealed that child sex, urban/rural and wealth quintile were not significant modifiers of the association between in-utero PM2.5 and HAZ or stunting. Subgroup-specific associations are displayed in Table 3. In all subgroups, the confidence intervals overlap widely.

Table 3 Subgroup specific associations (95% CI) (for HAZ), and OR (95% CI) (for stunting) between the outcomes of interest and in-utero PM2.5 derived by including interaction terms between sex, urban/rural, wealth quintile with in-utero PM2.5 in fully-adjusted models (Table 2)
Full size table

We evaluated the relationship between in-utero PM2.5 and HAZ and stunting in our study using fully adjusted models for each country in our study (except for South Africa because of the small number of children in the dataset) and then conducted a pooled analysis to compare country-specific results against the summary estimate of the meta-analysis. Forest plots from country-specific regression and the pooled analysis for the outcomes HAZ and stunting are presented in Fig. 4A and B, respectively. The I-squared from these analyses (24 and 37%) suggest that the association for all outcomes within individual countries exhibits moderate variation. Given the large CIs of the associations observed, the results from the pooled analyses suggest that individual country-level studies using DHS data may not have sufficient power to detect a relationship due to the small sample size within each country dataset.

Fig. 4
figure 4

Forest plots from country-specific regression and the pooled analysis for the outcomes HAZ and stunting

Full size image

When we evaluated the impact of trimester-specific PM2.5 on HAZ and stunting, we observed that exposure to PM2.5 in trimester 1, was most strongly associated with increased stunting Model 1, OR: 1.007 (95% CI: 1.002, 1.012). When we accounted for correlations between exposures in different trimesters, by using adjusted trimester exposures as described in Methods, we again fund that PM2.5 in trimester 1, was still most strongly associated with stunting OR: 1.006 (95% CI: 1.002, 1.011) (Table 4).

Table 4 Associations between the outcomes (HAZ and stunting) and PM2.5 experienced during different trimesters from fully-adjusted models
Full size table

Discussion

In this study we evaluated associations between exposure to ambient in-utero PM2.5 and stunting and HAZ using data from 58 DHS surveys conducted in 32 African countries between 2005 and 2019.

Using fully adjusted models, we observed a lower HAZ and increased stunting with higher in-utero PM2.5 concentrations (Table 2), with statistically significant associations between in-utero PM2.5 and stunting OR: 1.016 (95% CI: 1.002, 1.030). The associations were robust to different model specifications (Tables S3, S4), and were also robust when evaluating the impact of PM2.5 on children born in the same household (Tables S6, S7).

Child sex, wealth quintile, urban/rural were not significant effect modifiers of the association between in-utero PM2.5 and HAZ and stunting. Confidence intervals of the associations derived for subpopulations overlapped widely (Table 3).

We observed moderate heterogeneity in the association between PM2.5 and stunting in our country-specific analyses (Fig. 4), likely due to different baseline policies that impact stunting such as access to nutrition (food prices, food availability), health care systems, etc., as well as differences in activity patterns occupational exposures, other population characteristics, built environment, and particulate matter chemical composition. Other studies have demonstrated differences in the health associations for particulate matter in different locations [22, 23]. However, due to the limited number of observations within each country, we do not have the statistical power to detect robust relationships between PM2.5 and stunting or HAZ.

When evaluating trimester-specific associations, we also observed the highest associations between PM2.5 in trimester 1 (but not trimesters 2 and 3) and stunting, even after adjusting for exposures in other time windows and other covariates (Table 4). This suggests that PM2.5 experienced in early pregnancy likely has the greatest impact on stunting. More work is needed to evaluate critical windows of exposure to PM2.5 for HAZ and stunting.

Our study has several limitations. First, the dataset does not include a measure of gestational age, and therefore, assume that each child in our sample is carried to term, which is reflected in our calculation of in-utero exposures. Second, our results may suffer from residual confounding from omitted variables that are correlated both with the environmental exposures and the outcomes considered in this study. Third, the exposure of interest in our analysis is PM2.5, which is based on modelled data rather than monitoring data. Although this dataset has been validated using the global distribution of ground-based monitors (R2 = 0.81), there is a lack of surface-monitors in Africa and more research is required to evaluate the modeled estimates in the countries considered in this study. More work is also needed to examine the sensitivity of the calculated health impact of the different environmental exposures to the exposure product used (for example in deSouza et al. [12],). Fourth, our sample pools together data on all births in the 5 years preceding the survey date for women of reproductive age (15–49 years) who lived in the sampled households. Since these data on births are reported by mothers, however, we may be missing data on children in the households whose mothers have died or who were not present at the time of the survey. Sixth, research has shown that anthropometry is a complex indicator that captures genetic, environmental, behavioral factors, as well as exposure to disease. The health outcome: anthropometric failure must be complemented by other diet and food based measures in future work to measure the impact of PM2.5 on undernutrition. Seventh, the study cannot disentangle the impacts of different particulate matter chemical composition and sources, which are likely to vary across the study area due to different sources, meteorology, etc. Eighth, the population characteristics considered are not the same across all the study locations, and the impact of PM2.5 is likely different in different locations. Finally, while this is one of the first (perhaps the first) to investigate this topic for Africa, even so large parts of the continent are not included and more work is needed on these other regions. Nevertheless, our study provides important insights into the relationship between air pollution and stunting, and demonstrates the importance of reducing PM2.5 concentrations in African countries to protect children.

References

  1. Bell ML, Ebisu K, Belanger K. Ambient air pollution and low birth weight in Connecticut and Massachusetts. Environ Health Perspect. 2007;115(7):1118–24. https://doi.org/10.1289/ehp.9759.

    Article  CAS  Google Scholar 

  2. Black RE, Victora CG, Walker SP, Bhutta ZA, Christian P, de Onis M, et al. Maternal and child undernutrition and overweight in low-income and middle-income countries. Lancet. 2013;382(9890):427–51. https://doi.org/10.1016/S0140-6736(13)60937-X.

    Article  Google Scholar 

  3. Bryce J, Boschi-Pinto C, Shibuya K, Black RE. WHO estimates of the causes of death in children. Lancet. 2005;365(9465):1147–52. https://doi.org/10.1016/S0140-6736(05)71877-8.

    Article  Google Scholar 

  4. Burnett R, Chen H, Szyszkowicz M, Fann N, Hubbell B, Pope CA, et al. Global estimates of mortality associated with long-term exposure to outdoor fine particulate matter. Proc Natl Acad Sci. 2018;115(38):9592–7. https://doi.org/10.1073/pnas.1803222115.

    Article  CAS  Google Scholar 

  5. Calderón-Garcidueñas L, Mora-Tiscareño A, Ontiveros E, Gómez-Garza G, Barragán-Mejía G, Broadway J, et al. Air pollution, cognitive deficits and brain abnormalities: a pilot study with children and dogs. Brain Cogn. 2008;68(2):117–27. https://doi.org/10.1016/j.bandc.2008.04.008.

    Article  Google Scholar 

  6. Call M, Gray C, Jagger P. Smallholder responses to climate anomalies in rural Uganda. World Dev. 2019;115:132–44. https://doi.org/10.1016/j.worlddev.2018.11.009.

    Article  Google Scholar 

  7. Caulfield LE, de Onis M, Blössner M, Black RE. Undernutrition as an underlying cause of child deaths associated with diarrhea, pneumonia, malaria, and measles. Am J Clin Nutr. 2004;80(1):193–8. https://doi.org/10.1093/ajcn/80.1.193.

    Article  CAS  Google Scholar 

  8. Corsi DJ, Subramanyam MA, Subramanian S. Commentary: measuring nutritional status of children. Int J Epidemiol. 2011;40(4):1030–6. https://doi.org/10.1093/ije/dyr108.

    Article  Google Scholar 

  9. Currie J, Neidell M. Air pollution and infant health: what can we learn from California’s recent experience? Q J Econ. 2005;120(3):1003–30. https://doi.org/10.1093/qje/120.3.1003.

    Article  Google Scholar 

  10. DerSimonian R, Laird N. Meta-analysis in clinical trials. Control Clin Trials. 1986;7(3):177–88. https://doi.org/10.1016/0197-2456(86)90046-2.

    Article  CAS  Google Scholar 

  11. deSouza P, Braun D, Parks RM, Schwartz J, Dominici F, Kioumourtzoglou M-A. Nationwide study of short-term exposure to fine particulate matter and cardiovascular hospitalizations among Medicaid enrollees. Epidemiology. 2020;32(1):6–13. https://doi.org/10.1097/EDE.0000000000001265.

    Article  Google Scholar 

  12. deSouza PN, Dey S, Mwenda KM, Kim R, Subramanian SV, Kinney PL. Robust relationship between ambient air pollution and infant mortality in India. Sci Total Environ. 2022;815:152755. https://doi.org/10.1016/j.scitotenv.2021.152755.

    Article  CAS  Google Scholar 

  13. Dominici F, Peng RD, Bell ML, Pham L, McDermott A, Zeger SL, et al. Fine particulate air pollution and hospital admission for cardiovascular and respiratory diseases. JAMA. 2006;295(10):1127–34. https://doi.org/10.1001/jama.295.10.1127.

    Article  CAS  Google Scholar 

  14. Ferkol T, Schraufnagel D. The global burden of respiratory disease. Ann Am Thoracic Soc. 2014;11(3):404–6. https://doi.org/10.1513/AnnalsATS.201311-405PS.

    Article  Google Scholar 

  15. Fleischer NL, Merialdi M, van Don elaar A, Vadillo-Ortega F, Martin RV, Betran AP, et al. Outdoor air pollution, preterm birth, and low birth weight: analysis of the World Health Organization global survey on maternal and perinatal health. Environ Health Perspect. 2014;122(4):425–30. https://doi.org/10.1289/ehp.1306837.

    Article  CAS  Google Scholar 

  16. Goyal N, Canning D. Exposure to ambient fine particulate air pollution in utero as a risk factor for child stunting in Bangladesh. Int J Environ Res Public Health. 2018;15(1):Article 1. https://doi.org/10.3390/ijerph15010022.

    Article  Google Scholar 

  17. Grantham-McGregor S, Cheung YB, Cueto S, Glewwe P, Richter L, Strupp B. Developmental potential in the first 5 years for children in developing countries. Lancet. 2007;369(9555):60–70. https://doi.org/10.1016/S0140-6736(07)60032-4.

    Article  Google Scholar 

  18. Gray C, Wise E. Country-specific effects of climate variability on human migration. Clim Chang. 2016;135(3):555–68. https://doi.org/10.1007/s10584-015-1592-y.

    Article  Google Scholar 

  19. Harris I, Osborn TJ, Jones P, Lister D. Version 4 of the CRU TS monthly high-resolution gridded multivariate climate dataset. Scientific Data. 2020;7(1):Article 1. https://doi.org/10.1038/s41597-020-0453-3.

    Article  Google Scholar 

  20. Health Effects Institute. State of global air 2019. MA: Health Effects Institute Boston; 2019.

    Google Scholar 

  21. Hou HY, Wang D, Zou XP, Yang ZH, Li T-C, Chen YQ. Does ambient air pollutants increase the risk of fetal loss? A case–control study. Arch Gynecol Obstet. 2014;289(2):285–91. https://doi.org/10.1007/s00404-013-2962-1.

    Article  CAS  Google Scholar 

  22. Kioumourtzoglou MA, Schwartz J, James P, Dominici F, Zanobetti A. PM2.5 and mortality in 207 US cities: modification by temperature and city characteristics. Epidemiology. 2016;27(2):221.

    Google Scholar 

  23. Masselot P, Sera F, Schneider R, Kan H, Lavigne É, Stafoggia M, et al. Differential mortality risks associated with PM2.5 components: a multi-country, multi-city study. Epidemiology. 2021;33(2):167–75.

    Article  Google Scholar 

  24. Pradhan M, Sahn DE, Younger SD. Decomposing world health inequality. J Health Econ. 2003;22(2):271–93. https://doi.org/10.1016/S0167-6296(02)00123-6.

    Article  Google Scholar 

  25. Prendergast AJ, Humphrey JH. The stunting syndrome in developing countries. Paediatr Int Child Health. 2014;34(4):250–65. https://doi.org/10.1179/2046905514Y.0000000158.

    Article  Google Scholar 

  26. Randell H, Gray C. Climate change and educational attainment in the global tropics. Proc Natl Acad Sci. 2019;116(18):8840–5. https://doi.org/10.1073/pnas.1817480116.

    Article  CAS  Google Scholar 

  27. Sinharoy SS, Clasen T, Martorell R. Air pollution and stunting: a missing link? Lancet Glob Health. 2020;8(4):e472–5. https://doi.org/10.1016/S2214-109X(20)30063-2.

    Article  Google Scholar 

  28. Solon G, Haider SJ, Wooldridge JM. What are we weighting for? J Hum Resour. 2015;50(2):301–16. https://doi.org/10.3368/jhr.50.2.301.

    Article  Google Scholar 

  29. Spears D, Dey S, Chowdhury S, Scovronick N, Vyas S, Apte J. The association of early-life exposure to ambient PM2.5 and later-childhood height-for-age in India: an observational study. Environ Health. 2019;18(1):62. https://doi.org/10.1186/s12940-019-0501-7.

    Article  CAS  Google Scholar 

  30. Stieb DM, Chen L, Eshoul M, Judek S. Ambient air pollution, birth weight and preterm birth: a systematic review and meta-analysis. Environ Res. 2012;117:100–11. https://doi.org/10.1016/j.envres.2012.05.007.

    Article  CAS  Google Scholar 

  31. Tan-Soo J-S, Pattanayak SK. Seeking natural capital projects: Forest fires, haze, and early-life exposure in Indonesia. Proc Natl Acad Sci. 2019;116(12):5239–45. https://doi.org/10.1073/pnas.1802876116.

    Article  CAS  Google Scholar 

  32. van Donkelaar A, Martin RV, Li C, Burnett RT. Regional estimates of chemical composition of fine particulate matter using a combined geoscience-statistical method with information from satellites, models, and monitors. Environ Sci Technol. 2019;53(5):2595–611. https://doi.org/10.1021/acs.est.8b06392.

    Article  CAS  Google Scholar 

  33. Victora CG. The association between wasting and stunting: an international perspective. J Nutr. 1992;122(5):1105–10. https://doi.org/10.1093/jn/122.5.1105.

    Article  CAS  Google Scholar 

  34. WHO Expert Committee on Physical Status: the Use and Interpretation of Anthropometry (1993: Geneva, S., & Organization, W. H. (1995). Physical status: The use of and interpretation of anthropometry, report of a WHO expert committee. Geneva: World Health Organization. https://apps.who.int/iris/handle/10665/37003.

  35. Wilson JG, Kingham S, Pearce J, Sturman AP. A review of intraurban variations in particulate air pollution: implications for epidemiological research. Atmos Environ. 2005;39(34):6444–62. https://doi.org/10.1016/j.atmosenv.2005.07.030.

    Article  CAS  Google Scholar 

  36. Zhang Q, Körnich H, Holmgren K. How well do reanalyses represent the southern African precipitation? Clim Dyn. 2013;40(3):951–62. https://doi.org/10.1007/s00382-012-1423-z.

    Article  Google Scholar 

  37. World Health Organization. World Health Statistics Report. Geneva: World Health Organization; 2015.

Download references

Acknowledgements

KM acknowledges support from the Population Studies and Training Center at Brown University, which receives funding from the NIH (P2C HD041020).

Author information

Authors and Affiliations

  1. Department of Urban and Regional Planning, University of Colorado Denver, Denver, CO, USA

    Priyanka N. deSouza

  2. CU Population Center, University of Colorado Boulder, Boulder, CO, USA

    Priyanka N. deSouza

  3. Department of Energy, Environmental, and Chemical Engineering, Washington University in St. Louis, St. Louis, MO, 63130, USA

    Melanie Hammer

  4. Department of Physics and Atmospheric Science, Dalhousie University, Halifax, NS, Canada

    Melanie Hammer

  5. Department of Geography and Environmental Sciences, University of Colorado Denver, Denver, CO, USA

    Peter Anthamatten

  6. Boston University School of Public Health, Boston, MA, USA

    Patrick L. Kinney

  7. Division of Health Policy & Management, College of Health Science, Korea University, 145 Anam-ro, Seongbuk-gu, Seoul, 02841, South Korea

    Rockli Kim

  8. Interdisciplinary Program in Precision Public Health, Department of Public Health Sciences, Graduate School of Korea University, Seoul, 02841, South Korea

    Rockli Kim

  9. Harvard Center for Population and Development Studies, Bow Street, Cambridge, MA, 02138, USA

    Rockli Kim & S. V. Subramanian

  10. Department of Social and Behavioral Sciences, Harvard T.H. Chan School of Public Health, 677 Huntington Avenue, Boston, MA, 02115, USA

    S. V. Subramanian

  11. School of the Environment, Yale University, New Haven, CT, USA

    Michelle L. Bell

  12. Spatial Structures in the Social Sciences, Brown University, Providence, RI, USA

    Kevin M. Mwenda

  13. Population Studies and Training Center, Brown University, Providence, RI, USA

    Kevin M. Mwenda

Authors
  1. Priyanka N. deSouza
    View author publications

    You can also search for this author in PubMed Google Scholar

  2. Melanie Hammer
    View author publications

    You can also search for this author in PubMed Google Scholar

  3. Peter Anthamatten
    View author publications

    You can also search for this author in PubMed Google Scholar

  4. Patrick L. Kinney
    View author publications

    You can also search for this author in PubMed Google Scholar

  5. Rockli Kim
    View author publications

    You can also search for this author in PubMed Google Scholar

  6. S. V. Subramanian
    View author publications

    You can also search for this author in PubMed Google Scholar

  7. Michelle L. Bell
    View author publications

    You can also search for this author in PubMed Google Scholar

  8. Kevin M. Mwenda
    View author publications

    You can also search for this author in PubMed Google Scholar

Contributions

PD, SVS, RK, KM conceptualized the study. PD curated the data, conducted the formal analysis, and wrote the first draft. All authors developed the methodology and revised and edited the manuscript.

Corresponding author

Correspondence to Priyanka N. deSouza.

Ethics declarations

Competing interests

The authors declare that they have no competing interests.

Additional information

Publisher’s Note

Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.

Supplementary Information

Additional file 1: Table S1.

Study sample with filtered observations by country. Table S2. Associations (95% Confidence Intervals) between in-utero PM2.5 and HAZ, and the Odds Ratios (95% Confidence Intervals) corresponding to associations between in-utero PM2.5 exposure and stunting for an increase of 10 μg/m3 in PM2.5. Standard errors presented are clustered at the cluster-level. Table S3. Associations (95% Confidence Intervals) between in-utero PM2.5 and HAZ, and the Odds Ratios (95% Confidence Intervals) corresponding to associations between in-utero PM2.5 and stunting for an increase of 10 μg/m3 in PM2.5. Standard errors presented are clustered at the cluster-level. Note that sample sizes vary across Models because some fixed effects categories lack within-category variation in the independent variable (resulting in that category being dropped). The main model is highlighted. Table S4. Associations (95% Confidence Intervals) between in-utero and early-life PM2.5 and HAZ, and the Odds Ratios (95% Confidence Intervals) corresponding to associations between in-utero and early-life PM2.5 and stunting for an increase of 10 μg/m3 in PM2.5. Standard errors presented are clustered at the cluster-level. Note that sample sizes vary across Models because some fixed effects categories lack within-category variation in the independent variable (resulting in that category being dropped). . The main model is highlighted. Table S5. Associations (95% Confidence Intervals) between early-life PM2.5 and HAZ, and the Odds Ratios (95% Confidence Intervals) corresponding to associations between early-life PM2.5 and stunting for an increase of 10 μg/m3 in PM2.5. Standard errors presented are clustered at the cluster-level. Note that sample sizes vary across Models because some fixed effects categories lack within-category variation in the independent variable (resulting in that category being dropped). . The main model is highlighted. Table S6. Associations (95% Confidence Intervals) between in-utero PM2.5 and HAZ, and the Odds Ratios (95% Confidence Intervals) corresponding to associations between in-utero PM2.5 and stunting for an increase of 10 μg/m3 in PM2.5 derived from fully-adjusted models (Model 6 in Table S3), using household-fixed effects instead of cluster-specific fixed effects. Table S7. Associations (95% Confidence Intervals) between in-utero and early-life PM2.5 and HAZ, and stunting for an increase of 10 μg/m3 in PM2.5 derived from fully-adjusted models (Model 6 in Table S4), using household-fixed effects instead of cluster-specific fixed effects.

Rights and permissions

Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data.

Reprints and permissions

About this article

Check for updates. Verify currency and authenticity via CrossMark

Cite this article

deSouza, P.N., Hammer, M., Anthamatten, P. et al. Impact of air pollution on stunting among children in Africa. Environ Health 21, 128 (2022). https://doi.org/10.1186/s12940-022-00943-y

Download citation

  • Received:

  • Accepted:

  • Published:

  • DOI: https://doi.org/10.1186/s12940-022-00943-y

Environmental Health

ISSN: 1476-069X