Cumulative effect of PM2.5 components is larger than the effect of PM2.5 mass on child health in India

Air pollution is a major global health risk, with 6.67 million deaths attributable to the combined effects of household and ambient air pollution. In 2019, over 99% of the world’s population was exposed to PM2.5 levels exceeding the WHO annual guideline of 5 µg m−3. In India, exposure has risen over the last three decades; in 2017, 76.8% of the population experienced PM2.5 above the national standard (40 µg m−3). Children under 5 years are particularly vulnerable, with PM2.5 linked to adverse pregnancy outcomes (e.g., low birth weight, stillbirth, preterm birth), early childhood morbidities (e.g., acute respiratory infections), impaired growth, stunting, and anaemia. PM2.5 is a mixture of multiple chemical species with varying toxicity and source profiles. Consequently, reliance on total PM2.5 mass may underestimate health risks if more toxic species drive outcomes. Evidence from LMICs on species-specific and sector-specific associations with child health is sparse. This study aims to estimate the associations of ambient PM2.5 mass, its chemical components, and sectoral source contributions with three under-5 child health outcomes—low birth weight (LBW), anaemia, and acute respiratory infection (ARI)—in India, and to assess expected health benefits of meeting clean air targets. The study integrates NFHS-4 survey data with high-resolution satellite-derived PM2.5 and WRF-CMAQ speciated and sectoral outputs, employing a two-stage modeling strategy to address multicollinearity and infer component-specific effects.

The paper situates its research within evidence that PM2.5 exposure in pregnancy and early life is associated with adverse outcomes including LBW, stillbirth, preterm birth, congenital anomalies, impaired growth, ARI, and anaemia. Prior studies, including meta-analyses, have implicated specific PM2.5 components (e.g., elemental carbon, ammonium, certain metals) as potentially more harmful than aggregate PM2.5 mass, but national-scale evidence from LMICs is limited. Existing LMIC studies report associations of PM2.5 with ARI and anaemia in children but rarely explore chemical constituents or sectoral sources to quantify cumulative effects. The authors highlight the gap that PM2.5 mass may underestimate true health risks if more toxic species disproportionately contribute to effects, and that component-source attribution can inform targeted mitigation strategies.

Study design and data sources: Cross-sectional analysis linking National Family Health Survey (NFHS-4, 2015–2016) child-level health and sociodemographic data to ambient PM2.5 mass, chemical component, and sectoral exposure estimates. The NFHS-4 used a two-stage stratified sampling design across all districts of India with PSUs defined as villages (rural) or census blocks (urban). Final analytical sample: 177,072 observations for LBW and ARI after excluding records with missing exposure, outcomes, and covariates; anaemia analyses excluded records with missing anaemia status. Key outcomes: LBW (<2500 g) from records or maternal report; anaemia (Hb <11 g/dL via HemoCue Hb 201+); ARI symptoms within past two weeks (cough with rapid/difficult breathing). Covariates: Child sex; maternal education and parity; for ARI: child age; for LBW: maternal age categories and BMI; for anaemia: maternal Hb, per capita dietary iron intake; household wealth quintile; residence (urban/rural); passive smoking; cooking fuel type (clean, solid, kerosene, others). Exposure assessment: Monthly PM2.5 at 1 km × 1 km resolution derived from MODIS-MAIAC AOD converted to surface PM2.5 using MERRA-2 dynamic and diurnal scaling factors calibrated to CPCB monitors (annual r=0.97, RMSE=7.2 µg m−3). PSU-level exposures assigned using geocodes. Exposure windows: Pregnancy-average PM2.5 for LBW (based on gestational age); life-course average PM2.5 from birth to survey for anaemia and ARI. PM2.5 composition: WRF v3.9.1–CMAQ v5.3.1 simulations (36 km resolution, 25 vertical levels) for 2016 using ERA5 meteorology and GAINS-ASIA emissions; boundary conditions from CAM-chem; performance against ground monitors: monthly R2=0.81, index of agreement 0.94. Component mass concentrations estimated by scaling modeled component mass fractions to satellite PM2.5 at 1 km resolution: Mi = Mi,model × PM2.5,satellite / PM2.5,model; regridded bilinearly to 1 km. Sectoral contributions: Subtraction method in WRF-CMAQ—control run with all emissions; sequential zero-out by sector (transport, small/medium industries, brick, major industries, power, domestic solid fuel, agricultural residue burning, construction, road dust, others including refuse burning, crematoria, NH3, biogenic, refineries, evaporative NMVOCs); scaled to satellite PM2.5 similar to components to estimate PSU-level sectoral exposures. Statistical analysis: To address multicollinearity of components, a two-stage approach was used. Stage I: regress PM2.5 mass on each component Pi to obtain residuals ResPM2.5 capturing PM2.5 variation not explained by Pi. Stage II: mixed-effects logistic regression with random intercepts for PSU: logit Pr(Y=1|u) = β0 + β1 Pj + β2 ResPM2.5 + γ(confounders) + ui, with ui ~ N(0,σ2). Outcomes modeled separately for LBW (gestational exposure), anaemia and ARI (life-course exposure). Stratified analyses: interaction between LBW status (LBW vs normal) and component exposures for anaemia and ARI. Nonlinearity: penalized cubic smoothing splines f(Pj) substituted for linear Pj in Stage II to explore exposure-response shapes. Whole-mass effects: adjusted ORs for PM2.5 mass estimated via cluster (PSU)-logistic regression. Cumulative effects: computed as β_cum = Σ mi βi (mi: mass fraction of component i; βi: component-specific regression coefficient), with SE assuming independence: Var(β_cum)=Σ mi^2 Var(βi); converted to OR per 10 µg m−3 PM2.5. Health benefits modeling: For meeting NAAQS (40 µg m−3) and WHO AQG (5 µg m−3), district-level ΔPM2.5 estimated; attributable fraction AF=(RR−1)/RR with RR=exp[log(OR)×ΔPM2.5], using OR per unit PM2.5 for both mass-only and cumulative-component scenarios; expected reduction E=AF×district prevalence. Analyses conducted in R 4.1.2.

Sample and exposure: Final analytical sample included 177,072 observations for LBW and ARI; national LBW prevalence 16.6% (16.4–16.8), anaemia 56.8% (56.6–57.1), ARI 2.8% (2.7–2.9). Median PSU-level annual PM2.5: 62 µg m−3 (IQR 52–79). Component and sector profiles: Dominant components: OC, NO3−, NH4+, SO4^2−, and others (chloride, sea salt, crustal minerals, water, unspecified); major contributing sectors include domestic, industrial, international transport, agriculture, and transport. Mass-only associations (per IQR increase in PM2.5): OR 1.15 (1.12–1.18) for LBW, 1.57 (1.54–1.59) for anaemia, 1.32 (1.24–1.40) for ARI. Component-specific associations (per IQR increase): For LBW—NO3− 1.17 (1.14–1.20), others 1.14 (1.11–1.17), NH4+ 1.13 (1.11–1.16), EC 1.11 (1.08–1.14), soil 1.09 (1.07–1.11), SO4^2− 1.07 (1.04–1.09), OC 1.05 (1.03–1.08). For anaemia—NO3− 1.36 (1.32–1.41), NH4+ 1.28 (1.25–1.31), others 1.25 (1.21–1.28), EC 1.21 (1.18–1.25), soil 1.18 (1.16–1.20), SO4^2− 1.14 (1.12–1.17), OC 1.12 (1.09–1.15). For ARI—NO3− 1.52 (1.42–1.61), EC 1.49 (1.40–1.58), OC 1.46 (1.37–1.55), others 1.33 (1.26–1.41), NH4+ 1.15 (1.09–1.21); soil 0.93 (0.89–0.96) and SO4^2− 0.81 (0.77–0.85) showed inverse associations for ARI. Sector-specific associations (per IQR increase): LBW—road dust 1.13 (1.11–1.14) highest; international 1.09 (1.07–1.10), industry 1.07 (1.05–1.08), agriculture 1.06 (1.05–1.07), others 1.04 (1.02–1.07), transport 1.05 (1.02–1.07); domestic 1.00 (0.97–1.02), power 0.91 (0.89–0.93). Anaemia—unorganized/others 1.19 (1.18–1.20) highest; international 1.11 (1.09–1.13), domestic 1.09 (1.06–1.11), road dust 1.09 (1.08–1.11), agriculture 1.08 (1.07–1.09), industry 1.04 (1.02–1.06), transport 1.03 (1.00–1.06); power 0.96 (0.94–0.98). ARI—domestic 1.30 (1.24–1.35) highest; transport 1.21 (1.14–1.28), others 1.21 (1.14–1.28), agriculture 1.10 (1.07–1.13); industry 0.90 (0.86–0.93), road dust 0.96 (0.92–0.99), power 0.96 (0.92–1.01), international 1.03 (0.99–1.07). Stratified by birth weight: Effects of components on anaemia and ARI were stronger among LBW children compared to normal birth weight. For anaemia—NO3− OR 1.47 (1.41–1.52) in LBW vs 1.34 (1.30–1.38) in normal; EC 1.30 vs 1.20; OC 1.18 vs 1.11. For ARI—NO3− 1.72 (1.60–1.85) vs 1.46 (1.37–1.56); EC 1.66 vs 1.45; OC 1.59 vs 1.42. Nonlinear exposure-response: For LBW, OC, NO3−, NH4+ showed monotonic increases up to midrange concentrations; EC, soil, others showed inverted U-shapes. For anaemia, most components (EC, NH4+, NO3−, OC, SO4^2−, others) exhibited monotonic increases; soil showed a decline at higher exposures. ARI showed varying nonlinear patterns. Cumulative effects vs PM2.5 mass (per 10 µg m−3 increase): Cumulative component-based ORs were substantially higher—LBW 1.23 (1.21–1.26), anaemia 1.49 (1.45–1.52), ARI 1.35 (1.29–1.41)—than mass-only ORs—LBW 1.05 (1.04–1.06), anaemia 1.10 (1.09–1.11), ARI 1.11 (1.08–1.13). Expected health benefits of meeting clean air targets: Using mass-only ORs per unit PM2.5 (LBW 1.005, anaemia 1.01, ARI 1.011), meeting NAAQS could reduce prevalence from 16.6% to 14.5% for LBW, 56.8% to 44.8% for anaemia, and 2.8% to 2.1% for ARI; achieving WHO-AQG could reduce to 11.6% (LBW), 32.9% (anaemia), and 1.5% (ARI). Using cumulative component ORs per unit (LBW 1.021, anaemia 1.041, ARI 1.03), meeting NAAQS could reduce prevalence to 15.7% (LBW), 50.7% (anaemia), 2.3% (ARI); WHO-AQG to 14.7% (LBW), 44.2% (anaemia), 2.1% (ARI). Overall, NO3−, NH4+, EC, and OC were most consistently associated with adverse outcomes; source attribution indicates domestic, transport, road dust, industry, agriculture, and international transport contribute to risks in at least one outcome.

The study directly addresses the question of whether chemical components and sources of PM2.5 have differential and cumulatively greater associations with child health than PM2.5 mass alone. Findings show that specific components—particularly nitrate (NO3−), ammonium (NH4+), elemental carbon (EC), and organic carbon (OC)—have stronger associations with LBW, anaemia, and ARI than total mass, and that summing component-specific effects produces substantially higher cumulative ORs than mass-only estimates. This implies that using total PM2.5 mass as a surrogate exposure metric may underestimate the true health burden. Sectoral analysis identifies domestically generated pollution (household solid fuels), transport, road dust, industry, agriculture, and international transport as important contributors to adverse outcomes, with domestic emissions showing the strongest association with ARI. Mechanistically, results align with plausible causal pathways: PM components can cross the placental barrier, induce oxidative stress and placental inflammation, impairing nutrient/oxygen transfer and foetal growth (LBW); in young children, PM-induced oxidative stress and inflammation can lower immunity and increase susceptibility to infections (ARI); chronic inflammation and hepcidin-mediated iron regulation may reduce iron absorption and erythropoiesis, contributing to anaemia. Policy relevance is high: component- and sector-specific evidence supports targeted mitigation (e.g., clean household energy, industrial emissions control, transport and road dust management), and health impact projections show substantial potential reductions in child morbidities by meeting NAAQS and WHO-AQG. The stratified analyses underscore greater vulnerability among LBW children, highlighting a subgroup for targeted protection.

This first comprehensive national assessment in India demonstrates significant associations between PM2.5 chemical components and sectoral contributions with LBW, anaemia, and ARI in under-5 children. Cumulative effects of components per 10 µg m−3 PM2.5 increase are markedly larger than effects estimated using total PM2.5 mass, indicating that mass-only metrics underestimate health risks. NO3−, NH4+, EC, and OC emerge as key harmful components, and domestic, transport, road dust, industrial, agricultural, and international sources contribute to risks. Children born with LBW are disproportionately affected by PM2.5 components. These findings support targeted, component- and sector-focused air quality interventions to achieve greater child health benefits. Future work should include longitudinal cohort studies and toxicological investigations to elucidate causal pathways and refine exposure-response functions, alongside improved emission inventories and speciation data to reduce uncertainties.

Key limitations include: (1) Assumption of residential stability for children during early-life exposure periods; (2) Use of modeled component mass fractions and sectoral contributions assumed representative over the entire exposure period, with uncertainties tied to emission inventories and modeling; (3) Potential unmeasured or residual confounding, suggested by observed nonlinear and non-monotonic exposure-response relationships; (4) Cross-sectional survey design with cluster-level exposure assignment limits causal inference and may attenuate or bias effect estimates; (5) Sparse ground-based speciated PM2.5 measurements in India necessitating reliance on model-based speciation and sectoral attribution.