Exposure-response relationships for personal exposure to fine particulate matter (PM2·5), carbon monoxide, and black carbon and birthweight: an observational analysis of the multicountry Household Air Pollution Intervention Network (HAPIN) trial

Abstract

Background: Household air pollution (HAP) from solid fuel use is associated with adverse birth outcomes, but data for exposure-response relationships are scarce. We examined associations between HAP exposures and birthweight in rural Guatemala, India, Peru, and Rwanda during the Household Air Pollution Intervention Network (HAPIN) trial.

Methods: The HAPIN trial recruited pregnant women (9-<20 weeks of gestation) in rural Guatemala, India, Peru, and Rwanda and randomly allocated them to receive a liquefied petroleum gas stove or not (ie, and continue to use biomass fuel). The primary outcomes were birthweight, length-for-age, severe pneumonia, and maternal systolic blood pressure. In this exposure-response subanalysis, we measured 24-h personal exposures to PM_2·5, carbon monoxide, and black carbon once pre-intervention (baseline) and twice post-intervention (at 24-28 weeks and 32-36 weeks of gestation), as well as birthweight within 24 h of birth. We examined the relationship between the average prenatal exposure and birthweight or weight-for-gestational age Z scores using multivariate-regression models, controlling for the mother's age, nulliparity, diet diversity, food insecurity, BMI, the mother's education, neonate sex, haemoglobin, second-hand smoke, and geographical indicator for randomisation strata.

Findings: Between March, 2018, and February, 2020, 3200 pregnant women were recruited. An interquartile increase in the average prenatal exposure to PM_2·5 (74·5 μg/m³) was associated with a reduction in birthweight and gestational age Z scores (birthweight: -14·8 g [95% CI -28·7 to -0·8]; gestational age Z scores: -0·03 [-0·06 to 0·00]), as was an interquartile increase in black carbon (7·3 μg/m³; -21·9 g [-37·7 to -6·1]; -0·05 [-0·08 to -0·01]). Carbon monoxide exposure was not associated with these outcomes (1·7; -3·1 [-12·1 to 5·8]; -0·003 [-0·023 to 0·017]).

Interpretation: Continuing efforts are needed to reduce HAP exposure alongside other drivers of low birthweight in low-income and middle-income countries.

Funding: US National Institutes of Health (1UM1HL134590) and the Bill & Melinda Gates Foundation (OPP1131279).

Figure 1

Distribution of (A) birthweight and time-weighted (B) PM₂_·₅, (C) black carbon, and (D) carbon monoxide The corresponding numeric data are provided in the appendix (p 4). Results are presented separately for each study site and in combination for the entire trial. Dots are individual datapoints. X-axes are log transformed in panels B, C, and D. Thick solid lines inside the boxes are the medians. The lower and upper hinges (ie, the ends of the boxes) correspond to the 25th and 75th percentiles. The whiskers (ie, the lines beyond the boxes) extend from the hinge to 1·5 IQR. The panel-wide dotted vertical lines are study-wide medians. In panel A, the shaded area indicates low birthweight (<2500 g). In panel B, the dashed line is the WHO interim target level one annual guideline value of 35 μg/m³.

Figure 2

Exposure–response relationships between birthweight and prenatal (A) PM_2·5, (B) black carbon, and (C) carbon monoxide personal exposures Dashed lines correspond to the 95% CI and vertical dashes along the x-axis are observed measurements.

Figure 3

Exposure–response relationships between weight-for-gestational age Z scores and prenatal (A) PM_2·5, (B) black carbon, and (C) carbon monoxide personal exposures Dashed lines correspond to the 95% CI and vertical dashes along the x-axis are observed measurements.