Effects of Cooking with Liquefied Petroleum Gas or Biomass on Stunting in Infants

Abstract

Background: Household air pollution is associated with stunted growth in infants. Whether the replacement of biomass fuel (e.g., wood, dung, or agricultural crop waste) with liquefied petroleum gas (LPG) for cooking can reduce the risk of stunting is unknown.

Methods: We conducted a randomized trial involving 3200 pregnant women 18 to 34 years of age in four low- and middle-income countries. Women at 9 to less than 20 weeks' gestation were randomly assigned to use a free LPG cookstove with continuous free fuel delivery for 18 months (intervention group) or to continue using a biomass cookstove (control group). The length of each infant was measured at 12 months of age, and personal exposures to fine particulate matter (particles with an aerodynamic diameter of ≤2.5 μm) were monitored starting at pregnancy and continuing until the infants were 1 year of age. The primary outcome for which data are presented in the current report - stunting (defined as a length-for-age z score that was more than two standard deviations below the median of a growth standard) at 12 months of age - was one of four primary outcomes of the trial. Intention-to-treat analyses were performed to estimate the relative risk of stunting.

Results: Adherence to the intervention was high, and the intervention resulted in lower prenatal and postnatal 24-hour personal exposures to fine particulate matter than the control (mean prenatal exposure, 35.0 μg per cubic meter vs. 103.3 μg per cubic meter; mean postnatal exposure, 37.9 μg per cubic meter vs. 109.2 μg per cubic meter). Among 3061 live births, 1171 (76.2%) of the 1536 infants born to women in the intervention group and 1186 (77.8%) of the 1525 infants born to women in the control group had a valid length measurement at 12 months of age. Stunting occurred in 321 of the 1171 infants included in the analysis (27.4%) of the infants born to women in the intervention group and in 299 of the 1186 infants included in the analysis (25.2%) of those born to women in the control group (relative risk, 1.10; 98.75% confidence interval, 0.94 to 1.29; P = 0.12).

Conclusions: An intervention strategy starting in pregnancy and aimed at mitigating household air pollution by replacing biomass fuel with LPG for cooking did not reduce the risk of stunting in infants. (Funded by the National Institutes of Health and the Bill and Melinda Gates Foundation; HAPIN ClinicalTrials.gov number, NCT02944682.).

Figure 1. Screening, Randomization, and Follow-up.

Women could have had more than one reason for exclusion (i.e., categories are not mutually exclusive). Data are presented according to country in Table S2.

Figure 2.. Subgroup Analyses of the Effects of the Intervention on the Risk of Stunting at 12 Months of Age.

Shown is the analysis of the effects of the intervention on the risk of stunting at 12 months of age in the overall sample and in prespecified subgroups. Household food insecurity during the previous 30 days was assessed with the use of the Food and Agriculture Organization Food Insecurity Experience Scale. Gestation at the time of intervention refers to the time at which the women in the intervention group received an LPG cookstove. Exclusive breast‑feeding refers to the first 6 months of life. We defined stunting as a length‑for‑age z score that was more than two standard deviations below the median of the World Health Organization Multicentre Growth Reference Study standard for a given age and sex. We estimated the relative risk using a log‑binomial regression model for stunting at 12 months of age as a function of the trial‑group assignment and randomization strata. The number of stunted infants with length measurements and the total number of infants with length measurements who were born to women in each trial group are presented. A 98.75% confidence interval was used for the total sample included in the analysis (primary outcome), and 95% confidence intervals were used for each subgroup. The 95% confidence intervals for the subgroup analyses were not adjusted for multiplicity and should not be used in place of hypothesis testing.