Optimizing Exposure Measures in Large Scale Household Air Pollution Studies: Results from the Multi-Country Hapin Trial

Abstract

BackgroundRepeated exposure measurements of household air pollution may provide better estimates of average exposure but can add significantly to costs.MethodsIn a randomized trial of liquefied petroleum gas versus biomass cookstoves (HAPIN) in four countries, we took supplemental personal 24-hour measurements on a 10% subsample for mothers and their infants, interspersed between protocol samples. Per study protocol, mothers had up to 5 post-randomization measurements during gestation and the first year following birth, while their infants had 3 measurements in their first year. For the subsample, we added up to 6 supplemental post-randomization samples for mothers and three for infants, measuring PM2.5, black carbon (BC) (mothers only), and carbon monoxide (CO) at each visit.Results310 mothers had both protocol (n=1026) and supplemental (n=1099) valid exposure measurements. For children, supplemental data sufficient for analysis were collected only in Guatemala and India; 94 infants had both protocol (n=317) and supplemental (n=234) samples. The geometric means for protocol and supplemental samples for mothers for PM2.5 were 37 µg/m3 (GSD), 2.7) and 38 µg/m3 (GSD 2.7), respectively. For infants, the corresponding figures were 42 ug/m3 (GSD 2.4) and 46 µg/m3 (GSD 2.6) . Mixed models comparing supplemental to protocol samples across visits, controlling for site, arm, season, pre vs post birth, and weekend sampling, found no significant differences between protocol and supplemental samples for any pollutant, for either mothers or infants, except for CO for mothers. The supplemental samples for mothers were 1% higher, 1% higher, and 8% lower on the log scale, compared to protocol samples, for PM2.5, BC, and CO respectively. For infants they were 1% and 2% higher, for PM2.5 and CO.ConclusionFor our supplemental sub-sample, exposure measurements did not markedly change average exposure estimation. A more parsimonious sampling strategy may yield reasonable estimates without materially compromising precision.

Keywords: household air pollution, personal samples, repeated measures, PM2.5