Optimizing Exposure Measures in Large-Scale Household Air Pollution Studies: Results from the Multicountry HAPIN Trial

Abstract

Repeated measurements of household air pollution may provide better estimates of average exposure but can add to costs and participant burden. In a randomized trial of gas versus biomass cookstoves in four countries, we took supplemental personal 24-h measurements on a 10% subsample for mothers and infants, interspersed between protocol samples. Mothers had up to five postrandomization protocol measurements over 16 months, while infants had three measurements over one year. For the subsample, we added up to 6 supplemental postrandomization samples for mothers and 3 for infants, measuring PM_2.5, black carbon (BC) (mothers only), and carbon monoxide (CO) at each visit. 310 mothers had both protocol (n = 1026) and supplemental (n = 1099) valid exposure measurements. For children, supplemental data sufficient for analysis were collected in only two countries; 94 infants had both protocol (n = 317) and supplemental (n = 234) samples. The geometric means for protocol and supplemental samples for mothers for PM_2.5 were 37 μg/m³ and 38 μg/m³, respectively, while for infants, they were 42 μg/m³ and 46 μg/m³. Mixed models comparing supplemental to protocol samples, controlling for covariates, found few differences between protocol and supplemental samples. Supplemental analyses among control mothers with complete protocol measurements found that an average of three measurements explained 81% of the variance of the average of all six measurements.