This study assesses the effects of long-term pollution exposure on acute lower respiratory infection (ALRI), which is the first cause of death in children aged less than 5 years in developing countries. Previous observational studies have found that exposure to biomass fuel is associated with a twofold increase in ALRI rates in children. The intervention in this experimental design reduced smoke pollution using a chimney stove instead of open wood fires. The primary outcome is physician-assessed pneumonia in children up to 18 months of age. Other outcomes are fieldworker-assessed pneumonia, seven other characteristics of physician-diagnosed pneumonia that provide clinical confirmation of the condition (radiography, oxygen saturation) or could account for a differential effect (presence of virus) and the exposure-response function. The study population had high infant mortality, ALRI incidence and exposure to household air pollution. Most study children showed severe under-nutrition and growth retardation.
The sample size calculation considered a 20 per cent drop-out rate. The study households had no migration plans, used open fires for cooking in enclosed kitchens and were occupied by pregnant women or children younger than 4 months of age. After parallel randomisation by concealed permuted blocks of 10 homes, 269 households received chimneys and 265 continued using open wood fires (controls). Trained fieldworkers made weekly revisions of the children's health status and referred children presenting symptoms to physicians who had previous standardisation training and were masked to intervention status. To assess exposure-response, the fieldworkers ensured the functioning of the stoves. Because the secondary outcome analysis was by intention to treat, the researchers measured exposure misclassification through an exposure-response analysis. The fieldworkers obtained carbon monoxide exposure data for all study children every 3 months. Multiple imputations were used to predict the missing pneumonia observations on the physician-assessed outcomes for children who were not properly analysed by doctors after fieldworker referral.
The treatment did not significantly reduce physician-diagnosed pneumonia in the children under study. The 22 per cent observed reduction after multiple imputations was 3 percentage points smaller than the target reduction for the sample size. After multiple imputations, 149 cases arose in the experimental group and 180 in the control group (case-control rate ratio [RR] = 0.78, 95% confidence interval [CI] 0.59 to 1.06; reduction = 22 per cent, 95% CI = ?.06 to ?.41; p = 0.095). As for the physician-assessed cases, those with chest radiographic confirmation had an RR of 0.74 (95% CI = 0.42 to 1.15, p = 0.231). Nevertheless, severe (hypoxaemic) cases showed a reduction by one third (RR = 0.68, 95% CI = 0.36 to -1.33, p = 0.026). There was no observed differential effect on cases with negative results on respiratory syncytial virus (RR = 0.79, 95% CI = 0.53-1.07, p = 0.192) compared with those with positive results (RR = 0.76, 95% CI = 0.42-1.16, p = 0.275). However, the reduction in severe pneumonia was larger for cases negative for respiratory syncytial virus (RR = 0.54, 95% CI = 0.31-0.91, p = 0.026) than for positive cases (RR = 0.87, 95% CI = 0.46-1.51, p = 0.633).
Finally, the chimney stoves reduced mean exposure by 50 per cent, and the exposure distributions for the groups overlapped to a great degree. Therefore, the lack of a significant observed effect might respond to insufficient exposure reduction or insufficient power, and the exposure misclassification (overlapping distributions) might have introduced additional error. The exposure-response analysis shows that a larger exposure reduction to less than the mean in the experimental group is associated with a greater reduction in the risk for pneumonia. Children might have been exposed to smoke from open fires in their neighbours' homes. Moreover, chimney stoves do not reduce emissions but vent them to the immediate outdoors. Therefore, fuel interventions producing even lower average exposures are recommended.