A comparison of two definitions of abnormality on pulmonary outcome in epidemiologic studies

Abstract

Definition of abnormal outcome on lung function testing has not been well standardized. Observed test results have been compared with predicted values based on age, height, and sex in 2 ways: as percent of the predicted (PP) and in relation to limits established by the 95% confidence interval around the predicted (CI). The present study examines these 2 methods of comparison with respect to effect on pulmonary outcome in 377 railroad workers and 193 pipefitters exposed to asbestos. Using accepted PP and CI criteria to define abnormal test results, we classified subjects as restricted or obstructed or normal by each method. The proportion restricted or obstructed was higher (p = 0.0001) if definitions were based on PP rather than on CI: restriction, 12.5% by PP versus 7.2% by CI; obstruction, 18.4% by PP versus 16.0% by CI. Twenty-seven percent (n = 47) of those abnormal by PP were normal by CI. Independent indicators of disease were used to compare this discordant group with groups normal (n = 391) or abnormal (n = 129) by both criteria. The subjects in the discordant group were older, with more years of asbestos exposure, longer latency, and more pleural plaques (p less than 0.05). The occurrence of chronic bronchitis and dyspnea was similar to that in the abnormal group. In our study, the method of comparison had a significant effect on outcome; PP appeared a better measure of disease. Our findings suggest that age and underlying risk of the study population should be considered in selecting a method of comparison in epidemiologic studies of respiratory disease.