Does regression analysis of lung function data obtained from occupational epidemiologic studies lead to misleading inferences regarding the true effect of smoking?

Abstract

Exposure-response studies of the relationship between ventilatory function and dust exposure in workers are often quantified using linear regression methods. In coal miners, this technique has indicated that average effects of smoking and moderate dust exposure are roughly equivalent. However, the validity of direct comparison of the average effects of smoking and dust exposure has been questioned, the argument being that smoking causes severe effects in a minority, but leaves the remainder largely unaffected. This hypothesis was studied by examining distributions of FEV1 in a group of working coal miners where mean effects associated with both smoking and dust exposure have been detected. Overall, the results suggest that comparison of average effects of smoking and dust exposure derived from linear regression analysis is valid and not misleading.