Numerous epidemiological surveys have tried to evaluate the effects of exposures relevant to pulmon ary disease.1 A number of studies have analysed forced expiratory volume in one second (FEVJ using linear regression analysis, thus enabling adjustment for age, height, and sex as well as potential confound ing variables. Within the last few years, studies analysing dichotomous FEV\ variables with different cutoff points in multivariate regression models have been published.2 3 As there seems to be no obvious reason for dichotomising the continuous FEV13 we have com pared the results of the two different approaches to analysis of FEV\.