Selection effects of repeatability criteria applied to lung spirometry.

The potential for introducing bias in studies of pulmonary function by the exclusion of subjects with nonrepeatable measurements was examined in a cohort of Vermont granite workers followed for five years. At each annual survey, a "test failure" was defined as a test in which the two largest forced expiratory volumes in one second (FEV1) differed by more than 200 ml. "Persistent test failure" was defined in terms of 1) the number of test failures for each worker over the six surveys and 2) the difference between the two best efforts at each survey, averaged over all surveys for each worker. The rate of FEV1 loss was estimated for each subject based only on repeatable measurements. It is widespread practice to exclude subjects from analysis who do not perform repeatable lung function tests. The authors found that subjects with persistent test failure were losing FEV1 at a faster rate than subjects without. The results suggest that the application of rigid repeatability criteria may bias epidemiologic findings by the exclusion of many subjects with accelerated loss of lung function.