Loss of Statistical Power Associated with the Use of a Job-Exposure Matrix in Occupational Case-Control Studies

Abstract

Case-control studies are important sources of information on the possible associations between occupational exposures and disease. One of the main methodologic difficulties is the retrospective assessment of occupational exposures among study subjects. The job-exposure matrix, or JEM, has been proposed as a means of assigning exposure to study subjects, given only information on their job titles. The purpose of the present analysis was to compare the performance of a JEM with a presumably more valid approach, namely, the assessment of exposure by a team of expert hygienists and chemists. The comparison was based on a data set generated by experts in the course of a large cancer case-control study, from which was created a simple JEM. For each of 160 substances, the statistical power of the original (“gold standard”) rating was compared with that of the JEM. For most substances, there was considerable exposure misclassification and consequent loss of power using the JEM. Data are presented to show how this misclassification varies among substances; this information can guide others in the choice of substances for inclusion in a JEM.