Increased precision using countermatching in nested case-control studies

Abstract

Nested case-control studies in occupational cohorts are often used to estimate exposure effects when development of detailed exposure estimates for all cohort members is too costly. Duration of exposure, which can act as a surrogate for cumulative exposure, is often readily available for all cohort members. Langholz and others have recently proposed a method of control selection called countermatching, which uses data on the surrogate to determine which controls are selected from the risk set for a given case. This method may increase precision relative to the usual random sampling of the risk set. We compare countermatching with random sampling in a nested case-control study of silicosis among miners. Data on cumulative exposure were in fact available for all cohort members, enabling estimation of the parameter of interest in the full cohort. We conducted nested case-control analyses using 100, 20, 10, and 3 controls per case using random sampling and additional analyses using 3 controls per case with two different methods of countermatching. All analyses were replicated 50 times to explore the statistical properties of the estimated exposure parameter. We found that one of the countermatching methods markedly increased efficiency compared with random sampling. Countermatching using 3 controls per case yielded an approximate 25% increase in relative efficiency compared with random sampling; it was approximately equivalent to random sampling using 10 controls.