The authors compared the relative effectiveness of two distinct follow-up designs in prospective cohort studies--the active approach, based on direct contact with study subjects, and the passive approach, based on record linkages with population-based cancer registries--utilizing available information from the New York University Women's Health Study (WHS) and the New York State Cancer Registry (NYSCR). The analyses were limited to breast cancer cases identified during the period 1985-1992, for which follow-up was considered reasonably complete by both the WHS and the NYSCR. Among 12,947 cohort members who reported a New York State address, 303 pathologically confirmed cases were identified through active follow-up and 284 through record linkage. Sixty-three percent of cancers were identified by both sources, 21% by the WHS only, and 16% by the NYSCR only. The agreement was appreciably better for invasive cancers. The percentage of cases identified only by the NYSCR was increased among subjects whose active follow-up was incomplete, as well as among nonwhites, obese patients, and parous patients. This suggests that relying on either type of follow-up alone may introduce certain biases in evaluating risk factors for breast cancer. Combining both approaches appears to be a better strategy in prospective cohort studies.