Speaker
Bryan M. Langholz, Professor, Division of Biostatistics, Department of Preventive Medicine, School of Medicine, University of Southern California
Abstract
The nested case-control model, in which case-control sets are sampled independently from event time risk sets, is proposed as a way to connect cohort to case-control study methods for disease incidence data. Continuous and grouped time cohort data lead to individually matched and unmatched case-control structures. With risk sets sampled independently, full cohort analysis methods based on event time risk sets generalize to weighted versions for case-control data and often the same software can be used. The nested case-control approach is a "prospective model with sampling" alternative to the "retrospective model" generally used to represent case-control studies in the epidemiology and biostatistics literature. The nested case-control model has the advantage of better representing reality and providing some unity to apparently disparate approaches to epidemiologic research.