ABSTRACT
Cohort studies are considered to be the most reliable study design in epidemiology, while case-control studies are easier and quicker to implement. The logistic regression model, which is typically used in traditional case-control studies (cf. Chapters 3 and 4), does not explicitly consider the time aspect of the development of a disease. In contrast, in cohort studies of the time to death of a particular cause or the time to onset of a specific disease (generically denoted “failure”), it is common to use Cox’s regression model to assess the influence of risk factors and other covariates on mortality or morbidity. In this chapter, we will assume that failures (i.e., deaths or disease occurrences) in the cohort are modelled 286by Cox’s regression model, and we discuss how various cohort sampling designs make it possible to adopt Cox regression also for certain types of case-control data. The purpose of the chapter is both to give an overview of case-control methods for failure time data and to provide a background for the succeeding chapters in Part IV of the handbook.
