In epidemiological research it is common to follow large cohorts over time with respect to exposure and outcome development. This can, however, be expensive, time consuming or even logistically impossible. A much used alternative is case-control studies where only a suitably chosen subset of the full cohort is used for data collection and analysis. One such study design is the nested case-control design where m controls are sampled for each case. It is required that the controls are alive and event-free at the time the case experienced the event. Due to this requirement we say that the controls are matched to the cases on time. The controls are usually also matched on additional background factors for instance sex, or year of birth.
Due to the matching it has traditionally been considered impossible to reuse the controls i.e. use the controls for other cases than they originally were sampled for. However, Samuelsen (1997) introduced a method for breaking the matching which permits reuse of controls. This method involves weighting the likelihood contributions with inverse sampling probabilities. These probabilities must be estimated from the data and the thesis compares different estimators for these sampling probabilities along with some alternatives to inverse probability weighting. I specifically look at the complications of additional matching in relation to inverse probability weighting for which there have been little focus.
Even though the methodology has been available for more than 15 years, it has not yet been used in practice by the epidemiological community. One reason for this may be that there has been no available software for carrying out such analyses. By explaining the concept in an applied manner and developing an R-package that both estimates the weights and carries out the regression, I hope that the methodology will become more popular among the epidemiologists in the future.
List of papers I-IV. The papers are removed from the thesis due to publisher restrictions.
Paper I. Støer NC and Samuelsen SO (2012). Comparison of estimators in nested case-control designs with multiple outcomes. Lifetime data analysis 18(3):261–283. doi:10.1007/s10985-012-9214-8
Paper II. Støer NC and Samuelsen SO (2013). Inverse probability weighting in nested case-control studies with additional matching - a simulation study. Statistics in Medicine 32(30):5328–5339. doi:10.1002/sim.6019
Paper III. Støer NC, Meyer HE and Samuelsen SO (2014). Reuse of controls in nested case-control studies. Epidemiology 25(2):315–317. doi:10.1097/EDE.0000000000000057
Paper IV. Støer NC and Samuelsen SO. multipleNCC: Inverse probability weighting of nested casecontrol data in R. Manuscript.