Nested case–control studies: should one break the matching?
- 469 Downloads
In a nested case–control study, controls are selected for each case from the individuals who are at risk at the time at which the case occurs. We say that the controls are matched on study time. To adjust for possible confounding, it is common to match on other variables as well. The standard analysis of nested case–control data is based on a partial likelihood which compares the covariates of each case to those of its matched controls. It has been suggested that one may break the matching of nested case–control data and analyse them as case–cohort data using an inverse probability weighted (IPW) pseudo likelihood. Further, when some covariates are available for all individuals in the cohort, multiple imputation (MI) makes it possible to use all available data in the cohort. In the paper we review the standard method and the IPW and MI approaches, and compare their performance using simulations that cover a range of scenarios, including one and two endpoints.
KeywordsCase–cohort Competing risks Cox regression Inverse probability weighting Matching Multiple imputation Nested case–control
Most of this research was done when Ørnulf Borgan was visiting the Department of Medical Statistics at London School of Hygiene and Tropical Medicine the spring of 2014. The department is acknowledged for its hospitality and for providing the best working facilities. We also want to thank Nathalie Støer for letting us use her new R package multipleNCC before it was made publicly available.
- Bartlett JW, Seaman SR, White IR, Carpenter JR (2014) Multiple imputation of covariates by fully conditional specification: Accommodating the substantive model. Stat Methods Med Res. doi: 10.1177/0962280214521348
- Borgan Ø, Samuelsen SO (2013) Nested case–control and case–cohort studies. In: Klein JP, van Houwelingen HC, Ibrahim JG, Scheike TH (eds) Handbook of survival analysis. Chapman and Hall/CRC Press, Boca Raton, Florida, pp 343–367Google Scholar
- Meng X (1994) Multiple-imputation inferences with uncongenial sources of input. Stat Sci 9:538–558Google Scholar
- Støer NC, Samuelsen SO (2014) multipleNCC: weighted Cox-regression for nested case-control data. http://CRAN.R-project.org/package=multipleNCC, R package version 1.0
- Van Buuren S, Groothuis-Oudshoorn K (2011) Mice: multivariate imputation by chained equations in R. J Stat Softw 45:1–67Google Scholar