The analysis of interval-censored observations
Interval-censored data occur in basically all empirical research, but is rarely taken into account properly. Appropriate methodology has been developed for a large number of problems. Despite the available statistical technology, interval-censoring is often swept under the carpet. A notorious example is the analysis of cancer trials with progression-free survival as outcome, in which almost invariably interval-censoring is ignored. We argue that one of the reasons that the appropriate technology is not used is the ignorance of many statisticians what the effects are of ignoring interval censoring, but also what software is available. In the forthcoming CRC Press book (Survival Analysis with Interval-Censored Data: A Practical Approach with examples in R, SAS and BUGS by Kris Bogaerts, ArnoštKomárek, Emmanuel Lesaffre, 2017), we give an overview of the pitfalls of ignoring interval censoring, of the different approaches both in a frequentist and a Bayesian context and the currently available statistical software.
The course will be split up into four sessions (two in the morning and two in the afternoon). The contents of the sessions is:
Session 1: First an introduction is given to why one needs to take into account interval censoring, but also introduction to the example data sets will be given. Then a brief review of survival methods for the analysis of right-censored data for single and two-sample problems is given. In the third part we discuss the non-parametric Turnbull estimator (counterpart of a classical Kaplan-Meier). Then we show how significance tests for right-censored data need to be adapted for the analysis of interval-censored data. Illustrations with R analyses of some of the data sets will be given in-between the theoretical parts.
Session 2: Main results on Cox regression models and AFT regression models for right-censored are reviewed and then extended to interval censoring. Briefly bivariate interval censoring will be treated, and frailty models. The emphasis in this session is more on the applications rather than the theoretical concepts.
Session 3: A brief introduction to parametric Bayesian methods will be provided and to Bayesian software such as BUGS-related software (WinBUGS/OpenBUGS/JAGS) with their R-interface that can handle Bayesian models with interval-censored data.
Session 4: This session will be devoted to mainly regression models together with corresponding frailty/random effects versions. Also some more complex settings will be discussed. Illustrations will be given with BUGS and R.
Note: there will be focus on R and BUGS software, but there will be also indicated what SAS software is available.
Learning Outcomes:
Knowledge: why interval censoring needs to be addressed, what methods are available but even more importantly what software is available on the different platforms: R, SAS, BUGS and how to use the software.
At the end of the course, participants will be able to judge when interval censoring is important to take into account and how to take it into account.
Instructor:
Emmanuel Lesaffre is Professor of Biostatistics at I-Biostat, KU Leuven, Belgium. His research interests include Bayesian methods, longitudinal data analysis, statistical modelling, analysis of dental data, interval censored data, misclassification issues and clinical trials. He has written more than 400 papers in peer-reviewed statistical and medical journals. He is the founding chair of the Statistical Modelling Society, past-president of the International Society for Clinical Biostatistics and fellow of ISI and ASA. He (co-)authored five books among which the recently published Wiley book Bayesian Biostatistics (2012) together with Andrew Lawson. He has taught many statistical courses on a variety of topics in regular Master programs, but also short-courses on-site both at national as well as international level. The audience were medical students and researchers, engineers, mathematicians and statisticians.
Textbook:
CRC Press book: Survival Analysis with Interval-Censored Data: A Practical Approach with examples in R, SAS and BUGS by Kris Bogaerts, ArnoštKomárek, Emmanuel Lesaffre, 2017,
Material to bring along:
A laptop could be useful, but is not necessary since annotated output will be provided but no hands on computing.