CMStatistics 2020: Start Registration
View Submission - CMStatistics
Title: Reference-based sensitivity analysis for time to event data Authors:  James Carpenter - London School of Hygiene and Tropical Medicine (United Kingdom) [presenting]
Andew Atkinson - University of Bern (Switzerland)
Tim Clayton - LSHTM (United Kingdom)
Mike Kenward - Ashkirk (United Kingdom)
Abstract: Survival analysis typically assumes censoring at random, i.e. that, conditional on covariates in the model, the distribution of event times is the same, whether they are observed or censored. When trial patients who remain in follow up stay on their assigned treatment, then analysis under this assumption broadly addresses the de jure, or while on treatment strategy estimand. In such cases, we may well wish to explore the robustness of our inference to more pragmatic assumptions about the behaviour of patients post censoring. Such questions can be explored for trials with continuous outcome data using reference-based multiple imputations. This has two advantages: (a) it avoids the user specifying numerous parameters describing the distribution of patients' post-withdrawal data and (b) it is, to a good approximation, information anchored, so that the proportion of information lost due to missing data under the primary analysis is held constant across the sensitivity analyses. We bring this approach to the survival context, proposing a class of reference-based assumptions appropriate for survival data. We report a simulation study exploring the extent to which the multiple imputation estimator (using Rubin's variance formula) is information anchored in this setting and then illustrate the approach by re-analysing data from a randomized trial, which compared medical therapy with angioplasty for patients presenting with angina.