Tuesday, Aug 8: 8:30 AM - 10:20 AM
Invited Paper Session
Metro Toronto Convention Centre
Section on Statistics in Epidemiology
Subgroup analysis has long been done in clinical trials to assess heterogeneity of treatment effects. Most often, this is a post hoc analysis - albeit pre-specified subgroups may be defined in the protocol - with little attention to Type 1 Error control or bias in treatment effect estimates. Consequently, findings in such subgroups are and should be viewed skeptically. Thus, the common approach to subgroup analysis rarely leads to clarification of whether heterogeneous treatment effects occur across subgroups and in fact often add confusion to the interpretation of results. So, why do we do them so commonly and so casually? This talk will focus on subgroup identification - in contrast to subgroup analysis - by which is meant a more disciplined analytical approach that seeks to control Type 1 Errors and corrects for bias in treatment effect estimates. It will be argued that by taking a more thoughtful, rigorous approach results can be interpreted more meaningfully and labelling of a new treatment can be more clear and confident when describing any heterogeneous treatment effects.
Individualized treatment rules (ITRs) inform tailored treatment decisions based on the patient's information, where the goal is to optimize clinical benefit for the population. When the clinical outcome of interest is survival time, most of current approaches typically aim to maximize the expected time of survival. We propose a new criterion for constructing ITRs that optimize the clinical benefit with survival outcomes, termed as the adjusted probability of a longer survival. This objective captures the likelihood of living longer with being on treatment, compared to the alternative, which provides a straightforward interpretation to communicate with clinicians and patients. We develop a new method to construct the optimal ITR by maximizing a nonparametric estimator of the adjusted probability of a longer survival for a decision rule. Simulation studies demonstrate the reliability of the proposed method across a range of different scenarios. We further perform data analysis using data collected from a randomized Phase III clinical trial (SWOG S0819).
The heterogeneity of treatment effects is challenging and may not be replicated in new trials. There is a simple reason for this: Clinical trials are powered to provide reliable inference on the overall treatment effect. Estimates of subgroup treatment effects will be variable and unreliable in a single trial. As data alone cannot provide definitive answers on subgroup effects, it is important that a workflow around assessing treatment effect heterogeneity should include clinical considerations. A structured approach should be followed to identify variables that modulate the treatment effect and predicting possibly heterogeneous treatment effects on new patients. The workflow includes methodological learnings from the 2022 Analytics Subgroup Challenge at Novartis.