Bridging Early Insights for Enhanced Oncology Decision-Making
Chair: Miaomiao Ge (Boehringer Ingelheim) and Wenqiong Xue (Boehringer Ingelheim)
Speaker: Matt Psioda, PhD (GSK)
Title: Using Win Odds to Improve Commit to Phase 3 Decision-Making in Oncology
Abstract: Making good commit to phase 3 decisions in drug development is a challenging problem. This is especially true in oncology where due to the variety of emerging therapies, the relationships between the overall survival registration endpoint, and measures used in phase 2 such as progression-free survival and objective response, are often poorly understood. We present a novel framework for phase 2 decision-making based on a three-endpoint win odds and formal quantitative decision-making criteria. We assess the suitability of the win odds, and its interpretation for decision-making purposes as the reciprocal of the survival hazard ratio using both a real clinical trial case study, and data generated with a multi-state model in a comprehensive simulation study derived from real data. Using a multi-state model to jointly simulate correlated patient level endpoints provides a clear understanding of the effects simulated in the different endpoints and creates clinically realistic data for testing the performance of assessment methods. We conclude that the integration of all available overall survival data with progression-free survival or objective response data into a single decision-making endpoint for patients without a survival event introduces precision, and as a result, the commit to phase 3 decision-making improves compared with other methods.
Speaker: Kevin Kunzmann, PhD (Boehringer Ingelheim) and ChengXue Zhong, PhD (Boehringer Ingelheim)
Title: Sensitivity of ‘Probability of Go’ in Early Oncology to Study Population Heterogeneity – a MAP Prior Based Approach to More Robust Decision Making
Abstract: In early oncology trials, (objective tumor) response based on the RECIST criteria is often used as primary endpoint to establish the activity of a treatment. A quantitative go/no-go decision boundary on the number of responders typically informs the decision to further development a compound. To predict the probability of crossing the go boundary during a trial, projections about the outcomes of non-assessable individuals still ‘at risk’ of a response and of individuals still to be recruited can be made. Heterogeneity of the study population can be modeled with models incorporating predictive covariates and/or time-at-risk but these complex models are often difficult to fit in small sample size situations. Instead, an approach using the well-established metanalytic predictive prior framework to model heterogeneity can be used to conduct a sensitivity analysis of probability of go. The sensitivity analysis allows to judge the impact different levels of heterogeneity have on the predicted probability of go, thus supporting robust decision making.
Speaker: Bo Huang, PhD (Pfizer)
Title: Analysis of Response Data for Assessing Treatment Effects in Oncology Studies
Abstract: In comparative studies, treatment effect is often assessed using a binary outcome that indicates response to the therapy. Commonly used summary measures for response include the cumulative and current response rates at a specific time point. The current response rate is sometimes called the probability of being in response (PBIR), which regards a patient as a responder only if they have achieved and remain in response at present. The methods used in practice for estimating these rates, however, may not be appropriate. Moreover, whereas an effective treatment is expected to achieve a rapid and sustained response, the response at a fixed time point does not provide information about the duration of response (DOR). As an alternative, a curve constructed from the PBIR over the entire study period may be considered, which can be used for visualizing how rapidly patients responded to therapy and how long responses were sustained. The area under the PBIR curve in the data observation window is the restricted mean DOR (RMDOR) in the ITT population, which enables valid statistical comparison between treatment groups based on DOR, making it a powerful and useful endpoint for assessing treatment effect of drugs that have higher response rate, shorter time-to-response, and longer time being in response. Statistical inference and applications in oncology clinical trials will be presented.
