Abstract
In an adaptive clinical trial, researchers use accumulating data on effectiveness and/or safety endpoints to make planned modifications to the design of the trial as it is being run. Perhaps the oldest family of adaptive trial designs are sequential designs. These enable researchers to sequentially monitor the outcome data from the trial, and use the observed signal to make decisions regarding whether to stop or continue recruitment to the trial. Various proposals exist in the literature, with these generally seeking a decision rule or stopping boundary that minimises the expected sample size of the trial subject to certain design constraints (for example, type I and II error rates). However, traditional approaches do not explicitly consider the delay between recruitment and observation of endpoints, or economic factors such as the costs of the research, the number of patients expected to benefit from the research, or the costs associated with replacing an existing standard treatment with a new one. Chick et al. (2017) propose a Bayesian decision-theoretic model of a fully sequential two-arm clinical trial, which does account for these factors, and uses the accumulating cost-effectiveness data from the trial to guide decision-making regarding the continuation or cessation of trial recruitment.
During this presentation, I will provide a brief account of the rationale, assumptions and solution of the Bayesian decision-theoretic model of Chick et al. I will then present contrasting findings from two recently published case-studies that retrospectively applied this model to data from the PROximal Fracture of the Humerus: Evaluation by Randomisation (ProFHER) trial and the Hydroxychloroquine Effectiveness in Reducing symptoms of hand Osteoarthritis (HERO) trial.
I will then discuss preliminary findings and ongoing work from a third case-study using data from the Dupuytren’s Interventions Surgery versus Collagenase trial. The DISC trial found collagenase injection and manipulation to be inferior to surgical treatment with respect to the majority of the clinical effectiveness endpoints that were assessed in the trial, but found that it was very likely to be cost-effective over a short time horizon. Like many other clinical trials, recruitment to the DISC trial was severely impacted by the COVID-19 pandemic, with recruitment taking at least 12 months longer than it likely would have done had the pandemic not occurred. This case study builds on the previous ones in two key respects. Firstly, we are directly comparing the decision rule from the model of Chick et al. based on cost-effectiveness outcomes, with decision rules indicated by other sequential approaches based on frequentist and Bayesian analyses of clinical effectiveness outcomes. Secondly, we will examine the role that these sequential monitoring of outcome data could have played in guiding decisions regarding the expected benefits of continuing recruitment following the onset of the pandemic.
Collegamento Microsoft Teams
Organizzazione
Martin Forster