TL12: Statistical Assessment of Comparative Effectiveness in Clinical Trials
*Isaac Nuamah, Janssen Research & Development 

Keywords: comparative effectiveness, meta-analysis, mixed treatment comparisons, network meta-analysis

The American Recovery and Reinvestment Act of 2009 allocated over $1 billion to support comparative effectiveness research (CER) in the US, and the Patient Protection and Affordable Care Act of 2010 provided sustained federal funding through 2019. An important aspect of CER is determining the relative clinical effectiveness of similar prescription drugs. In the absence of head-to-head comparisons that compare drugs with each other within studies, extensions to traditional meta-analytic approaches have been proposed. These multiple treatment comparisons can thus be used to infer the comparative effectiveness of two treatments either through direct head-to-head evidence or through the combination direct and indirect evidence. A number of methods of multiple treatment comparisons are available including Bucher’s method of adjusted indirect comparison, Lumley’s method of network meta-analysis, and Bayesian mixed treatment comparison (MTC). These advances in evidence synthesis approaches provide a way to use the evidence base and compare the effectiveness of drug treatment options simultaneously.

The ISPOR task force recently presented recommendations on the use and application of network meta-analysis. From a regulatory perspective, the Canadian Agency (Indirect Evidence: Indirect Treatment Comparisons in Meta-Analysis) and the UK NICE Evidence synthesis series have recently been published. The US FDA, to enhance regulatory science, has recent projects on comparative effectiveness research. In this roundtable we will discuss the regulatory and statistical issues associated with these recent advances, and if/how that can be used in regulatory applications.

1. What are the expectations and experiences with comparative effectiveness in clinical trials? 2. Are regulatory guidelines sufficiently clear about what is needed to incorporate CER principles in drug applications? 3. What bottlenecks remain and what statistical considerations come into play?