PURPOSE: The EMA have recently published document ICH E9 R1 outlining the need to proactively link design, analysis, and interpretation of PRO data in clinical trials. We discuss an iterative approach to designing, analyzing and interpreting trial data, rather than the stepwise approach that is often used.

DESCRIPTION: First, we will discuss trial design processes, including situations and examples, where proactively and iteratively linking trial design with analysis can have clear benefits. We will also discuss challenges where it is either not practical or possible to identify a priori clear links between design and analysis. The role of patient input in identifying meaningful PRO endpoints will also be raised.

Next, we will describe a variety of analytic techniques that can be planned a priori to better understand PRO data, particularly in situations with intensive data collection (e.g., diaries) and large amounts of potentially informative missing data. A few examples will be shared, including pattern mixture models, extended pattern mixture models, and joint models. In addition, more complicated relationships often exist between treatment and PRO outcomes (i.e., indirect/mediated effects), but traditional analyses often do not capture them. We will describe alternative, robust, structural equation modeling approaches for examining these complex relationships and present recent examples from oncology trials to demonstrate their power and informativeness.

Finally, we will discuss the impact of linking trial design and analysis from an evaluator’s perspective: how can we get initial insights from regulators/HTA agencies and anticipate or at least identify potential biases, and analyze, interpret, and present results in a submission dossier that will increase success for drug approval or reimbursement?

There will be polling throughout this workshop to ask the audience their priorities, experiences, and challenges when trying to minimize bias; the pros/cons of different approaches; and an opportunity for questions at the end.