From objectives to meaningful results: the pivotal role of estimands in analyzing patient reported outcomes data from randomized cancer clinical trials

Music City Center 

Lacking clear PRO objectives results in uncertainty around the optimal approach for measuring PROs in cancer clinical trial adding confusion in their analysis and reporting. This can be observed in the current practice for utilizing patient experience data, typically characterized by inconsistent terminology, inflation of the number of analyses and potentially conflicting results hindering decision-making and cross-validation across studies. Using estimands that reflect relevant and concise PRO objectives can improve design, analysis, interpretation and communication of PRO results. We will highlight and exemplify recommendations in terms of analyses of PROs used in randomized trials based on consensus from various stakeholders. A particular topics is unobserved data which is a persistent problem in PRO analyses due to voluntary patient participation. It can stem from two distinct processes: intercurrent events and missing data. Intercurrent events inherently affect endpoint interpretation and therefore strategies to account for them must be aligned with the intended PRO objective. In particular, how to address terminal events such as death (common in oncology) must be decided a priori in function of the objective. On the other hand, missing data, often a non-random occurrence, requires strategies that consider potential informative relationships and assess overall robustness with respect to the missing data assumptions. of the results to various assumption about the missing data. Reporting the extent of data availability at each assessment timepoint by two standardized different metrics, completion rates and available data rates is recommended. As these rates will differ over time due to patient selection. Attrition and missing data will reduce observed data over time leading to selection bias. This bias impacts common methods where time trends are inferred from repeated cross-sectional analyses presented as a longitudinal series.

randomized clinical trial

patient-reported outcome

estimand

standards

cancer