All Times EDT
Keywords: estimands, time-to-event, patient-reported outcomes, health-related quality of life, competing risk
Health-related quality of life (HRQoL), specifically symptom burden and functional status, contribute to the interpretation of progression-free survival. Recent regulatory guidance on the incorporation of symptoms, function and tolerability in anti-cancer clinical trials elevates the patient experience in the endpoint strategy. Patient-focused time-to-event endpoints align directly to survival endpoints, but require special consideration in their definition due to the subjective nature of the measures (eg., PROs). Particular attention to the definition of events, including intercurrent events, and estimation approach is necessary. Additional issues include cancer type (e.g., metastatic) and treatment modality (e.g., chemotherapy vs immuno-therapy). Defining intercurrent events is critical to clearly evaluate of time-to-event endpoints in oncology, as they are often related to treatment tolerability leading to discontinuation or switching. Statistically, how does one best estimate a time-to-HRQoL endpoint in the presence of intercurrent events that alter the patient journey? When the events due to other causes reduce the possible number of actual events due to the cause of interest, how do we best estimate time-to-event? One method is a competing risks model using the cumulative incidence function and sub-distribution hazard functions to estimate the cause-specific risk. Another approach is to consider those who are censored due to a competing event as still at risk for the event with post-censoring observations coming from the distribution of people who have not experienced that event. Establishing ways in which to handle intercurrent events outside the event of interest in this context helps to clarify estimands for time-to-event of PRO endpoints. We present challenges of defining estimands for therapeutic decision making and accurately estimating the risk of HRQoL endpoints. We illustrate the variability of estimands and related estimators through case-studies.