How should overall survival be analysed in randomised clinical trials in cancer if participants receive subsequent treatment lines? A stakeholder consultation

Background Overall survival is used to assess clinical effectiveness in cancer clinical trials. In practice, it may be influenced by intercurrent events post-randomisation. The decisions made on how to address intercurrent events, change the interpretation of the results.

An example is when participants stop their trial intervention and start subsequent anti-cancer interventions (treatment lines) during trial follow-up. At present, there is no evidence on the views of all stakeholders about this intercurrent event or consensus on how it should be addressed. The aim of this work was to understand the perspectives of all stakeholders and to obtain consensus through a qualitative study to guide future methodological work.

Methods A modified Rand/UCLA appropriateness method was implemented. Stakeholder views were collected using an online questionnaire and discussed at a focus group. The questionnaire included items on, the different methods for addressing an intercurrent event, data collection following an intercurrent event, statistical assumptions, and data presentation. Analysis was descriptive incorporating a conventional content approach. Consensus was defined a priori.

Results One hundred three stakeholders (30 statisticians or other data analysts, 6 payers or industry partners, 22 healthcare professionals and 45 patient, carer or members of the public) completed the questionnaire between 3/8/2022 and 30/9/2022. Seventy-nine percent of respondents thought it important to consider the potential effect of subsequent treatment lines.

Consensus was reached on most questionnaire items. Stakeholders agreed that statistical assumptions were applicable only in “Some Scenarios” and that results should be presented using both a visual and summary measure. The focus group discussed different methods for addressing an intercurrent event and items around data collection where consensus was unclear. Seven participants attended (two patients/carers, one healthcare professional, three statisticians and one payer) with K-LR and PW. Attendees agreed that the treatment policy approach should be considered in future work as it was the most realistic, and that data collection was acceptable with informed consent.

Discussion This work demonstrates that all stakeholder groups are interested in how subsequent treatment lines may impact overall survival and provides evidence on what future methodological work in the area should consider. The next step of this work will investigate whether it is possible to estimate the overall survival treatment effect in a hypothetical scenario where participants who received second-line therapy all received the same second-line therapy. This will aim to complement the existing treatment policy approach and quantify the impact of subsequent treatments.