Frangou, Eleni; Choodari-Oskooei, Babak; Parmar, Mahesh; Meade, Angela; RAMPART Trial Team; (2024) Navigating through the re-design of a multi-arm, multi-stage (MAMS) trial in light of external data. Presented at: International Conference of Trial Methodology (ICTMC) 2024, Edinburgh, UK.

Preview

Text ICTMC2024_Poster_V1.0_24Sep2024.pdf - Published Version Download (373kB) | Preview

Abstract

Introduction: Clinical trials to establish the efficacy of new agents in the adjuvant cancer setting typically take many years to complete. External evidence along the way can impact recruitment and reporting plans. The RAMPART MAMS, international, platform trial is testing the PD-L1 inhibitor durvalumab (arm B) and in combination with the CTLA-4 inhibitor tremelimumab (arm C) against active monitoring (arm A, standard of care), in patients with resected renal cell carcinoma (RCC). It opened to accrual in 2018. In late 2022, a change in the standard of care for many RCC patients based on the results of the KEYNOTE-564 trial meant it was no longer feasible to continue to randomise to RAMPART and recruitment was closed earlier than planned. Can we, then, modify the design of the trial while maintaining its integrity and achieve similar power as originally planned? We present the challenges and solutions we employed to make the most of the available randomised evidence in RAMPART. / Methods: We asked the Independent Data Monitoring Committee to give permission to use control arm data to model the survival distribution and inform revised timelines. Without any knowledge of the research arms’ data, we modified the assumed control arm event rate and target effect size. As the primary outcome is disease-free survival, we used Stata’s nstage, artpep and artsurv to update the sample size and analysis timelines. In the modified design, we replaced the planned interim analyses with a combined superiority analysis (B and C vs A) to precede the individual arm comparisons and permit an earlier read-out of potential efficacy. Analyses by arm will follow several years later. To obtain the critical value for the pairwise comparisons and control the FWER, we used Dunnett’s method and modelled the survival distributions using flexible parametric models which do not require proportional hazards. / Results: We, successfully, predicted the timing of the first superiority analysis based on the revised number of control arm events. By targeting a larger treatment effect, we maintained the overall power to 80% and the FWER at 2.5% (one-sided). / Discussion: Early recruitment cessation resulted in a total sample size 45% of that originally planned. Our goal was to make the most of the available randomised evidence when the trial in its original format would not have been viable. By changing the targeted effect size, we maintain similar power for our planned analyses.

Download activity - last month

Export as JSONCSVXML

Download activity - last 12 months

Export as CSVXMLJSON

Downloads by country - last 12 months

Export as XMLCSVJSON

Archive Staff Only

View Item