How to design a ROCI (Response Over Continuous Intervention) randomised trial: guidance and a case study

An important challenge in late-phase drug development is selection of the optimal parameters for treatment administration, including dose, duration and frequency. Whilst these are (near-)continuous parameters, conventional non-inferiority trial designs are not well suited to explore multiple options across a range of values. The ROCI (Response Over Continuous Intervention) randomised trial design is an alternative to standard non-inferiority trials, addressing limitations in selecting optimal treatment regimens, but practical guidance in its application is limited. We outline key design considerations for ROCI trials, including modelling the treatment–response relationship, selecting treatment levels and arms, defining optimality criteria, and determining power and sample size. A flexible fractional polynomial approach is recommended to estimate the treatment–response relationship, and two power definitions—optimal and acceptable—are considered. We apply these principles to REFINE-Lung, which investigates whether extended pembrolizumab dosing intervals in advanced non-small cell lung cancer (NSCLC) maintain efficacy while reducing unnecessary overtreatment. Sample size calculations use simulation-based methods or an analytic formula for binary outcomes. REFINE-Lung was designed with five administration frequency arms, exploring the range within which the optimal frequency is likely to exist (6-, 9-, 12-, 15-, and 18-weekly dosing), with the primary endpoint 2-year overall survival. Initial sample size calculations targeted 1,750 patients to provide 80% optimal power, but in order to accelerate completion this was adjusted to 1,100 patients targeting acceptable power. An interim analysis strategy allowed adaptation based on early data on progression-free survival. This approach balances statistical robustness with feasibility while ensuring ethical equipoise in dose de-escalation. The ROCI design is an efficient alternative to standard non-inferiority trials for optimising continuous aspects of treatment administration. By sharing information across multiple treatment levels and leveraging flexible modelling approaches, ROCI trials improve efficiency and reliability in identifying optimal treatment regimens. The REFINE-Lung trial case study demonstrates applying these principles, providing a framework for future trials aiming to refine treatment administration strategies.