# The REFRACT trial: implementation of Bayesian power priors in a randomised, sequential phase II adaptive platform trial

## TL;DR

The REFRACT trial uses a Bayesian adaptive design to evaluate new treatments for relapsed or refractory follicular lymphoma with fewer patients.

## Contribution

It introduces Bayesian power priors to share control data across sequential treatment rounds in an adaptive phase II trial.

## Key findings

- The trial design allows for sequential evaluation of three novel therapies using adaptive randomisation ratios.
- Control data from earlier rounds are shared and weighted at 75% of active control patients using power priors.
- The design improves operating characteristics while reducing the number of patients needed in the control arm.

## Abstract

REFRACT is a randomised trial aimed at rapidly evaluating multiple novel therapies against standard treatment for relapsed or refractory follicular lymphoma (rrFL) using a minimal number of patients. To this end, we designed a prospective, adaptive, sequentially randomised clinical trial to allow multiple novel therapies to be assessed sequentially against a control arm of investigator choice standard therapy (ICT).

REFRACT uses a Bayesian power priors approach enabling the sharing of control arm data from previous treatment rounds. The design allows for the randomisation ratio to be changed and fixed to 1:4 in later treatment rounds resulting in fewer patients being recruited to the control arm.

Following extensive simulations, we arrived at the selected design of three sequential treatment rounds, each with a control group and a novel experimental arm assessed for the primary outcome of complete metabolic response (CMR) at 24 weeks. Patients in Round 1 are randomised using a 1:1 allocation, with Rounds 2 and 3 randomised using a 1:4 allocation, in favour of experimental treatment. Using Bayesian power priors, data from control patients in earlier rounds will be shared to improve the operating characteristics in the current round. Previous control arm patients will be weighted at 75% of an active control patient within the prior, with opportunities for adjustment should control treatments change over time.

With the use of power priors and an adaptive design this trial will sequentially evaluate three novel treatment regimens in a disease that urgently requires additional treatment options. REFRACT opened to recruitment in July 2023.

EudraCT: 2022–000677-75; 10-Feb-2022.

ClinicalTrials.gov: NCT05848765; 08-May-2023.