CTNI-16. MONOGERM, A NOVEL PROOF-OF-PRINCIPLE BAYESIAN PHASE II TRIAL DESIGN OF CARBOPLATIN OR VINBLASTINE MONOTHERAPY INDUCTION PRIOR TO RADIOTHERAPY FOR INTRACRANIAL GERMINOMA

Abstract BACKGROUND Current European standard-of-care for localised intracranial germinoma is multi-agent chemotherapy (carboPEI: carboplatin/etoposide/ifosfamide) followed by definitive radiotherapy, with excellent survival. MonoGerm is a de-escalation, non-inferiority trial aiming to reduce toxicity. Twelve-week carboplatin (PMID:8039122) AUC10 or vinblastine (PMIDs:32642701/34520101) induction will be evaluated to test if as effective as carboPEI from SIOP-CNS-GCT-II. A novel trial design was required to answer this question pragmatically/safely. METHODS Clinical trials in rare diseases recruit slowly, allowing continuous monitoring of efficacy outcomes. Efficacy-transition-pathways (ETP) are innovative visual tools to aid determination of trial design parameters, and an extension of the dose-transition-pathways concept introduced for dose-finding trials (PMID:28733440). RESULTS MonoGerm includes two monotherapies, with each single arm recruiting six cohorts of three patients, with interim assessment after each recruited cohort and final analysis at 18 patients (total n=36). Insufficient tumour volume response (<30%) at 6-week safety MRI results in 12-weeks carboPEI. Primary outcome is radiological complete response (CR) by 12-weeks of induction monotherapy. A beta-binomial conjugate analysis will generate posterior probability distributions, combining observed trial data as realisations from a binomial distribution with a minimally informative Beta (1,1) prior. Decision criteria to allow early stopping at interim analyses and go/no-go decisions at final analysis are based on probabilities from these posterior distributions. ETP visually maps out parameters used to assert decisions after each interim assessment as a pyramid decision tree. For each recruited cohort and every CR outcome, estimates of the true CR rate and probabilities with associated decisions are mapped out. ETP allows clear communication between statisticians, clinicians, and patient-public-involvement (PPI) teams, facilitating informed decisions in an efficient/realistic trial design. CONCLUSION MonoGerm, a novel Bayesian de-escalation trial, funded by Little Princess Trust (https://www.littleprincesses.org.uk/), uses ETP and continuous monitoring with built-in stopping rules to ensure patient safety in this treatment de-escalation trial.