The trial will use a Bayesian approach to estimation and decision-making in each treatment arm.
The statistical analysis plan will be a simple beta-binomial conjugate analysis combining the observed trial data with a minimally-informative prior distribution. The posterior probability distribution will be used to:
(i) estimate the true efficacy rate with 95% credible intervals to indicate the level of uncertainty and
(ii) determine whether the probability that the true efficacy rate is greater than a clinically relevant critical value is sufficiently high (>0.80) to declare the treatment worthy of consideration (referred to as a GO decision).
Each treatment arm (or sub-group) will initially aim to recruit 15 patients. At regular intervals, an independent Data Monitoring Committee (DMC) will assess the data for futility and efficacy. At these stopping points, a posterior probability distribution will be used to calculate the predicted probability of success given the current observed data.
Should a treatment arm be deemed worthy of further consideration based on the analysis of 15 patients, then there will be an option to expand recruitment to a larger number of patients to provide evidence to support a marketing authorisation application.
Glo-BNHL trial will aim to recruit 30 patients per year (across all arms) with recruitment for 7 years.