I am a clinical researcher, and I wanted to discuss a specific clinical trial design question. A recently reported randomized non-inferiority trial has demonstrated non-inferiority of a shortened (hypofractionated) course of radiation therapy in the primary curative treatment of head and neck squamous carcinoma. The trial had 800 patients and met the non-inferiority criteria for both locoregional control and late toxicity.
We are conducting a similar non-inferiority trial in head and neck squamous carcinoma, but in an expanded cohort of both patients treated primarily with radiation therapy and also patients who receive radiation therapy in the post-operative adjuvant setting, with a quality of life endpoint. However, locoregional control and overall survival are key secondary endpoints. There are 600 patients in this second trial. While the absolute locoregional control may be slightly different in this trial we are really looking at the hazard ratio for the non-inferiority margin
Given the similar (but not identical) patient population, I was wondering
- if a Bayesian design confirmation on non-inferiority of locoregional control can be built into this trial with prior knowledge of the first trial. (it would be such a waste to recruit hundreds of extra patients in a frequentist design)
- How would we calculate power for the Bayesian survival (time-to-event) endpoint in this scenario
- Are there analogous trials which we can learn from, or publications that we should read in this context?
- Would a meta-analysis of the two trials be better than planning a Bayesian approach?
Happy to answer any questions, and thanks in advance for any guidance.