A natural history model of oesophageal adenocarcinoma to inform power calculations for the BEST4 screening trial

Objective Oesophageal adenocarcinoma (OAC) has a poor prognosis, often due to its late stage at diagnosis. Individuals diagnosed with Barrett's oesophagus (BO), a precursor condition to OAC, are offered surveillance with the goal of identifying and treating dysplastic BO, preventing progression to cancer. The BEST4 Screening randomised controlled trial was designed to demonstrate a reduction in (i) incidence of stage II + or fatal OAC and (ii) OAC mortality through targeted capsule sponge screening in high-risk individuals. Here we describe the simple natural history model we developed to inform the sample size calculation for the BEST4 screening trial. Methods We considered a range of annual transition probabilities between states to evaluate the impact on the incidence of (i) stage II + or fatal OAC and (ii) OAC mortality, and therefore the statistical power to identify a reduction in each of these endpoints. We additionally considered a range of baseline prevalences for each state, as well as for screening uptake, capsule sponge sensitivity and treatment efficacy. Results Under the assumptions considered, the BEST4 trial will have >90% power to detect a difference in incidence of stage II + or fatal OAC by 5.5 years, and mortality by 7.5 years, under most parameter combinations, and by 8.5 years and 12 years, respectively, under all combinations. The model is freely available to download. Conclusion A simple semi-Markov natural history model can be used to explore the power of a clinical trial with group-sequential analysis under a wide range of scenarios regarding the natural history of disease.