Post hoc reanalysis suggests stereotactic radiotherapy may improve local control in medically inoperable Stage I non-small cell lung cancer

INTRODUCTION: The results of the LUSTRE trial investigating stereotactic body radiotherapy in early-stage non-small cell lung cancer were inconclusive. A Bayesian analysis could potentially address this by integrating prior knowledge. METHODS: The LUSTRE trial was a randomized-controlled superiority trial conducted from 2014 to 2020. Patients with medically inoperable Stage I non-small cell lung cancer were randomized 2:1 to stereotactic radiotherapy (intervention) or conventional radiotherapy (control). The primary outcome was time-to-local control, defined as the period from randomization to absence of a primary tumor or marginal failure during follow-up. Primary analyses used Bayesian multivariable Cox regression, adjusting for strata (tumor size, tumor location, and clinical center). A neutral prior distribution was used for primary analysis, with optimistic (obtained from meta-analyses of 2 previous trials) and pessimistic prior distributions/knowledge for sensitivity analyses. RESULTS: Among 233 patients, 154 received stereotactic radiotherapy, and 79 received conventional radiotherapy. Using a noninformative neutral prior distribution, there was a 91% probability that the posterior adjusted hazard ratio (aHR) for local control with stereotactic radiotherapy was less than 1 (posterior aHR: 0.63, 95% credible interval, 0.32, 1.25). Incorporating the optimistic prior distribution, the posterior aHR was 0.62 (95% credible interval, 0.37, 1.02), indicating a 97% probability of benefit. Under the pessimistic prior distribution, the posterior aHR was 0.97 (95% credible interval, 0.71, 1.34), suggesting a 57% probability of benefit. CONCLUSIONS: Post hoc Bayesian reanalyses provided additional insight into the posterior probability of local control benefit with stereotactic radiotherapy. LUSTRE TRIAL REGISTRY ID: NCT01968941 (date of registration: October 21, 2013).