Abstract
This paper presents a reanalysis of a randomized clinical trial conducted by the Cancer and Leukemia Group B (CALGB, Bethesda, MD, USA). This trial found a significant benefit of combination chemotherapy followed by irradiation (CTRT) in comparison to radiotherapy alone (RT) for the treatment of nonsmall cell lung cancer. The validity of the results obtained and the decision to terminate taken by the CALGB, were assessed using sequential methods. The reliability and efficiency of sequential methods were also assessed for this study. Two sequential designs were used: the triangular and the restricted procedure. Initial analyses were conducted with the data from patients actually recruited, adjusting for important prognostic variables at any interim analysis. As a confirmatory technique, a continuation of the trial was simulated, sampling extra patients under the assumption of no treatment difference, preserving the effect of the prognostic variables. Using the results from the 155 patients recruited by the CALGB (88 deaths at termination and 136 after follow-up), the sample path stayed within the continuation region of both sequential designs considered. An underpowered sequential analysis showed significant superiority of CTRT over RT (95% confidence interval (95% CI) 0.50-0.96, p=0.03 for the triangular; 95% CI 0.37-0.88, p=0.01 for the restricted procedure). Conventional analysis of the follow-up data also showed significant superiority of CTRT. The trial extended with simulated data ended at 60 months with 251 patients (178 deaths), showing significant superiority of CTRT under both designs (95% CI for hazard ratio 0.55-0.97). The two sequential procedures would have led to the same conclusion as that reached by the Cancer and Leukemia Group B, still achieving considerable savings in patients recruited and time over the conventional design. The data simulated under the rather conservative null hypothesis did not reverse the positive result claimed by the Cancer and Leukemia Group B.
