FIG 3.

Intention-to-treat analysis of overall survival (OS) by randomly assigned treatment. As of December 11, 2016, 76 deaths were reported. The median follow-up time was estimated to be 53 months. There was insufficient evidence to reject the null hypothesis of no superiority of vaginal cuff brachytherapy plus three cycles of carboplatin and paclitaxel chemotherapy (VCB/C) over radiation therapy (RT) with respect to OS. The log-rank test statistic was −0.756 (one-tailed test P = .57). The estimated treatment hazard ratio (HR) was 1.04 (regimen II relative to regimen I). The (1-α) × 100% Wald CI was 0.664 to 1.632 for a two-sided α = 0.05 (0.025 in each tail) and 0.713 to 1.518 for a two-sided α = 0.10 (0.05 in each tail). An effect size of 0.51 (49% decrease in hazard) was not contained in these CIs.