Estimating Bayesian priors for a non-inferiority randomized trial of gemcitabine-based chemoradiation for muscle invasive bladder cancer.

Abstract

575 Background: Gemcitabine (gem) is a convenient radiosensitizer for muscle invasive bladder cancer (MIBC). However, the effectiveness and toxicity of gem have not been studied in a randomized control trial (RCT) against non-gem radiosensitization. This analysis aims to generate informative Bayesian prior distributions for effectiveness and toxicity outcomes to be used in a non-inferiority RCT of gem versus non-gem radio-sensitizers. Methods: We extracted all patients (pts) treated with curative intent concurrent chemoradiotherapy for node negative MIBC at a single institution between 2010-2022 and divided into gem vs non-gem cohorts. To minimize impact of confounding, we used freedom from distant metastases (FFDM) as our primary measure of effectiveness, censoring for both loss to follow up and death. We used a non-informative uniform prior to estimate the posterior distribution of hazard ratios (HR) for FFDM. We used a non-informative Jeffrey’s prior to estimate the posterior distribution of odds ratios (OR) for dichotomized acute grade 2 or higher gastrointestinal (G2+ GI) and genitourinary (G2+ GU) toxicities. We adjusted for age, T-stage, prior non-MIBC, neoadjuvant chemotherapy in all analyses. Results: A total of 106 pts were included (50 = gem, 56 = non-gem, of which 27 [48%] received mitomycin/5FU); 81% had cT2 disease and median radiation dose was 6480 cGy (range 4500-6480 cGy). Compared to pts in non-gem cohort, pts in gem cohort were older (mean age: 80 vs 76), had less advanced primaries (T3-T4: 14% vs 23%), were less likely to get neoadjuvant chemotherapy (12% vs 29%), were less likely to have prior non-MIBC (10% vs 13%), were less likely to complete full course radiotherapy (88% vs 94.6%), and had shorter follow-up (median 1.3 versus 2.3 years). The 2-year FFDM for the gem and non-gem were 62% and 66%. Table 1 shows rates of acute GI/GU toxicity. The median posterior HR (high probability density (HPD)) for FFDM was 1.54 (0.63-2.86). The OR (HPD) for grade 2+ GI toxicity was 2.4 (HPD 0.78-5.44) and grade 2+ GU toxicity was 0.91 (HPD 0.29-1.82) in gem vs non-gem cohorts. Conclusions: Our analysis suggests skeptical priors for FFDM (50% chance of at least 54% worse HR) and acute G2+ GI toxicity (50% chance of at least 140% worse OR) toxicity and a neutral prior for acute G2+ GU toxicity for a non-inferiority RCT of gem vs non-gem radiosensitization. Validation in a second institutional cohort will be available for our presentation. [Table: see text]