Recursive Partitioning Analysis to Define Prognostic Groups Using a Large Single Institution Experience of Multimodality Radiation Therapy for Bladder Cancer

Abstract

Listen

Translate article

<h3>Purpose/Objective(s)</h3> We analyzed a cohort of patients who received definitive trimodality therapy (TMT) for bladder cancer at a single high-volume institution. We hypothesized that a recursive partitioning analysis (RPA) including patient and tumor characteristics would allow stratification of patients into clinically meaningful prognostic groups for local recurrence. <h3>Materials/Methods</h3> A retrospective cohort analysis was performed for bladder cancer patients treated with definitive radiation or chemoradiation therapy from 2006 to 2021 at our institution. We assessed overall and disease-free survival for the cohort and as a function of clinical and pathologic features. We performed an RPA for multivariate decision tree analysis to associate outcomes with individual clinical and pathologic variables. <h3>Results</h3> 175 patients were identified who were treated with definitive radiation (28%) or chemoradiation (72%) and had adequate follow-up data for analysis. Median follow up time was 23 months (Median age=78 [IQR 71-84], 77% male). While Grade 1-2 toxicities were common, Grade ≥3 toxicities were rare and there were no treatment related deaths. Cumulative 2yr and 5yr CSS were 79% and 61%, respectively. Radical cystectomy (RCx) eligible patients (35%) had a significantly longer OS after bladder sparing therapy compared to RCx ineligible patients (5y OS 54% vs 29%, Log-Rank p=0.003), however no difference in cancer specific survival was seen between groups (5y CSS 61% vs 58%, Log-Rank p = 0.851). Tumor size greater than 5cm did not correlate with risk of recurrence (Log-Rank p=0.194). The most common site of recurrence in the cohort was distant (35%), followed by locoregional (19%) and bladder only (6%). Nine patients (5%) underwent salvage cystectomy (7 for invasive local recurrence) after initial definitive radiation therapy. Using RPA and decision tree analysis, the presence of CIS and older age at treatment were the strongest predictors of local recurrence. The RPA results confirmed no significant association between tumor size and risk of local recurrence. <h3>Conclusion</h3> Multimodality bladder preservation had CSS comparable to modern surgical series and high rates of functional organ preservation. Cystectomy candidacy was a strong predictor of overall survival but not cancer specific survival, suggesting group level mismatches in competing risks of death. Extensive CIS was associated with increased risk of local failure, but tumor size may not be a reliable predictor of treatment response. RPA prognostic groups were generated which may be used to guide clinical decision making and improve patient selection for bladder preservation.