Liam C Macleod1, Mina M Fam2, Jonathan G Yabes3, Nathan E Hale4, Robert M Turner5, Samia H Lopa3, Jeffrey R Gingrich6, Tudor Borza7, Ted A Skolarus8, Benjamin J Davies5, Bruce L Jacobs5. 1. Department of Urology, University of Pittsburgh Medical Center, Pittsburgh, PA; Department of Urology, Asante Rogue Regional Medical Center, Medford, OR. Electronic address: liamcmacleod@gmail.com. 2. Department of Urology, Jersey Shore University Medical Center, Neptune, NJ. 3. Department of Medicine, University of Pittsburgh, Pittsburgh, PA. 4. Department of Urology, Charleston Area Medical Center, Charleston, WV. 5. Department of Urology, University of Pittsburgh Medical Center, Pittsburgh, PA. 6. Division of Urology, Department of Surgery, Duke University, Durham, NC. 7. Department of Urology, University of Wisconsin, Madison, WI. 8. Division of Oncology, Department of Urology, University of Michigan, Ann Arbor, MI; Dow Division for Urologic Health Service Research, Department of Urology, University of Michigan, Ann Arbor, MI; VA Health Services Research & Development, Center for Clinical Management Research, VA Ann Arbor Healthcare System, Ann Arbor, MI.
Abstract
BACKGROUND: We use observational methods to compare impact of perioperative chemotherapy timing (ie, neoadjuvant and adjuvant) on overall survival (OS) in muscle-invasive bladder cancer because there is no head-to-head randomized trial, and patient factors may influence decision-making. PATIENTS AND METHODS: Using Surveillance, Epidemiology, and End Results-Medicare data, we identified patients receiving cystectomy for muscle-invasive bladder cancer diagnosed between 2004 and 2013. Patients were classified as receiving neoadjuvant or adjuvant chemotherapy. Propensity of receiving neoadjuvant chemotherapy was determined using gradient boosted models. Inverse probability of treatment weighted survival curves were adjusted for 13 demographic, socioeconomic, temporal, and oncologic covariates. RESULTS: We identified 1342 patients who received neoadjuvant (n = 676) or adjuvant chemotherapy (n = 666) with a median follow-up of 23 months (interquartile range, 9-55 months). Inverse probability of treatment weighted adjustment allows comparison of the groups head-to-head as well as counterfactual scenarios (eg, effect if those getting one treatment were to receive the other). The average treatment effect (ie, "head-to-head" comparison) of adjuvant compared with neoadjuvant on OS was not significant (hazard ratio, 1.14; 95% confidence interval, 0.99-1.31). However, the average treatment effect of the treated (ie, the effect if the neoadjuvant patients were to receive adjuvant instead) was associated with a 33% increase in risk of mortality if they were given adjuvant therapy instead (hazard ratio, 1.33; 95% confidence interval, 1.12-1.57). CONCLUSION: Significant treatment selection bias was noted in peri-cystectomy timing, which limits the ability to discriminate differential efficacy of these 2 approaches with observational data. However, patients with higher propensity to receive neoadjuvant therapy were predicted to have increased OS with approach, in keeping with existing paradigms from trial data.
BACKGROUND: We use observational methods to compare impact of perioperative chemotherapy timing (ie, neoadjuvant and adjuvant) on overall survival (OS) in muscle-invasive bladder cancer because there is no head-to-head randomized trial, and patient factors may influence decision-making. PATIENTS AND METHODS: Using Surveillance, Epidemiology, and End Results-Medicare data, we identified patients receiving cystectomy for muscle-invasive bladder cancer diagnosed between 2004 and 2013. Patients were classified as receiving neoadjuvant or adjuvant chemotherapy. Propensity of receiving neoadjuvant chemotherapy was determined using gradient boosted models. Inverse probability of treatment weighted survival curves were adjusted for 13 demographic, socioeconomic, temporal, and oncologic covariates. RESULTS: We identified 1342 patients who received neoadjuvant (n = 676) or adjuvant chemotherapy (n = 666) with a median follow-up of 23 months (interquartile range, 9-55 months). Inverse probability of treatment weighted adjustment allows comparison of the groups head-to-head as well as counterfactual scenarios (eg, effect if those getting one treatment were to receive the other). The average treatment effect (ie, "head-to-head" comparison) of adjuvant compared with neoadjuvant on OS was not significant (hazard ratio, 1.14; 95% confidence interval, 0.99-1.31). However, the average treatment effect of the treated (ie, the effect if the neoadjuvant patients were to receive adjuvant instead) was associated with a 33% increase in risk of mortality if they were given adjuvant therapy instead (hazard ratio, 1.33; 95% confidence interval, 1.12-1.57). CONCLUSION: Significant treatment selection bias was noted in peri-cystectomy timing, which limits the ability to discriminate differential efficacy of these 2 approaches with observational data. However, patients with higher propensity to receive neoadjuvant therapy were predicted to have increased OS with approach, in keeping with existing paradigms from trial data.