Conditional survival after radical cystectomy for bladder cancer: evidence for a patient changing risk profile over...

Standard survival statistics do not take into consideration the changes in the weight of individual variables at subsequent times after the diagnosis and initial treatment of bladder cancer.

To assess the changes in 5-yr conditional survival (CS) rates after radical cystectomy for bladder cancer and to determine how well-established prognostic factors evolve over time.

We analyzed data from 8141 patients treated with radical cystectomy at 15 international academic centers between 1979 and 2012.

Radical cystectomy and pelvic lymph node dissection.

Conditional cancer-specific survival (CSS) and overall survival (OS) estimates were calculated using the Kaplan-Meier method. The multivariable Cox regression model was used to calculate proportional hazard ratios for the prediction of mortality after stratification by clinical characteristics (age, perioperative chemotherapy status) and pathologic characteristics (pT stage, grade, lymphovascular invasion, pN stage, number of nodes removed, margin status). The median follow-up was 32 mo.

The 5-yr CSS and OS rates were 67.7% and 57.5%, respectively. Given a 1-, 2-, 3-, 5- and 10-yr survivorship, the 5-yr conditional OS rates improved by +5.6 (60.7%), +8.4 (65.8%), +7.6 (70.8%), +3.0 (72.9%), and +1.9% (74.3%), respectively. The 5-yr conditional CSS rates improved by +5.6 (71.5%), +9.8 (78.5%), +7.9 (84.7%), +7.2 (90.8%), and 5.6% (95.9%), respectively. The 5- and 10-yr CS improvement was primarily noted among surviving patients with advanced stage disease. The impact of pathologic parameters on CS estimates decreased over time for both CSS and OS. Findings were confirmed on multivariable analyses. The main limitation was the retrospective design.

CS analysis demonstrates that the patient risk profile changes over time. The risk of mortality decreases with increasing survivorship. The CS rates improve mainly in the case of advanced stage disease. The impact of prognostic pathologic features decreases over time and can disappear for long-term CS.