Using real world evidence to assess clinical outcomes in patients diagnosed with bladder cancer

Introduction: The current 10-year survival rate for patients with bladder cancer stands at around 50%. It is imperative that the prognostic value of factors surrounding treatments and demographics for bladder cancer patients are identified to allow for improvements in these survival rates. To that end, this thesis consists of the following three projects: 1. Neoadjuvant chemotherapy for muscle-invasive bladder cancer: an investigation on survival; 2. Effect of a delay in radical treatment on survival in bladder cancer patients; 3. A mediation analysis to explain socio-economic differences in bladder cancer survival. Methods: This thesis utilised a mixture of analyses to investigate the three projects which were carried out using both single-centre and nationwide data from two databases: The Guy’s and St Thomas’ Hospital cystectomy database, and the BladderBaSe. Project 1 utilised Cox regression models on data from both databases for all patients with clinical T2-T4, N0, M0 disease. Propensity score matching was also carried out for the BladderBaSe data. The second project was carried out through a systematic review and meta-analyses to collate as much information regarding delay to cystectomy as possible. The second half of this project consisted of a survival analyses conducted on data extracted from the BladderBaSe. Cox proportional hazards models were carried out to assess the impact of a longer time between TURBT and either NAC or radical cystectomy. Project 3 also consisted of a systematic review and was followed by both survival and mediation analyses in order to first ascertain a relationship between SES and survival, and then to disentangle the heterogeneity in these survival outcomes by identifying any potential mediators of the relationship. Results: Project 1: From the 944 patients analysed in the BladderBaSe study, those who had received NAC had higher 5-year survival proportions and decreased risk of both overall and bladder cancer-specific death (HR=0.71, 95% CI: 0.52–0.97 and HR=0.67, 95% CI: 0.48–0.94), respectively, as compared to patients who did not receive NAC. The PS matched cohort showed similar estimates, but with larger statistical uncertainty (Overall death: HR=0.76, 95% CI: 0.53–1.09 and bladder cancer-specific death: HR=0.73, 95% CI: 0.50–1.07). In the Guy’s and St. Thomas cystectomy data, an absolute survival benefit of 18% was observed in the NAC patients (n=62) when compared to those who did not receive NAC (n=172). The Cox models also revealed a reduced risk of death in the NAC patients (HR=0.57, 95% CI: 0.36-0.92), compared to those who did not receive NAC. Project 2: The systematic review identified nineteen studies which were eligible for inclusion (17,532 patients). Ten of these studies were included in the meta-analyses for which a longer delay between bladder cancer diagnosis and radical cystectomy resulted in a pooled hazard ratio of 1.34 (95% CI: 1.18-1.53) for overall death. For delay between TURBT and cystectomy, the pooled hazard ratio was 1.18 (95% CI: 0.99-1.41) for overall death. A pooled hazard ratio of 1.04 (95%CI: 0.93-1.16) was calculated for a longer delay between NAC and radical cystectomy. The BladderBaSe study identified no clear pattern of association between patients who experienced a longer time between TURBT and radical cystectomy (n=2,160 patients). The Cox models suggested an increased risk of bladder cancer-specific death in patients who had a longer time between TURBT and receiving NAC (HR=3.45, 95% CI: 1.11-10.73). Project 3: The AFT models revealed patients (both NMIBC and MIBC) with a medium or high SES had an increased survival both overall and bladder cancer-specific when compared to patients with a low SES (n=37,755 patients). In the NMIBC patients, Charlson Comorbidity Index was found to mediate this relationship by 10% (percentage of the total effect explained by mediator) and hospital type by 4%. The time from referral to TURBT was a considerable mediator (14%) in the MIBC patients only. The subsequent systematic review identified a wealth of studies investigating the relationship between each of the proposed mediators with survival in bladder cancer patients. There was however found to be a paucity in studies investigating the associations between said mediators and SES in bladder cancer specifically. Conclusions: This thesis has used real world evidence to help explore factors which affect the survival of bladder cancer patients. The effect of NAC has been extensively studied in RCTs but evidence using observational data is somewhat lacking. However, results from this thesis suggest its regimen imposes a survival benefit in patients even in a real world setting and future studies should focus on identifying exactly which patients benefit most. This thesis has also highlighted the complexity of the relationship between treatment delay and survival for those in need of radical cystectomy. Though the meta-analyses imply a negative impact on survival when radical treatment is delayed, the effect of this delay in patients receiving NAC requires more investigation. Finally, the associations observed between SES and survival using Swedish data for the first time are in line with existing literature. The current results suggest this relationship is contributed to by several factors. However, as the theoretical causal relationship could not be fully explained by the proposed mediators, more factors still need to be identified as further potential mediators. To the best of our knowledge this is the first time mediation analysis has been used to investigate the association between SES and survival in bladder cancer. This thesis has benefitted from the use of real word data to investigate questions not possible through the use of RCTS (i.e. the effect of a delay in cystectomy). Overall, it is hoped the results from this thesis can be used to build upon the knowledge of prognostic factors for patients with bladder cancer. Moreover, the results identify a need for a standardization of data collection methods for observational databases.