John Maringwa 1
1 Johnson & Johnson (, Netherlands)
Introduction:
Intermediate risk non-muscle invasive bladder cancer (IR-NMIBC) is a heterogeneous disease with variable clinical outcomes, making selection of appropriate therapeutic options for this population often challenging. Standard management of NMIBC typically includes initial tumor resection (transurethral resection of bladder tumor [TURBT]), often followed by intravesical chemotherapy or Bacillus Calmette-Guérin (BCG) immunotherapy [1][2] depending on risk stratification. The global BCG shortage has led to BCG prioritization of high risk (HR)-NMIBC patients. Intravesical chemotherapy is often used as the control group in studies investigating novel treatments. A network meta-analysis of 28 trials with chemotherapy as the common comparator [3] identified a wide range of chemotherapy treatments as control. Reliable quantitative benchmarks for disease-free survival (DFS) for these therapies are limited. This analysis aimed to quantify DFS associated with Mitomycin C (MMC) or gemcitabine using MBMA to inform control-arm assumptions, trial design, and hazard projections.
Methods:
A total of 161 studies in the NMIBC systematic literature review database published from 1990 through July 2025 were reviewed by a cross-functional team (clinicians, statisticians, data scientists). Of these, 46 studies included IR or a mixture of IR and HR-NMIBC patients. Fourteen studies reporting MMC data (19 arms overall; N=2,294) and 3 studies reporting gemcitabine data (3 arms overall; N=294) were included. Digitized Kaplan-Meier DFS curves were used to develop a DFS model assuming a Weibull distribution with treatment-specific parameters characterizing the hazard of disease recurrence, progression, or death. Inter-study variability was accounted for through random effects. The effect of risk factors (% of patients with grade 3 NMIBC, Stage T1, etc.) on the hazard was assessed using a proportional hazards model. Plots of reported and model-projected DFS were used to evaluate model adequacy. Numerical metrics including percent relative standard error (%RSE), i.e., the ratio of the standard error (SE) to the estimate of a parameter were also used for model evaluation. The expected 2- and 5-year DFS rates (95% prediction interval (PI); confidence interval (CI)) were simulated as a function of different risk factor values as applicable. The cumulative hazard in 3, 6, and 12-month intervals was calculated, stratified by treatment, to help assess whether the total risk accumulated within these respective time intervals was constant over time.
Results:
The model-projected 2-year and 5-year DFS rates were 68% (95% PI:31–86; CI:60-74) and 46% (9.4–74; 36-54)) for MMC, and 65% (45–91; 52-75) and 42% (20–83; 27-56) for gemcitabine, respectively. Although on average, the estimated DFS rates for MMC were numerically larger, the two chemotherapy options were comparable, as suggested by the overlapping PIs. The estimated hazard rates in successive yearly intervals decreased for both treatments, indicating that the risk of recurrence declines over time. For MMC, cumulative hazards (95%PI; CI) in years 1, 2, and 3 were 0.22 (0.080-0.60; 0.17–0.29), 0.16 (0.058-0.44; 0.12–0.21), and 0.14 (0.051-0.39; 0.11–0.18); corresponding values for gemcitabine were 0.24 (0.087-0.69; 0.16–0.36), 0.18 (0.064-0.51; 0.12–0.27), and 0.16 (0.056-0.45; 0.11–0.23). Overall, the cumulative hazard of recurrence, progression, or death in years 1, 2 and 3, were comparable between MMC and gemcitabine, commensurate with the similar landmark DFS rates. The effect of risk factors investigated was either minimal or could not be identified owing to prevalence of missing (unreported) information. Further exploration is ongoing.
Conclusion:
This MBMA provides quantitative DFS benchmarks for MMC and gemcitabine in IR-NMIBC patients and supports their use as comparable control therapies. The observed decrease in overtime in hazards suggests that assuming a constant hazard (i.e., exponential model) may be less robust for trial design in this setting and this suggests that study design should account for the piecewise exponential behavior of the DFS curve. These results can inform control-arm assumptions, sample-size calculations, and event projections for future trials in NMIBC.
References:
1. Chang, S. S., S. A. Boorjian, R. Chou, et al. (2016). “Diagnosis and Treatment of Non-Muscle Invasive Bladder Cancer: AUA/SUO Guideline.” J Urol 196(4): 1021-1029.
2. Flaig, T. W., P. E. Spiess, N. Agarwal, et al. (2018). “NCCN Guidelines Insights: Bladder Cancer, Version 5.2018.” J Natl Compr Canc Netw 16(9): 1041-1053.
3. Boehm, B. E., J. E. Cornell, H. Wang, et al. (2017). “Efficacy of bacillus Calmette-Guerin Strains for Treatment of Nonmuscle Invasive Bladder Cancer: A Systematic Review and Network Meta-Analysis.” J Urol 198(3): 503-510.
4. https://www.clinicaltrials.gov/study/NCT06319820?tab=study
Reference: PAGE 34 (2026) Abstr 11851 [www.page-meeting.org/?abstract=11851]
Poster: Drug/Disease Modelling - Oncology