Network meta-analysis of immune checkpoint inhibitors in BCG-naïve, high-risk non-muscle-invasive bladder cancer

Bacillus Calmette-Guérin (BCG) remains the standard first-line therapy for BCG-naïve high-risk non-muscle-invasive bladder cancer (NMIBC), yet recurrence and progression remain clinically significant. Multiple phase III trials have evaluated immune checkpoint inhibitors (ICIs) combined with BCG, but the relative efficacy of individual agents and the influence of carcinoma in situ (CIS) remain unclear. We performed a network meta-analysis (NMA) to compare ICI-based regimens and assess treatment ranking, with CIS stratification.

A frequentist NMA was performed using phase III randomized trials comparing ICI plus BCG vs. BCG alone in BCG-naïve high-risk NMIBC. Included trials were ALBAN (atezolizumab plus BCG), Combination of sasanlimab and alternative BCG Regimens to Evaluate outcomes with Subcutaneous anti-PD-1 Treatment (CREST) (sasanlimab plus BCG), and POTOMAC (durvalumab plus BCG). The primary endpoint was event-based disease control, defined as event-free survival (EFS) or disease-free survival (DFS), which were analyzed jointly given their comparable definitions across trials. Hazard ratios (HRs) and 95% confidence intervals (CIs) were extracted using BCG as the reference comparator. Treatments were ranked using P-scores. Sensitivity and CIS-stratified analyses were conducted.

Durvalumab plus BCG (HR 0.68; 95% CI 0.50-0.93) and sasanlimab plus BCG (HR 0.68; 95% CI 0.49-0.94) significantly improved event-based outcomes compared with BCG alone, whereas atezolizumab plus BCG did not (HR 0.98; 95% CI 0.71-1.36). Durvalumab and sasanlimab ranked highest by P-score (0.81 each), whereas atezolizumab (0.22) approximated BCG alone (0.16). CIS-stratified analyses were hypothesis-generating and revealed differences in event-based rankings, with durvalumab ranking highest in CIS-absent disease and sasanlimab ranking highest in CIS-present disease. Results were consistent across sensitivity analyses. Grade ≥3 TRAEs were more frequent with ICI plus BCG than BCG alone (21%-29% vs. 3.8%-8.8%), while CREST and POTOMAC reported no clinically meaningful decline in health-related quality of life with combination therapy.

In this NMA, durvalumab plus BCG and sasanlimab plus BCG significantly reduced events vs. BCG alone (HR 0.68 each), whereas atezolizumab plus BCG did not (HR 0.98; 95% CI 0.71-1.36). Pairwise indirect comparisons between active regimens were not statistically significant; point estimates and P-scores are consistent with a 2-tier efficacy pattern but do not establish between-agent differences. CIS-stratified subgroup analyses are exploratory and require confirmation.