The NIAGARA trial demonstrated improved event-free survival (EFS) with perioperative durvalumab plus gemcitabine-cisplatin (GC) vs. GC alone in muscle-invasive bladder cancer (MIBC). This analysis applied mixture cure models to explore whether this EFS benefit may reflect an increase in the proportion of patients achieving long-term event-free status, complementing standard hazard ratio analyses.
Reconstructed individual patient data (rIPD) were derived from published Kaplan-Meier curves. Parametric mixture cure models (Weibull, log-logistic, and log-normal) were fitted separately by treatment arm, with model selection based on the Akaike Information Criterion (AIC). Cure regression models decomposed treatment effects into cure and latency components. Sensitivity analyses assessed robustness to reconstruction uncertainty and follow-up duration.
The rIPD reproduced the published trial results with reasonable concordance (hazard ratio 0.68 vs. reported 0.68). The log-normal model provided the best fit. The estimated cure fraction (π) was 41% (95% confidence interval [CI]: 32%-48%) for the control arm and 60% (95% CI: 55%-65%) for the durvalumab arm, yielding a model-based difference (Δπ) of +19 percentage points (pp) (95% CI: +10 to +29). Akaike weights favored a cure-effect-only model (71%) over a combined cure-plus-latency model (29%), with a latency-only model receiving negligible support (<1%). Perturbation sensitivity analyses indicated that these bootstrap intervals underestimate total uncertainty, with individual-run estimates spanning negative values at plausible reconstruction noise levels.
In this exploratory analysis of reconstructed individual patient data, perioperative durvalumab was associated with an estimated +19 pp higher cure fraction (95% CI: +10 to +29); cure regression assigned an Akaike weight of 71% to a pure cure fraction mechanism vs. 29% to a combined cure-plus-latency mechanism. These estimates depend on the reconstructed dataset and on the assumption that the EFS curves have plateaued within the observed follow-up; perturbation sensitivity analyses indicated that at plausible reconstruction noise (SD ≥ 0.5 months) the range of individual-run estimates spans negative values.
Urologic oncology. 2026 Jun 05 [Epub ahead of print]
Shugo Yajima, Soichiro Yoshida, Wei Chen, Hiroshi Fukushima, Hajime Tanaka, Hiroyuki Sato, Akihiro Hirakawa, Hitoshi Masuda, Yasuhisa Fujii
Department of Urology, National Cancer Center Hospital East, Chiba, Japan; Department of Urology, Institute of Science Tokyo, Tokyo, Japan., Department of Urology, Institute of Science Tokyo, Tokyo, Japan. Electronic address: ., Department of Urology, Institute of Science Tokyo, Tokyo, Japan., Department of Clinical Biostatistics, Graduate School of Medical and Dental Sciences, Institute of Science Tokyo, Tokyo, Japan., Department of Urology, National Cancer Center Hospital East, Chiba, Japan.