Trial design drives outcomes: harmonising efficacy across BCG-unresponsive bladder carcinoma in situ trials

To evaluate how heterogeneity in response assessment protocols, including biopsy timing, complete response (CR) assessment, retreatment policies, and the handling of withdrawals, influences reported efficacy outcomes across single-arm registrational trials of patients with bacillus Calmette Guérin unresponsive (BCG-U) carcinoma in situ (CIS).

We reconstructed survival data from six registrational BCG-U CIS trials and conducted simulation analyses to test how trial design affects efficacy estimates. Specifically, we: (i) reclassified patients who underwent re-induction as non-responders; (ii) adjusted for false-negative rates of cystoscopy/cytology vs biopsy (as uniquely reported in the nadofaragene firadenovec trial, hence chosen as the analytical reference); (iii) harmonised CR assessment timing across trials. The exact binomial test was utilised for comparisons.

Re-induction reclassification reduced the 12-month CR rate for nogapendekin alfa inbakicept (NAI) + BCG from 44% to 33% and for cretostimogene grenadenorepvec from 45% to 38%. After adjusting for an 11% false-negative rate in trials lacking mandatory biopsies, the pembrolizumab 12-month CR rate declined to 17% (P = 0.07), and NAI, accounting for re-induction reclassification, achieved a CR rate of 29% (P = 0.4). TAR-200 outperformed the analytical reference, with a 12-month CR rate of 45% under mandatory biopsy protocols (P < 0.01).

Trial design artefacts significantly impact reported efficacy in BCG-U CIS studies. Published CR rates are not directly comparable across agents. Future trials should standardise biopsy timing and clearly define retreatment policies to ensure meaningful cross-study comparisons.