BCG-unresponsive CIS: published complete response rates are not comparable across agents

Design

Reconstruction and simulation analysis of six single-arm registrational trials

Population

Patients with BCG-unresponsive carcinoma in situ in six trials

Primary outcome

12-month complete response after harmonising trial design

Effect

NAI+BCG 44% to 33%; pembrolizumab adjusted to 17%; TAR-200 45%

This BJU International analysis reconstructed data from six registrational single-arm trials in BCG-unresponsive carcinoma in situ and simulated how design choices change reported efficacy, using the nadofaragene firadenovec trial as reference because it alone reported false-negative rates for cystoscopy and cytology against biopsy.

Counting patients who needed re-induction as non-responders lowered 12-month complete response for nogapendekin alfa inbakicept plus BCG from 44% to 33% and for cretostimogene from 45% to 38%. Adjusting for an 11% false-negative rate where biopsy was not mandatory lowered pembrolizumab's 12-month rate to 17%. TAR-200, tested with mandatory biopsies, retained a 45% 12-month rate and outperformed the reference.

These are the numbers used in tumour boards and patient conversations to choose bladder-sparing options before cystectomy. The analysis shows that differences in biopsy policy and retreatment rules can account for much of the apparent difference between agents. It is a modelling exercise on reconstructed data, not a head-to-head comparison.

• Do not compare headline complete response rates across BCG-unresponsive CIS trials without checking biopsy and re-induction rules.
• Ask whether a trial required mandatory biopsy when judging its complete response rate.
• Present bladder-sparing options to patients with realistic, design-adjusted expectations.
• Keep radical cystectomy in the discussion as the standard with the longest evidence.

Why it matters

Choices between costly bladder-sparing agents are being made on figures that are not like-for-like.

Don't overread it

Simulation on reconstructed data; it does not establish which agent is best.

The statistics, in plain English

The exact binomial test compares each adjusted rate with the reference trial's rate. P values of .07 and .4 mean those agents could not be distinguished from the reference after adjustment; TAR-200's P < .01 means it performed better than the reference.

Read the rest in the app

You have read your two free briefings this month. The app carries all 27 specialties, every morning, free — and this finding is waiting in it.

Get Daily Doctor — free

Scan to keep reading on your phone. No account needed to start.