ESMO Congress 2026

For patients with muscle-invasive bladder cancer (MIBC) who are ineligible for or decline cisplatin, the phase 3 evidence for perioperative enfortumab vedotin (EV) plus pembrolizumab is already established: KEYNOTE-905/EV-303 was designed from the start as a three-arm trial comparing pembrolizumab monotherapy, pembrolizumab plus EV, and radical cystectomy with pelvic lymph-node dissection alone ^[2]; the phase 3 results published in NEJM in 2026 randomized 344 patients (170 vs 174), with event-free survival (EFS) as the primary endpoint and median follow-up of 25.6 months ^[1]. Reviews have grouped EV-303/KEYNOTE-905 with NIAGARA and EV-304/KEYNOTE-B15 as perioperative studies delivering substantial improvements in EFS, overall survival and pathological response ^[3], and the Australian guide lists it among the landmark trials ushering in a new era of perioperative immunotherapy and ADCs ^[6].

What was presented at ESMO26 (Abstract 4733O, by Anders Ullén) was not a parent-trial update but an ad hoc exploratory three-arm analysis: 255 concurrently randomized, cisplatin-ineligible MIBC patients, with 85 on EV + pembrolizumab, 83 on pembrolizumab alone, and 87 on surgery alone, aimed at unpacking each drug's contribution. The efficacy showed a clear gradient: 2-year EFS of 75.8%, 55.8% and 41.0%, with EFS HRs versus surgery of 0.40 and 0.71; 3-year EFS of 69.5%, 49.6% and 37.4%; pCR of 58.8%, 25.3% and 10.3%, and pathological downstaging of 65.9%, 33.7% and 14.9%; OS pointed the same way, with HRs of 0.48 (95% CI 0.29–0.80) and 0.76 (0.48–1.20), and median OS of NR, 51.0 and 41.7 months. On safety, grade ≥3 adverse events were 44.0% vs 20.7% in the neoadjuvant phase and 60.0% vs 44.9% in the adjuvant phase. The meeting also featured ctDNA analyses of KEYNOTE-905 and KEYNOTE-B15 as an invited discussion, a separate parallel story.

This readout moves the question from whether perioperative EV + pembrolizumab works to why it works and which patients truly need both agents: pembrolizumab monotherapy was numerically better than surgery alone, but one live commentary explicitly noted that its confidence interval versus surgery crossed 1 and did not formally separate; adding EV lifted pCR by 48.6 percentage points over surgery, read as support for the biological rationale of combining an ADC with immune checkpoint inhibition, and it raised the question of which patients could safely de-escalate. A pathology-angle commentary cautioned that 41% of patients on EV+P still had residual disease, that membranous Nectin-4 combined with histological subtype may be the most informative combination currently available, and that with pCR carrying this much weight, the ypT0N0 diagnosis requires standardized sampling and central review.

The limitations must be stated clearly: this was an ad hoc, descriptive analysis without formal statistical testing and cannot serve as confirmatory evidence of each drug's independent contribution; one post tallied grade ≥3 TEAEs across the three arms at 42%, 70% and 69%, and similar rates between the two drug arms do not mean equivalent toxicity profiles, since neuropathy, skin and metabolic toxicity in a curative-intent setting are not interchangeable with immune-related events. Whether adjuvant EV is needed awaits clues from VOLGA, as noted on site; another commentary mentioned that the approval was based on KEYNOTE-905 (cisplatin-ineligible) and KEYNOTE-B15 (cisplatin-eligible), covering both populations.