Tom cites EV303 and EV304 as showing similar pathological complete response rates after three versus four neoadjuvant EV cycles. That undercuts the clinical assumption that more neoadjuvant EV necessarily adds benefit, though the transcript gives no direct detail on those trials' designs or patient populations.
In the ad hoc KEYNOTE-905 analysis, single-agent pembrolizumab produced a 25% pathological complete response, essentially the same as the 27% gemcitabine-cisplatin control rate cited from NIAGARA, despite a high-risk cisplatin-ineligible population. Brian and Tom call this genuine activity, while stressing it is an indirect comparison.
Tom reads the hazard ratios as a near-exact stepwise drop: 0.71 for pembrolizumab versus cystectomy and 0.40 for EV plus pembrolizumab, with each component apparently contributing about 0.3. He cautions the 85-patient arms are small, but says the EFS curves visually support this additive pattern.
Anders notes that in the primary KEYNOTE-905 analysis the control arm had grade 3 or higher adverse events in the 40-50% range, likely from surgery alone. That makes the 44% neoadjuvant rate with EV plus pembrolizumab look less alarming than a simple cross-arm comparison might suggest.
Tom points out that all patients received a year of adjuvant therapy in the experimental arms, but only about 25% of the ITT population completed all nine EV cycles, and little is known about who they were. The three-arm ad hoc analysis therefore cannot isolate the adjuvant phase's benefit or necessity.
Tom argues that if a patient is fit enough for cystectomy, they should be fit enough to start EV plus pembrolizumab, so the single-agent perioperative pembrolizumab scenario is mostly hypothetical in his view. The exception he describes is metastatic lymph-node-only disease with comorbidities, not curative-intent MIBC.
