Bayesian Network Meta-Analysis · Cisplatin-eligible MIBC

Perioperative EV-Pembrolizumab, GC-Durvalumab and dose-dense MVAC in muscle-invasive bladder cancer

Indirect comparison of three perioperative regimens against a common gemcitabine-cisplatin (GC) comparator across three phase III trials (EV-304/KEYNOTE-B15, NIAGARA, VESPER). Efficacy shown as hazard ratios (EFS, OS; lower favours the regimen) and odds ratios (pCR; higher favours the regimen). EV-304 efficacy and safety updated to the NEJM 2026 publication.

0.53EV+P EFS HR vs GC
(0.41–0.70)

0.65EV+P OS HR vs GC
(0.48–0.89)

2.6EV+P pCR OR vs GC
(2.0–3.5)

96%EV+P SUCRA
(EFS, ranked 1st)

2,364Participants
across 3 trials

3Phase III RCTs
star network

Indirect treatment comparisons

Direct effects versus GC and head-to-head indirect comparisons between experimental regimens. Bayesian NMA (fixed-effect), concordant with Bucher ITC.

EV+Pdd-MVACGC+DurvaGC (control)

0.350.50.711.52

Versus GC (direct)

EV+P

0.53 (0.41–0.70)

GC+Durva

0.68 (0.56–0.82)

dd-MVAC

0.74 (0.55–0.99)

Head-to-head (indirect)

EV+P vs GC+Durva

0.78 (0.56–1.10)

EV+P vs dd-MVAC

0.72 (0.48–1.10)

GC+Durva vs dd-MVAC

0.92 (0.65–1.30)

Hazard ratio (log scale) — HR < 1 favours the first regimen

EFS. All three regimens significantly improved EFS over GC. No head-to-head comparison between experimental regimens reached significance (all p ≥ 0.1).

Treatment rankings

SUCRA (surface under the cumulative ranking curve). Higher = more likely to be the best regimen for the selected endpoint.

EV+P

96%

GC+Durva

58%

dd-MVAC

45%

EV+P ranked first for all three endpoints. GC ranked last throughout.

Safety overview

Grade ≥3 adverse events, treatment discontinuation, and treatment-related deaths, by regimen.

Grade ≥3 AEs (%)

EV+P

76%

dd-MVAC

64%

GC+Durva

41%

Discontinuation (%)

EV+P

14.6%

dd-MVAC

41%

GC+Durva

15%

Treatment-related deaths (%)

EV+P

0.5%

GC+Durva

0.6%

dd-MVAC

1.2%

GC+Durva added no excess Grade ≥3 toxicity over GC; dd-MVAC least favourable (NNH = 5 for Grade ≥3). EV+P highest Grade ≥3 but lowest treatment-related death rate.

EV-304 adverse events by category NEW · NEJM 2026

Per-category adverse events (as-treated population), EV+P (N=403) vs GC (N=396) — the detailed safety breakdown unavailable in the ASCO GU abstract.

EventEV+P (%)GC (%)

Anemia

31.3

56.3

Neutropenia

14.6

55.1

Nausea

28.3

46.7

Pruritus

46.2

4.5

Constipation

26.6

37.1

Diarrhea

34

16.4

Alopecia

31.5

11.1

Decreased appetite

28.5

18.2

Thrombocytopenia

2.2

28.5

Fatigue

27.5

28

Rash

25.1

2.8

Urinary tract infection

25.1

21.2

Dysgeusia

24.8

9.3

AST increased

24.3

5.1

ALT increased

24.1

9.9

Asthenia

22.8

23.7

Weight decreased

21.6

6.3

Rash maculopapular

20.6

3.5

Leukopenia

3.2

18.9

Blood creatinine incr.

12.4

18.4

Peripheral neuropathy

17.1

3

EV+P skewed to cutaneous, gastrointestinal and neuropathic events (pruritus, rash, peripheral neuropathy); GC skewed to haematologic events (anemia, neutropenia, thrombocytopenia). AEs of special interest: EV-related skin reactions 63.5% and peripheral neuropathy 36.0%; pembrolizumab-related severe skin reactions 17.1% and hypothyroidism 12.2%.

Included trials

Effect estimates entered into the network (experimental arm vs intra-trial GC control).

Trial	Regimen	N (exp / GC)	pCR, exp vs GC	EFS HR	OS HR
EV-304 / KEYNOTE-B15	EV+P	405 / 403	55.8% vs 32.5%	0.53	0.65
NIAGARA	GC+Durva	533 / 530	37.3% vs 27.5%	0.68	0.75
VESPER (neoadjuvant)	dd-MVAC	218 / 219	39% vs 32%	0.74	0.71

VESPER: neoadjuvant subgroup (n=437), pCR on ITT denominator. EV-304 cystectomy rate 86.7% vs 89.6% (NEJM 2026). All EFS/OS hazard ratios versus intra-trial GC.