KEYNOTE-604

Background

Phase III, double-blind, placebo-controlled RCT (KEYNOTE-604). 453 patients with previously untreated extensive-stage SCLC. Stratified by ECOG PS (0–1 vs 2) and platinum agent. Treated/stable CNS metastases eligible.

Interventions and follow up

Arm A: Pembrolizumab 200 mg IV q3wk + etoposide 100 mg/m² d1–3 + cisplatin 75 mg/m² or carboplatin AUC5 d1, q3wk × 4 cycles, then pembrolizumab maintenance up to 35 cycles.
Arm B: Placebo + etoposide-platinum × 4 cycles, then placebo maintenance.
Primary endpoint: PFS and OS (dual primary)
mFollow up: 21.8mo

Results

mPFS: 4.5 vs 4.3mo, HR 0.75, 95% CI 0.61–0.91, P=.0023 (significant)
mOS: 10.8 vs 9.7mo, HR 0.80, 95% CI 0.64–0.98 — prespecified boundary not crossed (P=.0164 vs required P=.0128)
ORR: 70.6% vs 61.8%

Adverse events

Overall: Grade ≥3 events 76.7% vs 74.9%. Discontinuation due to AEs 14.8% vs 6.3%.
Immune-mediated: Any grade 24.7% vs 10.3%; hypothyroidism 17.3%; pneumonitis grade ≥3 2.2%.

Conclusions

Pembrolizumab + chemotherapy significantly improved PFS in ES-SCLC but failed to meet the prespecified OS significance boundary, a statistically negative trial for its dual primary endpoint despite a numerically improved OS trend.

Key Limitations

OS failed to cross prespecified alpha (P=.0164 vs required .0128) — a marginal but decisive miss. PFS benefit (absolute 0.2mo) is clinically negligible. Sequential alpha-spending raised the OS bar. Unlike IMpower133 and CASPIAN, pembrolizumab showed no clear OS benefit.

Clinical Context

Considered a negative trial for pembrolizumab in ES-SCLC, contrasting with IMpower133 (atezolizumab, HR 0.70) and CASPIAN (durvalumab, HR 0.75). Pembrolizumab lacks FDA approval for 1L ES-SCLC; atezolizumab and durvalumab remain the preferred IO agents. A key example of divergent outcomes among similar agents in one disease.

References