COMMIT (NRG-GI004/SWOG-S1610)

Background

Three-arm, prospective, open-label phase 3 (COMMIT; NRG-GI004/SWOG-S1610). First-line mismatch repair-deficient / microsatellite instability-high (dMMR/MSI-H) metastatic colorectal cancer, randomised 1:1:1. N=102 enrolled Nov 2017 – Mar 2025 (target 211 after redesign). Median age 63.3 y; 47.6% female. Rationale: roughly half of patients treated with single-agent PD-1 blockade progress within 12 months, and preclinical data suggest VEGF inhibition and cytotoxic chemotherapy synergise with PD-L1 blockade.

Results

Interventions and follow up: Arm A: mFOLFOX6 (oxaliplatin 85 mg/m², leucovorin 400 mg/m², 5-FU bolus 400 mg/m², 46-hour infusional 5-FU 2,400 mg/m²) + bevacizumab + atezolizumab (FFX/bev/atezo), n=41
Arm B: Atezolizumab 840 mg IV every 2 weeks, monotherapy, n=41
Arm C: mFOLFOX6 + bevacizumab (FFX/bev), n=20 — closed early after the KEYNOTE-177 readout
Primary endpoint: Progression-free survival (PFS) in the intention-to-treat population
mFollow up: 46 months (two continuing arms).
Results: PFS (primary): FFX/bev/atezo superior to atezolizumab alone — HR 0.439, 95% CI 0.23–0.84, P=.0103, below the prespecified critical value of 0.0152.
Median PFS: NR in the primary report.
ORR: 86.1% vs 46%.
12-month disease control rate: 64.7% vs 32.4%.
OS: NR.

Adverse events

Grade ≥3 (any attribution): 52 patients overall — 18 with atezolizumab monotherapy vs 34 with FFX/bev/atezo.
Haematologic: NR in the primary report.
Immune-mediated: NR in the primary report.
Chemotherapy/anti-VEGF-specific (neuropathy, hypertension, perforation): NR.
Discontinuation: NR.

Conclusions

In first-line dMMR/MSI-H metastatic colorectal cancer, adding mFOLFOX6 and bevacizumab to atezolizumab more than halved the risk of progression versus atezolizumab alone (HR 0.44) and nearly doubled objective response (86.1% vs 46%). COMMIT is the first randomised phase 3 signal that chemo-immunotherapy intensification improves PFS over checkpoint monotherapy in this biomarker-defined population, at the cost of roughly double the grade ≥3 event count and without demonstrated OS benefit.

Key Limitations

Severely underpowered relative to the original design: the FFX/bev control arm was closed after only 20 patients when KEYNOTE-177 read out, and total accrual reached 102 of a planned 211, leaving roughly 41 patients per analysed arm and a correspondingly wide hazard ratio confidence interval (0.23–0.84). Accrual spanned eight years (2017–2025), during which the frontline dMMR standard shifted first to pembrolizumab and then toward dual checkpoint blockade. Open-label. The comparator is atezolizumab, a PD-L1 agent that is not the established frontline standard — PD-1-directed therapy is — so the control arm may understate what current immunotherapy achieves, exaggerating the apparent benefit of intensification. PFS primary with immature OS and no survival benefit shown. Toxicity reported only as aggregate grade ≥3 counts, with no category-level breakdown.

Clinical Context

First-line dMMR/MSI-H metastatic colorectal cancer is treated with immunotherapy rather than chemotherapy, established by KEYNOTE-177 (pembrolizumab versus chemotherapy) and extended by CheckMate-8HW, in which nivolumab plus ipilimumab improved PFS over nivolumab alone and over chemotherapy; NCCN, ASCO and ESMO recommend checkpoint blockade as preferred first-line therapy in this population. COMMIT targets the residual problem of primary immunotherapy refractoriness, but does so by adding cytotoxic chemotherapy and anti-VEGF to a PD-L1 backbone rather than by intensifying immunotherapy itself. Whether chemotherapy intensification is preferable to dual checkpoint blockade is untested head-to-head, and the small sample argues against changing practice on COMMIT alone.

References