Why This Study Matters
DNA mismatch repair deficiency (dMMR) or microsatellite instability-high (MSI-H) status is reported in about 4–7% of metastatic colorectal cancer (mCRC). Single-agent PD-1 blockade is superior to chemotherapy in the first-line setting, yet the authors note that in KEYNOTE-177, 29.4% of patients had progressive disease as best response, and the 12-month PFS rate with pembrolizumab was 55.3%.
Preclinical models suggest that oxaliplatin can drive immunogenic cell death and that VEGF inhibition and chemotherapy can synergize with PD-L1 blockade. COMMIT was designed to test whether adding mFOLFOX6 plus bevacizumab to the anti-PD-L1 antibody atezolizumab improves outcomes in previously untreated dMMR/MSI-H mCRC.
Study Design
COMMIT was initially designed as a three-arm, open-label phase III trial that randomized eligible patients 1:1:1 to mFOLFOX6/bevacizumab (FFX/bev), atezolizumab monotherapy (840 mg IV every 2 weeks), or the combination FFX/bev/atezo. Randomization was stratified by BRAF V600E status, site of metastases (liver only versus extrahepatic), and prior adjuvant chemotherapy. After KEYNOTE-177 results, the FFX/bev arm was closed in June 2020 after 20 patients, leaving 1:1 randomization between the two remaining arms. Crossover was not allowed.
The 20 patients randomized to FFX/bev before that arm closed were not part of the formal comparison; with a median follow-up of 56.6 months they had a median PFS of 6.8 months, a median OS of 18.7 months, and an ORR of 50% (among 16 response-evaluable patients), which the authors describe as limited by small sample size. The primary endpoint was PFS in the intent-to-treat population. Prespecified secondary endpoints were ORR, OS, toxicity, 12-month PFS, disease control rate (DCR), duration of stable disease, and duration of response. Because of slow accrual, the sample size for the two remaining arms was reduced from 211 to 100 with a target HR change from 0.60 to 0.50, and accrual was suspended on March 31, 2025 after 82 patients were randomized between the two remaining arms.
102 patients enrolled (Nov 2017–Mar 2025); 82 randomized to the two compared arms (41 + 41); initial 1:1:1 randomization, with the FFX/bev arm closed after 20 patients; median follow-up 46 months (95% CI, 35–52).
Number of study sites: Not specified in source (the Appendix lists participating institutions). The trial was limited to North America.
Patient Population
Eligible adults (≥18 years) had previously untreated mCRC with locally assessed dMMR (by immunohistochemistry) or MSI-H status (by PCR or next-generation sequencing), measurable disease per RECIST v1.1, and ECOG performance status 0–2. Central pathology review was not performed. In the 82 patients in the two compared arms, median age was 63.3 years (range, 19.8–87.4), 47.6% were female, and BRAF V600E mutations were present in 23.2%. More than 80% had metastatic disease beyond the liver, and 13.4% had prior adjuvant CRC therapy. ECOG performance status 2 was recorded in 7.3% (5 of 82 patients) per Table 1; the Discussion cites 6%.
Primary Endpoint: Progression-Free Survival
A pre-planned interim analysis (data through March 31, 2025; 45 events; 65% information) reviewed by the data monitoring committee in July 2025 showed that FFX/bev/atezo improved PFS versus atezo alone (HR=0.44; 95% CI, 0.23–0.84; P=0.0103), crossing the prespecified O’Brien–Fleming efficacy boundary (one-sided P=0.01523). The abstract reports the HR to three decimals as 0.439. Updated results at the November 6, 2025 cutoff (median follow-up 46 months) gave HR=0.42 (95% CI, 0.22–0.80; P=0.0068).
| Endpoint | FFX/bev/atezo (n=41) | Atezo (n=41) |
|---|---|---|
| Median PFS | 24.5 months (95% CI, 10.1–NE) | 5.3 months (95% CI, 2.2–18.2) |
| 12-month PFS rate | 66.7% | 35.1% |
| 24-month PFS rate | 53.7% | 31.6% |
| ORR (response-evaluable: n=36 vs n=37) | 86.1% (n=31) | 46.0% (n=17) |
| Complete response | 36.1% (n=13) | 18.9% (n=7) |
| Partial response | 50.0% (n=18) | 27.0% (n=10) |
| Stable disease | 11.1% (n=4) | 21.6% (n=8) |
| Progressive disease as best response | 2.8% (n=1) | 32.4% (n=12) |
| 12-month DCR | 64.7% | 32.4% (P=0.009) |
| 24-month OS rate | 67% | 67% |
An inverse-probability-of-treatment-weighted sensitivity analysis was reported as consistent with these results. Duration of response (HR=0.86; 95% CI, 0.28–2.64; P=0.796) and duration of stable disease (HR=0.54; 95% CI, 0.22–1.31; P=0.167) did not differ significantly between arms. There was no difference in OS (HR=1.04; 95% CI, 0.47–2.28; P=0.90); the authors note the trial was not powered for survival and that more patients in the atezo arm received later chemotherapy (35% versus 8%).
Subgroup Analyses
The authors describe these as exploratory analyses. No subgroup demonstrated statistical significance after Bonferroni adjustment for multiplicity, and no interaction was significant.
| Subgroup | HR (95% CI) | Interaction P |
|---|---|---|
| Higher baseline tumor volume | 0.37 (0.16–0.86) | 0.45 |
| Elevated baseline CEA | 0.43 (0.20–0.90) | 0.47 |
| Age ≥60 | 0.40 (0.18–0.88) | 0.49 |
| Prior adjuvant chemotherapy | 0.28 (0.06–1.33) | 0.38 |
The Discussion also mentions numerical trends in right-sided primary tumors and BRAF V600E-mutant disease; subgroup statistics for those were Not specified in source text.
Safety Profile
Adverse events were more frequent and severe with the combination. Grade 3–5 events regardless of attribution occurred in 82.9% of patients receiving the combination versus 43.9% with atezo. The median duration of treatment was 11 months (range, 0.1–28) in the combination arm and 5 months (range, 0–26) in the atezo arm. There were five Grade 5 AEs: one in the atezo arm due to disease progression and four in the combination arm (one due to disease progression, two sudden deaths after cycles 16 and 18, and one hepatic hemorrhage after an extended right hepatectomy). Immune-related toxicities remained uncommon in both arms.
| Selected adverse event (any attribution) | Atezo, grade ≥3 | FFX/bev/atezo, grade ≥3 |
|---|---|---|
| Overall highest grade, all AEs | 18 (43.9%) | 34 (82.9%) |
| Neutrophil count decreased | 0 | 11 (26.8%) |
| Hypertension | 1 (2.4%) | 8 (19.5%) |
| Diarrhea | 0 | 5 (12.2%) |
| Sepsis | 0 | 4 (9.7%) |
| Oral mucositis | 1 (2.4%) | 4 (9.8%) |
| Thromboembolic event | 0 | 3 (7.3%) |
| Fatigue | 1 (2.4%) | 2 (4.9%) |
The Results text additionally reports one Grade 4 colonic perforation and one Grade 4 esophagitis in the combination arm.
In conclusion, combination therapy with FFX/bev/atezo demonstrates a statistically significant improvement in PFS compared to atezo, with no new safety signals.
— Rocha Lima et al., J Clin Oncol 2026 (Discussion). FFX/bev/atezo = mFOLFOX6 + bevacizumab + atezolizumab; PFS = progression-free survival.
Interpretation and Broader Context
The authors conclude that COMMIT, described as the first randomized trial evaluating chemotherapy plus anti-VEGF and anti-PD-L1-based therapy in treatment-naive dMMR mCRC, supports FFX/bev/atezo as a potential treatment option.
The authors caution that cross-trial comparisons should be interpreted cautiously. They note that the atezo-only arm’s 12- and 24-month PFS (35.1% and 31.6%) was lower than that reported for first-line pembrolizumab in KEYNOTE-177 (~55% at one year) and nivolumab in CHECKMATE 8HW, and they discuss possible contributors including PD-L1 versus PD-1 biology and local rather than central biomarker testing. CHECKMATE 8HW, the only randomized phase III trial comparing combination versus single-agent anti-PD-1 therapy, showed improved PFS with nivolumab plus ipilimumab (HR=0.62). These are cross-study benchmarks, not results from this trial. Future correlative analyses of tumor tissue and ctDNA may clarify which subsets require treatment intensification.
Limitations
Per the authors: enrollment was slow (102 patients) and accrual was suspended early, with the sample size for the two compared arms cut from 211 to 100 and 82 patients randomized; the trial was open-label without central radiology review; dMMR/MSI-H status was assessed locally rather than centrally; and the treatment landscape shifted during accrual with KEYNOTE-177 and CHECKMATE 8HW, which led to closure of the FFX/bev arm after 20 patients. The primary result rests on a small number of events, and the trial was not powered for overall survival, where no difference was seen. The subgroup analyses were exploratory, with no subgroup significant after Bonferroni adjustment. The atezolizumab-alone arm had lower PFS than reported for PD-1 monotherapy in other trials, so cross-trial comparisons should be interpreted cautiously. The trial was limited to North America.