Background
Single-arm, open-label phase 2 trial. 124 patients with previously treated MSI-H/dMMR metastatic colorectal cancer (mCRC), divided into Cohort A (≥2 prior lines, n=61) and Cohort B (≥3 prior lines, n=63). Evaluated pembrolizumab monotherapy in MSI-H/dMMR mCRC, a population historically resistant to chemotherapy but characterized by high neoantigen burden.
Interventions and follow up
Arm A: Pembrolizumab 200mg IV q3wk × up to 35 cycles
Arm B: Not applicable — single-arm study
Primary endpoint: Objective response rate (ORR)
mFollow up: 31.3 mo (Cohort A)
Arm B: Not applicable — single-arm study
Primary endpoint: Objective response rate (ORR)
mFollow up: 31.3 mo (Cohort A)
Results
ORR (Cohort A, ≥2L): 32.8% (95% CI 21.3–46.0%)
ORR (Cohort B, ≥3L): 34.9% (95% CI 23.3–47.7%)
mDOR (Cohort A): Not reached (NR)
mPFS (Cohort A): 2.3 mo
mOS (Cohort A): 31.4 mo
ORR (Cohort B, ≥3L): 34.9% (95% CI 23.3–47.7%)
mDOR (Cohort A): Not reached (NR)
mPFS (Cohort A): 2.3 mo
mOS (Cohort A): 31.4 mo
Adverse events
Immune-mediated: Immune-mediated AEs 21.8% — hypothyroidism (any grade) 10.5%, colitis grade ≥3 3.2%, pneumonitis grade ≥3 1.6%, adrenal insufficiency 1.6%
Overall: Grade ≥3 AEs 16.9% (Cohort A); treatment discontinuation due to AEs 8.1%
Overall: Grade ≥3 AEs 16.9% (Cohort A); treatment discontinuation due to AEs 8.1%
Conclusions
Pembrolizumab produced clinically meaningful and durable responses in patients with previously treated MSI-H/dMMR mCRC, with ORR ~33% across both cohorts and median OS exceeding 31 months — remarkable in a heavily pretreated population historically achieving median OS <6 months with third-line chemotherapy. The durability of responses (median DOR not reached) is the defining clinical feature.
Key Limitations
Key Limitations: Non-randomized single-arm design — ORR is the primary endpoint without a contemporaneous control arm. The mPFS of 2.3 months seems paradoxically short despite high ORR and long OS, reflecting the delayed kinetics of immune responses (pseudoprogression) and the distribution of responders vs non-responders. The 33% ORR means ~67% of patients do not benefit, with no validated predictive biomarker beyond MSI-H/dMMR to further refine selection. MSI-H comprises only ~4–5% of mCRC patients, limiting absolute impact on the overall mCRC population.
Clinical Context
KEYNOTE-164 supported FDA accelerated approval of pembrolizumab for MSI-H/dMMR mCRC across all lines in 2017, converted to regular approval in 2020. It has been superseded in the first-line setting by KEYNOTE-177 (pembrolizumab 1L vs FOLFOX/FOLFIRI, n=355 in DB) and CheckMate 8HW (nivolumab + ipilimumab 1L, n=646 in DB). For MSI-H patients who progress after first-line IO, KEYNOTE-164 demonstrates continued pembrolizumab activity in later lines.
References
References: Le DT et al, JCO 2020