Observation versus maintenance immunotherapy after PD-1-induced clinical complete response in dMMR/MSI-H colorectal cancer managed non-operatively: a multicenter cohort study.

BACKGROUND

Whether maintenance immunotherapy is necessary after clinical complete response (cCR) to programmed cell death protein 1 (PD-1) inhibition in mismatch repair-deficient/microsatellite instability-high colorectal cancer remains unclear. We evaluated its associations with survival and treatment-related toxicity.

METHODS

This multicenter cohort study included patients treated at five hospitals in China between January 2018 and March 2025 who achieved cCR after PD-1 inhibitor therapy and entered non-operative management. All patients underwent endoscopy and site-appropriate cross-sectional imaging to confirm cCR; 62.0% also underwent biopsy. Patients were categorized into observation and maintenance groups according to whether they received maintenance immunotherapy after cCR, defined as at least two additional treatment cycles. The primary endpoint was disease-free survival (DFS) from cCR; secondary endpoints were overall survival (OS), locoregional regrowth, distant metastasis, and treatment-related adverse events (TRAEs). Overlap weighting addressed measured confounding, with 6-week landmark and strict per-protocol analyses assessing robustness.

RESULTS

Among 318 patients treated with PD-1 inhibitors, 173 (54.4%) achieved cCR; 129 (74.6%) entered non-operative management (observation, n=66; maintenance, n=63), while 44 underwent surgery. The analytic cohort included 58 patients with a single rectal primary, 61 with a single colon primary, and 10 with synchronous colorectal primaries; 115 (89.1%) had stage I–III disease. Median follow-up from cCR was 32.2 months (IQR 20.4–44.4). Both locoregional regrowths occurred in the observation group within 2 years after cCR; both patients underwent salvage treatment and remained alive. No distant metastases occurred, and the only death was due to COVID-19 without recurrence. 3-year DFS was 96.2% with observation and 98.3% with maintenance (log-rank p=0.508); corresponding OS was 100.0% and 98.3% (p=0.321). Overlap weighting did not materially alter the DFS comparison (p=0.192), and sensitivity analyses yielded similar estimates. Across the entire treatment and follow-up, any TRAE was numerically more frequent in the maintenance group than in the observation group (65.1% vs 59.1%); 15.9% of patients in the maintenance group experienced grade ≥2 TRAEs during maintenance therapy.

CONCLUSIONS

Findings suggest that discontinuation after cCR may be reasonable within a watch-and-wait strategy for rectal cancer and provide preliminary support for non-operative management of colon cancer; prospective validation is warranted.