Neoadjuvant nivolumab and relatlimab in locally advanced MMR-deficient colon cancer: a phase 2 trial

Immune checkpoint blockade (ICB) is standard for several metastatic cancers, including dMMR colorectal cancer, and has moved into earlier disease stages. dMMR occurs in about 15% of non-metastatic colorectal cancers and leads to hypermutation and abundant neoantigens, rendering tumors highly immunogenic. Prior neoadjuvant ICB studies in dMMR colon cancer, notably NICHE-2 with nivolumab/ipilimumab, achieved exceptionally high pathologic response rates, while neoadjuvant chemotherapy (FOXTROT) yielded poor responses in dMMR tumors. LAG-3 is an inhibitory receptor associated with T-cell exhaustion; dual PD-1 plus LAG-3 blockade (nivolumab/relatlimab) has shown efficacy and a favorable toxicity profile in melanoma. Given high LAG-3 expression in dMMR colon tumor-infiltrating lymphocytes, the authors hypothesized that neoadjuvant nivolumab plus relatlimab would produce high pathologic response rates with acceptable safety in locally advanced, resectable dMMR colon cancer. The NICHE-3 phase 2 cohort tested two cycles of nivolumab 480 mg plus relatlimab 480 mg prior to surgery, with primary endpoint of pathologic response (≤50% residual viable tumor).

Background evidence includes: neoadjuvant dual ICB (nivolumab/ipilimumab) in dMMR colon cancer with ~98% pathologic responses and 68% pCR (NICHE-2), multiple smaller studies of neoadjuvant anti–PD-1 monotherapy in dMMR colorectal cancer with pCR rates ~53–88% but a higher fraction of limited/no response versus dual ICB, and FOXTROT demonstrating poor pathologic responses (~7%) for dMMR tumors with neoadjuvant chemotherapy. In melanoma, nivolumab/relatlimab showed efficacy in advanced and early-stage disease with relatively favorable toxicity versus other combinations, supporting exploration of PD-1+LAG-3 in dMMR colon cancer. dMMR tumors exhibit frequent LAG-3 expression on tumor-infiltrating lymphocytes, providing mechanistic rationale.

Design: Investigator-initiated, multicenter, non-randomized, open-label phase 2 cohort (NICHE-3; ClinicalTrials.gov NCT03026140) within the NICHE platform. A Simon’s two-stage design tested a null pathologic response rate of 70% versus one-sided alternative of 85%; stage 1 accrued 19 patients (proceed if >14 responses), then 40 additional patients for total n=59; reject null if ≥47 responses. Alpha 0.05, power 80%. Ethics and oversight: Approved by NedMec (joint IRB of the Netherlands Cancer Institute (NKI), UMC Utrecht, Princess Máxima Center) and conducted per ICH-GCP and the Declaration of Helsinki; written informed consent obtained. Population: Adults ≥18 years with resectable, locally advanced (≥cT3 and/or cN+) dMMR colon adenocarcinoma without distant metastases; WHO performance status 0–1; adequate organ function. dMMR by IHC loss of MLH1, PMS2, MSH2, and/or MSH6; sporadic dMMR defined by MLH1 promoter hypermethylation or somatic MMR mutations. Key exclusions: obstruction/perforation, prior chemotherapy/ICB, recent other malignancies unless low risk, immunodeficiency/autoimmune disease, or chronic steroids >10 mg prednisone equivalent daily. Intervention: Nivolumab 480 mg plus relatlimab 480 mg on day 1 and day 29 (±3 days), followed by surgical resection within 6–8 weeks after enrollment. Imaging with baseline CT of chest/abdomen/pelvis; centrally reviewed. Blood draws at baseline, on-treatment, and follow-up for safety and translational studies. Tissue obtained via baseline endoscopic biopsies; post-treatment resection specimens collected. Endpoints: Primary—pathologic response (≤50% residual viable tumor, RVT). Secondary—safety (immune-related adverse events, irAEs), major pathologic response (MPR; ≤10% RVT), pathologic complete response (pCR; 0% RVT), disease-free survival (DFS), and overall survival (OS). Treatment-related AEs leading to >2-week surgical delay deemed unacceptable. Postoperative AEs graded by Clavien-Dindo. Pathology: Central histopathologic assessment of entire resected primary tumor and all lymph nodes; response quantified by percent RVT. pCR required 0% RVT in primary tumor bed and tumor-draining lymph nodes; cases with pCR in primary but nodal metastases were classified as MPR. Staging per AJCC 8th edition. Statistics: Safety population included all treated patients; efficacy population those without major protocol deviations and evaluable for pathologic response. Binary endpoints reported with Clopper-Pearson 95% CIs; categorical as frequencies; continuous as medians with ranges. DFS analyzed by Kaplan-Meier; median follow-up by reverse Kaplan-Meier. Analyses performed in R 4.3.0 and SAS 9.4.

- Accrual and baseline: 67 screened; 59 enrolled and treated between 15 Dec 2022 and 4 Apr 2024. Median age 65 years (21–85), 54% female, 19% Lynch syndrome. Majority had cT4 (68%) and cN+ (63%) disease; 63% stage III. All 59 underwent surgery; R0 achieved in 100%; median lymph node yield 33 (9–104). - Primary endpoint met: Pathologic response ≤50% RVT in 57/59 patients (97%; 95% CI 88–100%). - Depth of response: MPR (≤10% RVT) in 54/59 (92%; 95% CI 81–97%); pCR (0% RVT) in 40/59 (68%; 95% CI 54–79%). Three patients receiving only one cycle all achieved MPR (one pCR). Partial responses (11–50% RVT) in 3 patients (12–30% RVT). Two non-responders (RVT 80% and 90%). Pathologic response observed across clinical stages with similar pCR rates in cT2/3 (74%) and cT4a/b (65%), and in cN0 (73%) and cN+ (65%); pCR lower with elevated baseline CEA ≥5 µg/L (50% vs 74%). Tumor-positive lymph nodes in 2 patients (one non-responder; one with 6% RVT); both declined adjuvant chemotherapy. - Outcomes: Median time from first dose to surgery 7.6 weeks (6.4–34.3). At median follow-up of 8 months (2–19), all patients alive; 58/59 (98%) disease-free; one recurrence (irresectable dMMR liver metastases) occurred in a patient with delayed surgery due to grade 4 hepatitis and nephritis requiring prolonged immunosuppression. - Safety: Any-grade irAEs in 47/59 (80%; 95% CI 67–89%); grade 3–4 irAEs in 6/59 (10%; 95% CI 4–21%). Common irAEs: infusion-related reactions 29%, thyroid dysfunction 22% (including 1 grade 3 hyperthyroidism), fatigue 20%, rash/dermatitis 17%. Grade 3 events: hepatitis 3%, colitis 3%, hyperthyroidism 2%; one grade 4 hepatitis. Endocrinopathies in 29% (mostly grade 1–2); 25% required chronic hormone replacement. Three patients had treatment-related surgical delays (4–26 weeks). Postoperative grade 3–4 complications in 7% (ileus 3%, surgical site infection 2%, intra-abdominal abscess 2%, anastomotic leak 2%, hemorrhage 2%).

A short neoadjuvant regimen of nivolumab plus relatlimab produced a 97% pathologic response rate and 68% pCR in locally advanced dMMR colon cancer, supporting robust antitumor activity of PD-1+LAG-3 blockade in this setting. These rates align with prior dual ICB (PD-1+CTLA-4) results in dMMR colon cancer and exceed pathologic responses historically seen with neoadjuvant chemotherapy in dMMR tumors. Responses were consistent across tumor stages and nodal status, though numerically lower pCR rates were observed with cT4 disease and higher baseline CEA, suggesting tumor burden may influence depth of response; the high MPR rate implies responses may deepen over time and could be affected by the interval to surgery. Compared with nivolumab/ipilimumab in NICHE-2, nivolumab/relatlimab showed similar efficacy but appeared to have higher rates of certain irAEs (thyroid dysfunction, adrenal insufficiency, colitis), potentially related to the higher relatlimab dose (480 mg) and two-dose schedule. Future NICHE cohorts will explore alternative dosing/schedules to maintain efficacy while reducing toxicity, and results will inform a larger international study. The high MPR/pCR rates reignite interest in organ preservation strategies for dMMR colon cancer, though accurate response assessment remains challenging; integration of ctDNA, PET-CT, and endoscopic evaluation may improve selection. The cohort targeted patients at radiographically high risk (≥cT3 and/or cN+), a group with substantial recurrence risk under standard chemotherapy, countering the notion that most are cured by surgery alone. Overall, the data strengthen the rationale for neoadjuvant immunotherapy as a preferred approach in resectable dMMR colon cancer and underscore the need for survival data and toxicity optimization.

Neoadjuvant nivolumab plus relatlimab achieved exceptionally high pathologic response (97%), MPR (92%), and pCR (68%) rates with acceptable safety in locally advanced, resectable dMMR colon cancer. The findings, consistent with and complementary to prior dual ICB studies, support neoadjuvant immunotherapy as a promising standard in this population. Future research should include larger, multicenter studies to confirm long-term DFS/OS benefits, refine dosing and scheduling to reduce irAEs while preserving efficacy, validate pathologic response as a surrogate endpoint for DFS, and develop reliable multimodal response assessment to enable organ preservation strategies in selected patients.

Key limitations include: single-arm, non-randomized design with cross-study comparisons limiting definitive efficacy and safety comparisons; reliance on radiographic staging, which is inaccurate for nodal status and can lead to overstaging and potential overtreatment, although cT4 assessment is more reliable; relatively short median follow-up (8 months) without mature survival outcomes; lack of an established surrogate endpoint in colon cancer (pathologic response is not yet validated for DFS in this disease); and potential contribution of treatment interval and dosing to toxicity profile. Radiographic response assessment after neoadjuvant immunotherapy is also unreliable, complicating organ-preservation decision-making.