Neoadjuvant nivolumab with or without relatlimab in resectable non-small-cell lung cancer: a randomized phase 2 trial

Lung cancer is the leading cause of cancer mortality worldwide, and many patients present with advanced disease. Checkpoint inhibitors targeting PD-1/PD-L1 and CTLA-4 have improved outcomes in metastatic non-small-cell lung cancer (NSCLC) and are increasingly being moved into perioperative settings. Delivering immunotherapy before surgery may offer advantages: it avoids postoperative delays, allows direct assessment of tumor response, and may be more effective while tumor-infiltrating lymphocytes remain in situ. Dual checkpoint blockade can theoretically overcome resistance seen with single-agent PD-1/PD-L1 therapy. Relatlimab targets LAG-3, which is co-expressed with PD-1 on exhausted T cells; preclinical and clinical melanoma data suggest synergistic activity with nivolumab. The NEOpredict-Lung trial tested whether short-course neoadjuvant nivolumab with or without relatlimab is feasible and safe without delaying curative surgery in resectable NSCLC, and explored signals of clinical activity and biological correlates.

Prior neoadjuvant immunotherapy studies in resectable NSCLC have shown feasibility and promising pathological responses using PD-1/PD-L1 antibodies alone or combined with chemotherapy (for example, NEOSTAR, CheckMate 816, and other trials). While chemoimmunotherapy increases pathologic response rates and event-free survival, it can obscure the contribution of the ICI component and add toxicity, with up to ~20% not reaching surgery in larger studies. Dual ICI regimens in metastatic NSCLC (e.g., nivolumab plus ipilimumab) have shown mixed benefit compared to PD-1/PD-L1-based regimens with chemotherapy. LAG-3 is a rational partner for PD-1 blockade due to complementary mechanisms and evidence from a phase 3 melanoma trial (relatlimab plus nivolumab) demonstrating improved responses and PFS over nivolumab alone, with manageable safety. These data underpin evaluation of PD-1/LAG-3 dual blockade in earlier-stage, resectable NSCLC.

Design: NEOpredict-Lung (NCT04205552) is an open-label, randomized, phase 2 trial conducted at three centers (Essen, Germany; Hasselt, Belgium; Amsterdam, The Netherlands). Sixty adult, treatment-naive patients with histologically or cytologically confirmed, resectable NSCLC (UICC 8th ed. stages IB, II, or selected IIIA) and ECOG PS 0–1 were enrolled. Key inclusion criteria included adequate organ function and curative resectability per multidisciplinary tumor board. Exclusion criteria included extensive mediastinal lymph node metastases (multilevel N2 or any N3), distant metastases, active autoimmune disease, significant cardiovascular disease, active infections (HBV, HCV, HIV), or prior ICI exposure. Randomization and treatment: Patients were randomized 1:1 (no stratification, no blinding) to receive two neoadjuvant doses every 14 days of nivolumab 240 mg IV (arm A) or nivolumab 240 mg plus relatlimab 80 mg IV (arm B). Surgery was planned within ≤43 days from treatment start, followed by standard-of-care adjuvant therapy as indicated; postoperative treatments were not part of the protocol. Endpoints: The primary endpoint was feasibility, defined as the number of patients proceeding to curatively intended surgery within 43 days of therapy initiation. Secondary endpoints included objective radiographic response (RECIST 1.1) pre-surgery, pathological responses (complete pathological response [pCR]: 0% viable tumor; major pathological response [MPR]: ≤10% viable tumor), R0 resection rate, disease-free survival (DFS) and overall survival (OS) at 12 months, safety/tolerability, and 90-day postoperative morbidity/mortality. Exploratory endpoints included FDG-PET/CT metabolic response (PERCIST) in a subset, peripheral blood and tumor immune phenotyping by flow cytometry, NanoString-based gene expression profiling (PanCancer Immune Profiling and PanCancer Pathway panels), and whole-exome sequencing to assess mutational spectra and subclonal dynamics. Assessments: All patients had baseline FDG-PET/CT; brain imaging by contrast-enhanced MRI or CT was performed in nearly all. Radiographic response was assessed immediately pre-surgery. Pathologic response was assessed using standardized criteria with site pathologists. Safety: Adverse events (AEs), serious AEs, and treatment-emergent AEs (TEAEs) were collected and graded; immune-related AEs were characterized. Statistics: The study was designed descriptively to confirm feasibility; up to 30 evaluable patients per arm were planned, with continuous monitoring for feasibility stopping rules. Secondary endpoints were analyzed exploratorily/descriptively; no formal between-arm comparative testing was planned.

Enrollment and treatment: Of 64 screened, 60 patients were randomized (30 per arm); 58 (97%) received both planned doses (one in each arm received only one dose due to immune-related AEs that resolved). All 60 patients (100%) proceeded to surgery within the protocol-defined timeline, meeting the primary endpoint. Surgical outcomes: Curative resection was achieved in 57/60 (95%) overall. Arm-level details: arm A (nivolumab) R0 resection in 30/30; arm B (nivolumab+relatlimab) R0 in 27/30, R1 in 1/30, pleural carcinosis detected intraoperatively in 2/30. One patient had an oligometastatic single bone lesion identified perioperatively and was managed with curative intent. Radiographic and metabolic response: No complete radiographic responses. Partial response by RECIST: 10% (arm A) vs 27% (arm B). Stable disease and progression proportions are depicted in Fig. 1d. In the Essen subset (n=31), partial metabolic response by PERCIST occurred in 38% in both arms; all MPRs had partial metabolic responses, and 67% of PMR patients achieved MPR. Pathological response: pCR occurred in 13% (4/30) with nivolumab and 17% (5/29 evaluable) with nivolumab+relatlimab. MPR rates were 27% (arm A) and 30% (arm B). Pathologic response (≤50% viable tumor) was observed in 60% (arm A) and 72% (arm B). Deeper responses clustered in PD-L1–positive tumors. Survival: With median follow-up of 12 months, 12-month DFS was 89% (arm A) and 93% (arm B); 12-month OS was 93% (arm A) and 100% (arm B). No patient with MPR relapsed; one MPR patient died of noncancer cause (pulmonary embolism). Safety: Any AE occurred in 92% overall during neoadjuvant therapy. Serious AEs: 30% (arm A) vs 33% (arm B). TEAEs: 53% (arm A) vs 63% (arm B). Grade ≥3 TEAEs: 10% (arm A) vs 13% (arm B). Common immune-related AEs included hyper- and hypothyroidism; grade 3 hyperthyroidism occurred in one arm A patient. Pneumonitis occurred in two patients per arm (mostly grade 1–2). No deaths occurred during neoadjuvant therapy; postoperative 90-day mortality was 3% (two deaths, both in arm A). Exploratory correlates: Responders showed increased peripheral CD8+ and CD8+GrzB+ T cells after 4 weeks. Tumors with MPR had lower intratumoral CD16+ neutrophils, CD14+ monocytes, and CD4+CD25+ regulatory T cells. Gene expression analyses showed induction of inflammatory and interferon-related genes (e.g., CXCL2, CXCR4) with treatment; LAG3 expression was induced by nivolumab monotherapy but not the combination. In arm B, MPR associated with suppression of myeloid-related programs (e.g., CD24, CXCL1, CXCL14, IL8, MIF, ISG15) and upregulation of NLRP3, CD27, IRF4, IL16. Genomic analyses revealed subclonal remodeling in tumors with partial responses, with evidence of both enrichment and depletion of subclones; in select cases, putative resistance alterations emerged in residual tumor (e.g., MYC copy gain, KRAS gain, IDH1 and STK11 variants).

Short-course neoadjuvant nivolumab with or without relatlimab met the feasibility endpoint, with all patients proceeding to surgery on schedule and a high R0 resection rate (95%). Safety was acceptable, with low rates of grade ≥3 TEAEs and no neoadjuvant treatment-related deaths. Secondary activity signals, including MPR and pCR rates, were comparable between arms within this small, noncomparative study and similar to prior neoadjuvant ICI experiences without chemotherapy. Early survival outcomes were favorable, and achieving MPR correlated with an absence of relapse at early follow-up. Translational analyses suggest that effective neoadjuvant PD-1±LAG-3 blockade augments cytotoxic CD8+ T cell responses and that reduced intratumoral myeloid and regulatory T-cell populations accompany MPR, particularly with the combination. Gene expression changes indicate consistent induction of inflammatory programs with therapy and a potentially more directed immune activation pattern with PD-1/LAG-3 dual blockade. Genomic analyses show rapid immunotherapy-driven clonal selection in partially responding tumors, highlighting dynamic resistance mechanisms. Collectively, these findings support the safety and operational feasibility of adding LAG-3 blockade to PD-1 inhibition in the perioperative NSCLC setting and provide mechanistic leads for future rational combinations.

NEOpredict-Lung demonstrates that two neoadjuvant doses of nivolumab with or without relatlimab are feasible and safe in resectable stage IB–IIIA NSCLC, with all patients proceeding to surgery and high R0 resection rates. Activity signals include MPR and pCR rates consistent with prior neoadjuvant ICI studies and favorable early DFS/OS. Exploratory analyses reveal immune activation patterns and reductions in myeloid and regulatory populations associated with response, and suggest immunotherapy-driven clonal remodeling in residual disease. Further evaluation of dual PD-1/LAG-3 blockade in larger, comparative perioperative trials is warranted, including biomarker-stratified designs, longer follow-up to confirm survival benefit, and exploration of rational triplet combinations to extend benefit to broader patient populations.

The study was open-label, non-comparative, and not powered for formal efficacy comparisons between arms. The sample size was moderate (n=60). Randomization was not stratified by PD-L1 status, leading to imbalances that could influence response rates. Patients with extensive mediastinal nodal disease were excluded, potentially contributing to high operation and R0 rates and favorable early survival. Median follow-up was relatively short (12 months) for survival endpoints. Findings from exploratory biomarker analyses require validation.