Between 9 August 2019 and 5 August 2022, among the 54 patients evaluated for eligibility, 52 were enrolled in this study. Twelve (23.1%), 16 (30.8%), and 24 (46.2%) of the 52 patients were confirmed with < 1%, between 1% and 49%, and ≥ 50% PD-L1 expression, respectively. Twelve patients with ≥ 50% PD-L1 expression received neoadjuvant nivolumab-exclusive (N) and 40 patients received nivolumab–chemotherapy (N/C) combinations. All the patients underwent at least one-cycle neoadjuvant treatment. Among the prespecified arms, the disease characteristics and baseline demographics were generally well balanced (Table 1).

When the database was locked on August 5, 2022, 48 (92.3%) patients had fully completed the three-cycle neoadjuvant treatment, whereas 46 (88.5%) patients had undergone surgery and at least one-cycle adjuvant nivolumab treatment (Supplementary Fig. S1). Eighteen (39.1%) patients had completed the adjuvant treatment as per the protocol, 24 (52.2%) were still undergoing trial treatment, and four (8.7%) had discontinued the adjuvant nivolumab treatment. The median time length of the adjuvant nivolumab treatment was 50.6 weeks (range 2.9–52.1).

Forty-six (88.5%) of the 52 patients visited the operating room with surgery intentions. Among the six patients without surgery, one received nivolumab treatment (8.3%, 1/12) and five were treated in the nivolumab plus chemotherapy arm (12.5%; 5/40). The most common reasons for not undergoing surgery were toxicity (n = 3, 1.9%), disease progression (n = 2, 3.8%), and declined (n = 1, 1.9%). The percentages of patients with disease progression or adverse events were undifferentiated among the treatment groups (Table 2; Supplementary Fig. S1).

Of the 46 (100.0%) patients who were initiated using a minimally invasive (video-assisted thoracoscopic surgery) approach, only two (4.3%) were converted to thoracotomy. Forty-one (89.1%) patients underwent lobectomy, three (6.5%) underwent pneumonectomy, and one (2.2%) underwent sleeve lobectomy and wedge resection respectively. Forty-five (97.8%) patients underwent R0 surgical resection and 26 (56.5%) achieved pathological lymph node downstaging.

After the final dose of neoadjuvant treatment, the median interval to surgical resection was 32 days (range 26–56). Five surgeries (5/46, 10.9%) were delayed (Table 2). No surgical morbidity or mortality due to the neoadjuvant treatment was observed.

For the intention-to-treat population, based on blinded independent pathological review (BIPR), the primary endpoint of MPR was 50.0% (26/52, 95% CI 35.8–64.2%), whereas the pCR rate was 25.0% (13/52, 95% CI 14.0–38.9%) (Fig. 1a). Among the patients with PD-L1 expression ≥ 50%, N/C induced a significantly higher MPR rate of 66.7% (A2, 8/12, 95% CI = 34.8–90.1%) than N (16.7%; A1, 2/12, 95% CI = 2.1–48.4%) (odds ratio = 14.85, 95% CI = 1.50–64.91, p = 0.009). Patients also more commonly obtained pCR after N/C (41.7%; 5/12, 95% CI 15.2–72.3) than after N (16.7%; 2/12, 95% CI 2.1–48.4%) (Fig. 1b). For PD-L1 expressions of 1–49% (B1) and < 1% (B2), the MPR rate after neoadjuvant N/C was 56.3% (B1, 9/16, 95% CI = 29.9–80.2%) and 58.3% (B2, 7/12, 95% CI = 27.7–84.8%), and the pCR rate was 18.8% (B1, 3/16, 95% CI = 4.0–45.6%) and 25.0% (B2, 3/12, 95% CI = 5.5–57.2%), respectively (Fig. 1c).

Pathological evaluation of responses to neoadjuvant nivolumab with or without chemotherapy on the basis of PD-L1 expression. a Pathological responses in resected primary lung tumors following neoadjuvant nivolumab treatment and nivolumab plus chemotherapy, based on the remaining viable tumor cells percentage, for each patient (n = 46) who received surgical resection. The black horizontal line represents the MPR threshold (90% regression). Clinical characteristics, pathological and preoperative radiologic responses are annotated for each patient. b Proportion of MPR/non-MPR and pCR/non-pCR in enrolled patients after neoadjuvant nivolumab and nivolumab plus chemotherapy (A1 & A2, PD-L1 ≥ 50%). c Proportion of MPR/non-MPR and pCR/non-pCR in enrolled patients after neoadjuvant nivolumab plus chemotherapy (B1, 1% ≤ PD-L1 ≤ 49%; B2, PD-L1 < 1%). d Proportion of MPR/non-MPR and pCR/non-pCR in resected patients following nivolumab plus chemotherapy and neoadjuvant nivolumab. e MPR/non-MPR in resected patients following nivolumab plus chemotherapy and neoadjuvant nivolumab based on PD-L1 expression. p values were calculated through Fisher’s exact tests. ASC adenosquamous carcinoma, CR complete response, LUAD adenocarcinoma, LUSC squamous-cell carcinoma, MPR major pathological response, N nivolumab monotherapy, N/C nivolumab plus chemotherapy, pCR pathologic complete response, PD progression disease, PD-L1 programmed death ligand-1, PR partial response, RECIST Response Evaluation Criteria in Solid Tumors Version 1.1, SD stable disease. * The preoperative PET-CT displayed an additional FDG uptake in the right supraclavicular lymph node then the patient underwent a wedge resection and the right supraclavicular lymph node sampling. No variable tumor in resected lung tissue but metastasis in the supraclavicular lymph node. † The patient received systematic mediastinal lymph node dissection and radical lobectomy. No variable tumor in primary tumor while metastasis in the hilar lymph node

As resected specimens were not available for six patients who received treatment without undergoing surgery, a sensitivity analysis was carried out to examine the MPR and pCR rates in those who underwent surgical resection. Among the 46 patients resected in the trial, a significantly higher incidence of MPR in the N/C group was observed than that in the N group among the patients with ≥ 50% PD-L1 expression (80.0% vs.18.2%, odds ratio = 14.85, 95% CI 1.50–64.91, p = 0.009). pCR was also more common in the N/C group (50.0% vs.18.2%, odds ratio = 4.17, 95% CI 0.46–59.60, p = 0.183) (Fig. 1d). In the post-hoc analysis, no significant correlations between the pathological responses and the baseline PD-L1 expression were identified in the patients who received neoadjuvant immunochemotherapy (MPR, A2 vs. B1 vs. B2, 80.0% vs. 64.3% vs. 63.6%, p = 0.428) (Fig. 1e).

In all treated populations, the overall response rate (ORR) was 55.8% (29/52, 95% CI 41.3–69.5%), whereas in the N group and N/C group, the ORRs were 41.7% (5/12, 95%CI 15.2–72.3%) and 60% (24/40, 43.3–75.1%), respectively (Supplementary Table S1 and Table S2). Additionally, two patients that achieved stable disease radiographically were assessed as pCR after surgery.

With a median follow-up of 25.1 months for the entire enrolled population (95% CI 22.0–27.7 months, range 0.07–32.7 months), neither the median EFS nor the median OS was reached. Nine (17.3%) of the 52 patients experienced disease progression or died, two of whom (3.8%) did not undergo surgery (Fig. 2). Thirty-six (78.6%) of the 46 patients with tumor resection were recurrence-free and alive, with an 18-month EFS rate of 64.8% (95% CI 51.9–81.0%). As the median 12.4-month follow-up in A2 was relatively shorter than the other three arms which was 27.6, 27.6, and 25.1 months in A1, B1, and B2, respectively, survival outcomes might need to be interpreted cautiously.

Swimmer plot depicting events in all enrolled population (n = 52). The left column displays pathological responses and clinical characteristics of the patients. Each bar represents one patient. Seven patients (13.5%) who received surgery had experienced disease progression, and two (3.8%) of them had died. ASC adenosquamous carcinoma, LUAD adenocarcinoma, LUSC squamous carcinoma, N nivolumab monotherapy, N/C nivolumab plus chemotherapy, RECIST Response Evaluation Criteria in Solid Tumors Version 1.1

Overall, adverse events were manageable, without new concerns of safety in comparison to the known safety profiles for neoadjuvant nivolumab monotherapy and nivolumab plus chemotherapy. Of the 52 patients in the safety-evaluable population, 36 (69.2%) experienced at least one immune-related adverse event during neoadjuvant treatment, and 13 (25.0%) experienced grade ≥ 3 adverse events. Exhaustion (40.3%), appetite loss (32.7%), and anemia (28.8%) were among the most common grade 1/grade 2 immune-related adverse events. White blood cell count reduction (16.7%, n = 2) and neutrophil count reduction (10.0%, n = 4) were the most frequent grade ≥ 3 immune-related adverse events among the 12 patients receiving N and the 40 patients receiving N/C, respectively. Only one patient death among the three (5.8%) who died in the neoadjuvant phase was related with immune treatment (pneumonitis) (Supplementary Table S3).

The collected plasma samples at the pre-treatment period were analyzed in 38 patients. ctDNA was detected in 89.5% (34/38) of pre-treatment samples overall, including 21.1% (8/38) with stage II disease and 68.4% (26/38) with stage III disease (Fig. 3a, Supplementary Table S4).

Associations between circulating tumor DNA (ctDNA) and responses to neoadjuvant treatment. a Patients and tumor characteristics based on the baseline ctDNA status (T0). b ctDNA dynamics during neoadjuvant nivolumab monotherapy and nivolumab plus chemotherapy. Proportion of patients based on ctDNA positivity according to available numbers of samples per time point. c Sankey plot displaying ctDNA dynamics (clearance/non-clearance) during neoadjuvant treatment versus response (pCR/non-pCR). Analyses were carried out for the patients with positive ctDNA at T0 (baseline) and corresponding ctDNA test results at T2 (following completing neoadjuvant therapy). N nivolumab monotherapy, N/C nivolumab plus chemotherapy, pCR pathologic complete response

The ctDNA detection rate decreased during neoadjuvant treatment, from 89.5% before treatment (T0) to 34.2% after neoadjuvant treatment (T2), then continued to decrease to 27.6% after surgery (T3). Specifically, neoadjuvant N and N/C reduced ctDNA positivity by 42.9% (T3) and 18.2% (T3), respectively (Fig. 3b and Supplementary Table S5).

We next evaluated whether ctDNA clearance was associated with pCR rates. Among the patients with detectable ctDNA at T0 and underwent the corresponding T2 plasma measurements, 23 were ctDNA-negative at T2, with a ctDNA clearance rate of 67.6% (23/34) in the whole cohort. The percentage of patients with ctDNA clearance who received neoadjuvant N/C was significantly higher than those who received N (75.8% vs. 20%; odds ratio = 11.49, 95%CI 0.94–64.36, p = 0.029). As much as 39.1% of patients with undetectable ctDNA at T2 were without residual disease at surgery (pCR 9/23), in comparison to 6.7% of patients with detectable ctDNA at T2 (pCR 1/15) (Fig. 3c), showing a substantial association (odds ratio = 6.14, 95%CI = 0.84-Inf, p = 0.077). With a median 22.7-month follow-up (95%CI = 22.0–27.7 months, range 8.9–30.8 months) for all 38 patients, four experienced local recurrence and two experienced distant metastases, one of whom died (Fig. 4a). Notably, detectable ctDNA was observed at a minimum of one point in all patients with local and distant recurrences. Interestingly, the 18-month EFS rate for patients with ctDNA/MRD– (both T2 and T3) vs. those with ctDNA/MRD+ (either T2 or T3) was 93.8% vs. 47.3% (HR, 0.15; 95% CI: 0.04, 0.94; p = 0.005) (Fig. 4b).

Circulating tumor DNA (ctDNA) and survival outcomes. a Swimmer plot depicted events in patients with baseline ctDNA testing (n = 38). The left column displays pathologcial responses and clinical characteristics. Each bar represents one patient. b Patient survival according to ctDNA/MRD status at T2 (after the completion of neoadjuvant treatment) and T3 (1 month after surgery, prior to adjuvant treatment). Negative, ctDNA/MRD, assessed as negativity at both T2 and T3; Positive, ctDNA/MRD+, assessed as positivity at either T2 or T3. ASC adenosquamous carcinoma, LUAD lung adenocarcinoma, LUSC lung squamous carcinoma, MPR major pathological response, N nivolumab monotherapy, N/C nivolumab plus chemotherapy, pCR pathologic complete response