Neoadjuvant and perioperative immunotherapy in resectable non–small cell lung cancer: Evidence to date and remaining questions

Case of NSCLC with bulky N2 disease before successful perioperative chemo-immunotherapy.

Central Message.

Neoadjuvant and perioperative chemo-immunotherapy consistently improves survival and pathological response in locally advanced non–small cell lung cancer.

The neoadjuvant and perioperative use of chemo-immunotherapy in resectable non–small cell lung cancer (NSCLC) has reshaped the treatment approach for locally advanced disease and has considerably improved outcomes. The addition of immune checkpoint inhibitors in the neoadjuvant (before surgery alone) and perioperative (before and after surgery) settings aims to target micrometastatic disease before resection while tumor antigens are abundant.¹ To date, several landmark Phase 3 trials have demonstrated the improved outcomes, including event-free survival (EFS), overall survival (OS), pathological complete response (pCR), and major pathological response (Table 1).¹ However, these trials include a broad spectrum of patients, including those with 8th edition of the TNM Classification of Malignant Tumours stage IIA to IIIB disease, varying levels of programmed death-ligand 1 (PD-L1) expression, different oncogenic mutations, and both nodal-negative and multilevel N2 disease. This heterogeneity makes it difficult to determine which subgroups derive the most benefit from neoadjuvant or perioperative therapy.² Additionally, these regimens carry the risk that some patients may not proceed to surgery due to treatment-related toxicity, with Phase 3 trials reporting attrition rates of 15% to 22%, raising concerns about compromised local therapy.3, 4, 5, 6, 7 In the light of these ongoing challenges in real-world clinical practice, the objective of this article is to discuss the current evidence and updates on immunotherapy in resectable disease and to offer an outlook on future directions.

Table 1.

Landmark phase 3 trials have demonstrated the improved outcomes, including event-free survival (EFS), overall survival (OS), pathological complete response (pCR), and major pathological response (MPR)

Trial name	N, randomization	End points	Stages∗	Systemic plan	No. of neoadjuvant cycles	Treatment arm	Surgery (%)	R0 rate (%)	EFS at 2 y (%)	Median EFS (months)	EFS (95% CI)	OS at 2 y (%)	Median OS (m)	OS hazard ratio (95% CI)	pCR (%)	MPR (%)
CheckMate 816 (CTx-nivolumab)	358, 1:1	PCR, EFS	II-IIIB	Neoadjuvant	3	ICI	83.2	83.2	63.8	31.6 (30.2-NR)	0.63 (97.38% CI, 0.43-0.91); P = .005	82.7	NR	0.72 (95% CI, 0.532-998); P = .048	24.0	36.95
						Placebo	75.4	77.8	45.3	20.8 (14.0-26.7)		70.6	NR		2.2	8.8
CheckMate 77T (CTx-nivolumab)	461, 1:1	EFS	II-IIIB	Perioperative	4	ICI	78	89	N/A	N/A	0.58 (97.36% CI, 0.42-0.81); P = .00025	N/A	N/A	Immature	25.3	35.4
						Placebo	77	90	N/A	N/A		N/A	N/A		4.7	12.1
KEYNOTE-024 (CTx-pembrolizumab)	791, 1:1	EFS, OS	II-IIIB	Perioperative	4	ICI	82.1	92.0	62.4	NR (34.1-NR)	0.58 (95% CI, 0.46-0.72); P < .001	80.9	NR	0.72 (95% CI, 0.56-0.93); 1-sided P = .0052	18.1	32.2
						Placebo	79.4	84.2	40.6	17.0 (14.3-22.0)		77.6	45.5 (42.0-NR)		4.0	11:0
AEGEAN (Ctx-duravlumab)	740, 1:1	PCR, EFS	II-IIIB	Perioperative	4	ICI	80.6	94.7	63.3	NR (31.9-NR)	0.68 (95% CI, 0.53-0.88); P = .004	N/A	N/A	Immature	17.2	33.3
						Placebo	80.7	91.3	52.4	25.9 (18.9-NR)		N/A	N/A		4.3	12.3
Neotarch (CTx-toriplimab)	404, 1:1	MPR, EFS	III	Perioperative	3	ICI	82.2	95.8	64.7	NR (24.4-NR)	0.40 (95% CI, 0.28-0.57); P < .001	81.2	NR	Immature	24.8	33.3
						Placebo	73.3	92.6	38.7	15.1 (10.6-21.9)		74.3	30.4 (29.2-NR)		1.0	12.3
RATIONALE-315 (Ctx-tislelizumab)	453, 1:1	EFS, MPR	II-IIIA	Perioperative	3-4	ICI	84.1	95.3	68.3	NR (NR-NR)	0.56 (95% CI, 0.40-0.79); P = .0003	88.6	NR	Immature	40.7	56.2
						Placebo	76.2	93.1	51.8	NR (16.6-NR)		79.4	NR (35.0-NR)		5.7	15.0

CTx, Chemotherapy; ICI, immune checkpoint inhibitor; NR, not reached; N/A, not available.

^∗TNM Classification of Malignant Tumours, 8th Edition.

Quality of Trials and Evidence for Neoadjuvant or Perioperative Regimens

Multiple recent systematic reviews and meta-analyses have evaluated the overall quality and the risk of bias in Phase 2 and 3 randomized controlled trials (RCTs) of immunotherapy for resectable early-stage NSCLC, using the Cochrane Risk Bias Tool (version 2.0).^1,8,9 With respect to randomization, missing outcome data, and selection of reported results, all Phase 3 RCTs (CheckMate 816, CheckMate 77T, Keynote-671, AEGEAN, RATIONALE 315, and Neotorch) were judged to have low or limited risk of bias.^1,8,9 One systematic review identified a high risk of bias due to deviations from intended interventions and a high risk of bias in outcome measurement in the CheckMate 816, the NADIM-II, and the TD-FOREKNOW trials.⁹ The variability in outcome measurement particularly referred to the inaccurate consideration of tumor progression under neoadjuvant treatment when assessing EFS.⁹ Although the definition of pCR was consistent across the trials, the blinding of pathological assessment varied.⁹ The authors of the 2 other systematic reviews concluded that the overall trial quality was high with only limited risk of bias.^1,8 For randomization, missing outcome data, outcome measurement, and selection of reported results, all RCTs were rated as having a low risk of bias.⁸ Regarding publication bias, no significant bias across different PD-L1 subgroups was detected using the Egger test.⁸

Despite the consistent EFS benefits observed across 1 neoadjuvant and 5 perioperative Phase 3 trials with similar hazard ratios, it remains unclear how much of this benefit can be attributed specifically to the adjuvant component. To address this gap in evidence, a propensity score-weighted comparison using individual patient-level data from the CheckMate 816 and CheckMate 77T trials was performed.¹⁰ The study suggested that the perioperative use of nivolumab may further improve EFS (unweighted hazard ratio, 0.59; 95% CI, 0.38-0.92), in patients with both <1% and ≥1% PD-L1 expression, across all baseline disease stages, and regardless of pCR status. However, even with propensity score matching, the results should be interpreted with caution because key methodological differences between the 2 trials cannot be fully accounted for. These limitations include variations in baseline characteristics and the number of induction cycles (3 cycles in CheckMate 816 vs 4 cycles in CheckMate 77T). In addition, the analysis excluded patients in CheckMate 77T who did not receive adjuvant treatment due to postoperative complications, whereas no similar exclusion was possible in the CheckMate 816 cohort. This leads to a further selection bias toward patients with an uneventful recovery in the perioperative group. Additional evidence on this topic comes from 2 recent meta-analyses: an individual patient data meta-analysis of 1064 patients from 8 prospective trials found no statistically significant difference in EFS between patients treated with perioperative and neoadjuvant regimens.¹¹ Similarly, a network meta-analysis, including 3113 patients from 8 RCTs, found no clear differences between the 2 approaches in terms of pCR, EFS, or OS.¹² Subgroup analyses within that meta-analysis offer a more nuanced view and, indicating superior outcomes with neoadjuvant chemo-immunotherapy in patients with high PD-L1 expression (≥50%), nonsquamous histology, and in nonsmokers. Conversely, the perioperative approach may provide a greater benefit in patients with PD-L1 <1%, squamous cell carcinoma, or in those who did not achieve pCR.¹²

Circulating Tumor DNA as a Minimal Residual Disease Biomarker in Resectable NSCLC

The personalization and tapering of perioperative chemo-immunotherapy requires biomarkers that can reliably detect minimal residual disease (MRD) after induction treatment. Liquid biopsy-based analysis of circulating tumor DNA (ctDNA) has emerged as a promising, minimally invasive tool for risk stratification. Translational data from several clinical trials support the role of ctDNA as a prognostic marker. The post hoc analyses from the Phase 2 single-arm NADIM trial showed that baseline ctDNA mutant allelic fractions were predictive of OS and EFS and ctDNA clearance after neoadjuvant treatment was associated with significantly improved OS.¹³ Additional evidence comes from the CheckMate 816 trial, which monitored ctDNA dynamics during neoadjuvant treatment.³ Patients who cleared ctDNA after 2 cycles had higher pCR rates and greater reductions in tumor burden. Long-term follow-up confirmed a survival benefit for those who achieved ctDNA clearance before surgery, with markedly higher 5-year OS than patients with persistent ctDNA.³ In the AEGEAN trial, similar patterns emerged: all patients with initially detectable ctDNA who ultimately achieved pCR also cleared ctDNA by the end of therapy, whereas persistent ctDNA reliably predicted failure to achieve pCR, with more than 84% accuracy.¹⁴ In the TRAcking Cancer Evolution through therapy (Rx) study, a tumor-informed assay using anchored-multiplex pCR identified preoperative ctDNA as a predictor of OS in TNM eighth edition stage I to III lung adenocarcinoma, although this association was not observed in nonadenocarcinoma histologies.¹⁵

Taken together, these findings indicate that the presence of ctDNA after neoadjuvant treatment signals an elevated risk of relapse and may serve as a surrogate marker for poorer prognosis. However, ctDNA clearance is not yet suitable as a standalone guide for treatment de-escalation largely because of the limited sensitivity of current assays.¹⁶ Several limitations must be addressed before ctDNA can be used as an MRD assay for decision making after induction therapy or resection. Improving assay sensitivity, such as by phased variant sequencing, is needed to increase accuracy after induction or surgery.¹⁶ In the light of intratumoral heterogeneity, it also remains uncertain whether pretreatment biopsies alone are sufficient to generate robust tumor-informed signatures for disease monitoring.¹⁶ Well-designed prospective trials are necessary to validate the prognostic performance of ultrasensitive ctDNA assays and to test whether a ctDNA-guided strategy can identify patients who should continue or discontinue adjuvant therapy.¹⁷ In addition, non-ctDNA liquid biomarkers such as circulating tumor cells, micro RNAs, extracellular vesicles, or blood-based protein signatures have shown promise in treatment monitoring and may ultimately support a multilayered integration of biomarkers into a single assay.¹⁸

Complex Situations in Locally Advanced NSCLC

Complex clinical scenarios such as Pancoast tumors and multistation or bulky mediastinal nodal involvement remain central challenges in the management of locally advanced NSCLC. A key study addressing superior sulcus tumors (N0-1 M0) was the Southwest Oncology Group-Intergroup Trial S0220, which evaluated a trimodality approach using neoadjuvant chemoradiotherapy (cisplatin-etoposide and 45 Gy thoracic radiotherapy), followed by surgery and adjuvant docetaxel in 46 patients.¹⁹ The trial reported a very promising pCR or pathologic nearly complete response rate of 72%. However, despite excellent locoregional control, the 3-year EFS was 56%, and distant relapses, particularly in the brain, remained a major concern.¹⁹ More recently, the single-arm Phase 2 INCREASE trial evaluated a multimodal strategy in 29 patients with resectable or borderline resectable T3 or 4 N0 to 2 NSCLC, including 7 Pancoast tumor cases. This approach combined neoadjuvant dual immune checkpoint inhibition (ipilimumab and nivolumab) with concurrent chemoradiotherapy (50-60 Gy), followed by resection.²⁰ Similar to the Southwest Oncology Group-Intergroup Trial S0220, the trial achieved excellent local control, with a pCR rate of 58% and 100% R0 resections. Nonetheless, grade 3 or 4 treatment-related adverse events occurred in 70% of patients, indicating that this multimodality regimen should be reserved for those with good performance status.²¹ In the field of superior sulcus tumors or Pancoast tumors, further research is thus needed to clarify the role of induction chemo-immunotherapy.

In cases with multistation or bulky N2 disease, treatment decisions are particularly nuanced. Multistation N2 disease, defined as involvement of 2 or more mediastinal lymph node stations and bulky N2 disease, typically referring to nodes larger than 3 cm in short-axis diameter or conglomerated, have traditionally been considered borderline or unresectable.22, 23, 24 Recent data challenge this view. In NADIM-II, approximately two-thirds of patients had pathologically confirmed N2 disease, including multistation involvement. Despite this, high rates of nodal downstaging and favorable survival outcomes were achieved following neoadjuvant chemo-immunotherapy.²⁵ Likewise, the CheckMate 77T trial included 39.7% of patients with multistation N2 disease. Among them, 34% achieved pCR and 57% were downstaged to ypN0, translating into a notable EFS benefit at 12 months, approximately 80% in the experimental arm versus 38% in the control.⁷ To address current knowledge gaps in multistation and bulky N2 disease, real-world cohorts with granular staging and treatment data, as well as long-term follow-up, will be essential.

The outcomes of pneumonectomy for central tumors after neoadjuvant or perioperative chemo-immunotherapy also remain an ongoing point of discussion. In Keynote 671, CheckMate 816, CheckMate 77T, AEGEAN, and RATIONALE 315 trials, pneumonectomy was performed in 11.4%, 16.8%, 9.0%, 7.4%, and 8.4% of all patients in the immunotherapy-arms and in 12.3%, 25.2%, 13.5%, 7.8%, and 12.1% of patients in the placebo arms, respectively.^5,7,E1,E2 In the immunotherapy-treated arms, pneumonectomy-rates were thus reduced in CheckMate 816, CheckMate 77T, and RATIONALE 315, whereas no similar trend was seen in Keynote 671 or AEGEAN. A subgroup analysis from CheckMate 816 presented at the 2025 World Conference of Lung Cancer reported comparable OS in the pneumonectomy cohort (5-year OS was 80% in the nivolumab arm vs 53% in the chemotherapy arm) relative to the lobectomy cohort (5-year OS was 74% vs 67%). Moreover, long-term quality of life was not negatively influenced after pneumonectomy compared with lobectomy.^E1

Despite the proven benefit of neoadjuvant and perioperative chemo-immunotherapy in unselected NSCLC populations, the inclusion of patients with actionable driver mutations remains controversial.^E3 Growing evidence indicates that specific oncogenic alterations may confer varying sensitivity to immune checkpoint inhibition.^E3 The ImmunoTarget group examined immunotherapy responses across different NSCLC driver mutations and found that tumors harboring KRAS, BRAF, or MET exon 14 alterations were more likely to benefit than EGFR-, ALK-, or RET-driven cancers.^E4 A recent study identified KEAP1 and TP53 mutations as independent predictors of poor immunotherapy response in metastatic lung adenocarcinoma.^E5 Translational post hoc analyses from Phase 2 neoadjuvant trials suggest that tumors with STK11 comutations or reduced STK11/KEAP1 copy numbers are less likely to respond.^E3,E6, E7, E8 Among Phase 3 neoadjuvant and perioperative chemo-immunotherapy trials, only Keynote 671 and AEGEAN included subgroups of EGFR- or ALK-mutated NSCLC, although small sample sizes hamper a meaningful interpretation of pathological response and survival.^E9 In CheckMate 816, CheckMate 77T, Neotorch, and RATIONALE 315, patients with EGFR and ALK mutations were excluded.^E9 Given the compelling survival benefit of adjuvant osimertinib for resected, EGFR L858R, or exon 19-deleted NSCLC and with adjuvant alectinib for ALK-rearranged NSCLC, immune checkpoint inhibitors are not considered standard of care for these molecular subsets. Overall, the mutation-related differences in immunotherapy response underscore the need for further studies to refine treatment strategies based on individual tumor biology.^E3

Open Questions and Outlook

Chemo-immunotherapy has reshaped the management of resectable, locally advanced NSCLC. Across multiple Phase 3 trials, the addition of immune checkpoint inhibitors has consistently increased pCR rates and improved survival, establishing a new standard of care that clearly outperforms chemotherapy alone for stage IIA through IIIB disease.

Despite this progress, several practice-defining questions remain. Surgical outcomes in borderline-resectable scenarios, such as in bulky (Figure 1) or multistation N2 disease, still require further investigation. Ongoing studies like the MDT-BRIDGE trial (NCT05925530), which enrolls patients with borderline resectability, will be essential to define multidisciplinary pathways that align induction therapy with either surgery or definitive chemoradiotherapy.^E10 In parallel, high-quality real-world data registries with granular information on clinical staging, potentially even a centralized radiological imaging review, and comprehensive surgical data, could offer valuable subgroup analyses to help to clarify the outcomes of neoadjuvant and perioperative chemo-immunotherapy in borderline-resectable disease.

Figure 1.

A 66-year-old patient was admitted to our department with TNM (9th edition) stage cT4 cN2a (bulky) cM0 thyroid transcription factor-1-positive, TP53-mutated adenocarcinoma of the right upper lobe (A). PD-L1 expression in tumor cells was <1%. Restaging after 4 cycles of neoadjuvant chemo-immunotherapy showed a good radiological response (B). Robotic-assisted thoracic surgery extrapleural right upper lobectomy and systematic mediastinal lymphadenectomy were performed. Although surgical complexity after induction chemo-immunotherapy was increased due to pleural adhesions and hilar fibrosis, a minimally invasive approach remained feasible (C and D). Histology revealed a pathological complete response.

Based on current evidence, the role of an adjuvant continuation of immunotherapy or its de-escalation remains uncertain. ctDNA is a promising tool for MRD assessment and may help to eventually guide treatment de-escalation, but the limited sensitivity and lack of standardization in commercially available assays are still insufficient for that purpose today. The ideal sequence and duration of chemo-immunotherapy—neoadjuvant, perioperative, or adjuvant—also remains a central knowledge gap, with no robust evidence yet to favor 1 approach over another. Several ongoing Phase 3 trials aim to address this. The randomized Phase 3 PROSPECT-Lung trial (NCT06632367) compares perioperative versus adjuvant chemo-immunotherapy in resectable TNM eighth edition stage IIA through IIIB NSCLC. The randomized Phase 3 European Thoracic Oncology Platform Adaptive Oncology Platform Trial (ETOP-ADOPT) trial (NCT06284317) evaluates the benefit of perioperative versus neoadjuvant chemo-immunotherapy in R0-resected patients who do not achieve pCR.^E11 And for patients who do achieve pCR, the randomized Phase 3 INSIGHT (S2414) trial (NCT06498635) investigates perioperative versus neoadjuvant chemo-immunotherapy. Together, these trials are expected to help shed light on the complex landscape of immunotherapy for resectable early-stage NSCLC.

Conflict of Interest Statement

Dr Werner collects speaker fees from AstraZeneca and is on the advisory boards of MSD and BMS. Dr Opitz has received institutional grants from Roche and XVIVO; is on the advisory board and steering committee for AstraZeneca; is on the advisory boards for MSD, BMS, and Regeneron; is on the advisory board and has received an institutional grant from Medtronic; holds a proctorship and receives speaker fees from Intuitive; and has received speaker fees from Sanofi, Siemens, and Astellas. Dr Opitz is international director for the American Association for Thoracic Surgery; a member of the Thoracic Clinical Practice Standards Committee and the Thoracic Education Committee of the American Association for Thoracic Surgery; a board member of the European Society of Thoracic Surgeons and iMig; and an associate editor of the Journal of Thoracic and Cardiovascular Surgery.

The Journal policy requires editors and reviewers to disclose conflicts of interest and to decline handling or reviewing manuscripts for which they may have a conflict of interest. The editors and reviewers of this article have no conflicts of interest.