A systematic review and network meta-analysis of immune checkpoint inhibitors in operable NSCLC: neoadjuvant, adjuvant, or perioperative?

Abstract

Background

The optimal timing for immune checkpoint inhibitor (ICI) therapy in patients with operable non-small cell lung cancer (NSCLC) remains a subject of clinical uncertainty. This Bayesian network meta-analysis (NMA) aimed to compare the efficacy of ICIs administered in neoadjuvant, adjuvant, and perioperative settings.

Methods

A systematic review of randomized controlled trials (RCTs) published up to September 1, 2024, identified eight eligible studies comprising a total of 3,659 patients. Primary outcomes included disease-free survival (DFS) and event-free survival (EFS). A Bayesian framework was used to estimate hazard ratios (HRs) and 95% credible intervals (CrIs), with treatment efficacy ranked using the surface under the cumulative ranking curve (SUCRA). Risk of bias was assessed using the Cochrane Risk of Bias tool, and sensitivity analyses confirmed the robustness of the findings after exclusion of studies at higher risk.

Results

Perioperative ICIs demonstrated superior efficacy compared with neoadjuvant or adjuvant strategies. Specifically, perioperative toripalimab plus chemotherapy ranked first (SUCRA =0.99; rank 1 probability =87%), followed by perioperative nivolumab (SUCRA =0.94). Both regimens significantly reduced the risk of disease recurrence or progression compared to surgery alone or standard chemotherapy. While neoadjuvant and adjuvant ICIs provided moderate benefit, they ranked lower in overall efficacy. Sensitivity analyses excluding higher-risk studies did not substantially alter the results, confirming their robustness. No direct comparisons were available between all strategies, and heterogeneity in control arms, ICI agents, programmed death ligand-1 (PD-L1) inclusion thresholds, and follow-up duration limit cross-trial comparability. Additionally, long-term overall survival data were largely immature across studies.

Conclusions

These findings suggest that perioperative administration of ICIs, particularly toripalimab and nivolumab in combination with chemotherapy, may offer the most effective approach to improving DFS/EFS in resectable NSCLC. However, further head-to-head trials, longer follow-up, and biomarker-stratified analyses are needed to confirm these results and guide personalized treatment decisions.

Introduction

Non-small cell lung cancer (NSCLC) remains one of the leading causes of cancer-related mortality worldwide, with its high incidence and challenging prognosis contributing to a significant global health burden. While early detection and surgical intervention offer the most promising outcomes for patients with operable disease, the reality is that surgery alone is often insufficient to prevent disease recurrence. Recurrence rates remain unacceptably high even after complete resection, particularly in patients with locally advanced disease. This underscores the urgent need for adjunctive therapies that can mitigate the risk of relapse and improve long-term survival.

Historically, adjuvant chemotherapy has been the standard of care for patients undergoing curative-intent surgery for NSCLC. Although this approach has demonstrated modest survival benefits, its efficacy is limited, and many patients experience disease progression despite treatment. More recently, the advent of immune checkpoint inhibitors (ICIs) has revolutionized the treatment landscape for advanced and metastatic NSCLC. ICIs, which target immune regulatory pathways such as the programmed death ligand-1 (PD-1)/programmed death-1 and programmed death ligand-1 (PD-L1) axis, have shown remarkable efficacy in improving survival outcomes by enhancing the immune system’s ability to recognize and eliminate cancer cells. Their success in the metastatic setting has spurred interest in evaluating their role in earlier stages of the disease.

The rationale for incorporating ICIs into the treatment of operable NSCLC lies in their ability to induce durable responses and address micrometastatic disease, which is often undetectable at the time of surgery. Three key settings have emerged as potential opportunities for ICI integration: neoadjuvant, adjuvant, and perioperative. Neoadjuvant therapy involves administering ICIs before surgery, with the goal of reducing tumor burden, improving resectability, and eliciting a systemic immune response that could target micrometastases. Adjuvant therapy, administered after surgery, aims to eliminate residual microscopic disease and reduce the risk of relapse. The perioperative approach combines these strategies, with ICIs given both before and after surgery to maximize therapeutic benefit (1-3).

Despite the theoretical rationale supporting the use of ICIs in all three settings, the optimal timing for their administration remains uncertain due to mixed clinical trial results. For instance, the CheckMate 816 and NADIM II trials (4,5) demonstrated significant improvements in event-free survival (EFS) with neoadjuvant nivolumab plus chemotherapy, whereas the IMpower010 trial reported only modest DFS benefits with adjuvant atezolizumab. In contrast, recent perioperative trials such as NEOTORCH and CheckMate 77T (1,3) suggest that combining perioperative ICI therapy may yield superior outcomes. However, no head-to-head randomized controlled trials (RCTs) directly compare neoadjuvant, adjuvant, and perioperative ICI strategies, and differences in trial design, control arms, and biomarker inclusion criteria further complicate cross-study comparisons. In this context, a network meta-analysis (NMA) offers a methodologically robust approach to integrate both direct and indirect evidence across multiple RCTs, enabling a comprehensive comparison of the relative efficacy of ICI-based strategies. By modeling the treatment network, NMAs can estimate the comparative effectiveness of interventions that have not been directly compared in clinical trials, thereby informing clinical decision-making where conventional pairwise meta-analyses fall short. Given the growing number of RCTs evaluating ICIs in operable NSCLC but the absence of direct comparative trials, an NMA is ideally suited to address this pressing clinical question and to identify the most effective treatment strategy among neoadjuvant, adjuvant, and perioperative options. We present this article in accordance with the PRISMA-NMA reporting checklist (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-185/rc).

Methods

To evaluate the comparative efficacy of ICIs across neoadjuvant, adjuvant, and perioperative settings for operable NSCLC, a Bayesian NMA was conducted. This approach synthesizes both direct and indirect evidence from RCTs, providing a robust framework to compare treatment strategies that have not been directly compared in clinical studies.

The study began with a systematic literature search performed in three major databases: PubMed, Embase, and the Cochrane Library. The search was completed up to September 1, 2024, using a comprehensive query that incorporated the terms “immune checkpoint inhibitors”, “non-small cell lung cancer”, and the specific treatment settings (“neoadjuvant”, “adjuvant”, and “perioperative”) alongside key survival endpoints such as “disease-free survival”, “event-free survival”, and “relapse-free survival”. This ensured the identification of all relevant studies that met the inclusion criteria.

Studies were included based on the following criteria: population (P): adults (≥18 years) with operable, histologically confirmed stage IB–IIIB NSCLC; interventions (I): ICIs administered in the neoadjuvant, adjuvant, or perioperative setting, either as monotherapy or in combination with chemotherapy; comparators (C): surgery alone or surgery with platinum-based chemotherapy; outcomes (O): DFS, EFS, or relapse-free survival (RFS) as the primary or co-primary endpoint; study design (S): phase 2 or 3 RCTs. Exclusion criteria included: (I) studies with only metastatic or unresectable NSCLC populations; (II) non-randomized studies; (III) insufficient survival outcome data; and (IV) duplicate publications.

Data extraction was performed by two independent reviewers to minimize bias and ensure accuracy. Key details such as study design, patient demographics, treatment regimens, and reported outcomes were systematically collected. Special attention was paid to the hazard ratios (HRs) for DFS, RFS, and EFS, along with their corresponding 95% credible intervals (CrI). The Cochrane risk of bias (RoB 2.0) tool was used to assess the methodological quality of included RCTs. Domains evaluated included: (I) randomization process; (II) deviations from intended interventions; (III) missing outcome data; (IV) measurement of outcomes; and (V) selection of reported results. Each study was rated as “low risk”, “some concerns”, or “high risk” of bias. Discrepancies were resolved by discussion.

Statistical analysis

Statistical analyses were conducted using a Bayesian framework within the Generalized Mixed Treatment Comparisons (GeMTC) package in R (6). This framework uses Markov Chain Monte Carlo (MCMC) simulations to estimate treatment effects, allowing for a comprehensive synthesis of direct and indirect evidence. A fixed-effect model was chosen based on the assumption that all studies estimate the same underlying treatment effect, with differences arising solely from random variation. Convergence of the MCMC chains was assessed through diagnostics such as Gelman-Rubin statistics, trace plots, and posterior density plots, ensuring reliable parameter estimates. Node-splitting models were employed to evaluate potential inconsistencies between direct and indirect comparisons within the network, a critical step for validating the internal consistency of the NMA.

The surface under the cumulative ranking curve (SUCRA) was used to rank the efficacy of the different treatments, with higher SUCRA values indicating greater likelihood of a treatment being the most effective. Additional measures, including residual deviance and deviance information criterion (DIC), were used to assess the model’s goodness of fit and overall performance. Sensitivity analyses were performed to test the robustness of the findings by excluding studies with higher risks of bias, applying random-effects models, and evaluating the impact of individual studies on overall treatment rankings.

Results

The NMA included eight RCTs, enrolling a total of 3,659 patients with operable NSCLC (Figure 1, Table 1). These trials investigated the use of ICIs across neoadjuvant, adjuvant, and perioperative settings, making them suitable for the comparative analysis. The studies encompassed various treatment approaches, including neoadjuvant nivolumab (CheckMate 816 and NADIM II), adjuvant atezolizumab (IMpower010) and pembrolizumab (PEARLS/KEYNOTE-091), and perioperative strategies involving toripalimab (NEOTORCH), nivolumab (CheckMate 77T), pembrolizumab (KEYNOTE-671), and durvalumab (AEGEAN) (1-5,7-9).

Figure 1.

Flow diagram of included studies.

Table 1. Characteristics of included studies.

Author, year	Type of study	No. of patients	Race	Histology (ADK/SCC), %	Stage (%)	PD-L1+, %	EGFR/ALK+, %	Exp arm	Ctr arm	Median follow up (months)	Primary endpoint	HR of DFS/EFS (95% CI)	Risk of bias
Cascone, 2024 (1)	Phase 3	461	White (67.7 vs. 75.4)	49.3/50.7 vs. 49.1/50.9	IIIA–B (63.8 vs. 64.2)	≥1% (55.9 vs. 55.2)	0/0	NIVO + CT ×4 → S → NIVO ×12 months	Plac + CT ×4 → S → Plac ×12 months	25.4	EFS	0.58 (0.42–0.81)	L
Felip, 2021 (2)	Phase 3	1,005	White (71 vs. 76)	65/35 vs. 67/34	IIIA (40 vs. 42)	≥1% (56 vs. 51)	10 vs. 13/3 vs. 4	S → CT ×4 → ATEZO ×12 months	S → CT ×4 → BSC	32.2	DFS	0.81 (0.67–0.99)	L
Forde, 2025 (5)	Phase 3	358	Asian (47.5 vs. 51.4)	51.4/48.6 vs. 46.9/53.1	IIIA (63.1 vs. 64.2)	≥1% (49.7 vs. 49.7)	0/0	NIVO + CT × 4 → S	CT ×4 → S	29.5	EFS	0.63 (0.43–0.91)	L
											PCR	13.94 (3.49–55.75)†
Heymach, 2023 (7)	Phase 3	802	White (56.3 vs. 51.1)	53.6/46.2 vs. 47.9/51.1	IIIA (47.3 vs. 44.1)	≥1% (66.7 vs. 66.6)	0/0	DURVA + CT ×4 → S → DURVA × 12	Plac + CT ×4 → S → Plac ×12	–	EFS, PCR	0.68 (0.53–0.88)	L
Lu, 2024 (3)	Phase 3	404	China	22.3/77.7 vs. 22.3/77.7	IIIA (67.3 vs. 67.3)	≥1% (65.9 vs. 65.4)	0/0	TORIPA + CT ×3 → S → CT ×1 → TORIPA ×13	Plac + CT ×3 → S → CT ×1 → Plac ×13	18.3	EFS, PCR	0.40 (0.28–0.57)	L
O’Brien, 2022 (8)	Phase 3	1,177	White (76 vs. 78)	67/33 vs. 62/38	II (56 vs. 58)	≥1% (60 vs. 60)	7 vs. 6/1 vs. 1	S → PEMBRO × 18	S → Place ×18	35.6	DFS	0.76 (0.63–0.91)	L
Provencio, 2023 (4)	Phase 2 random	86	Spain	44/37 vs. 38/48	IIIB (11 vs. 31)	≥1% (60 vs. 65.2)	0/0	NIVO + CT ×3 → S → NIVO ×6 months	CT ×3 → S	26.1	PCR	5.34 (1.34–21.23)‡	M
											PFS§	0.47 (0.25–0.88)
Wakelee, 2023 (9)	Phase 3	797	White (63 vs. 59.8)	56.9/43.1 vs. 56.8/43.2	IIIA (54.7 vs. 56.2)	≥1% (65.2 vs. 62.2)	3.5 vs. 4.8/3 vs. 2.2	PEMBRO + CT ×4 → S PEMBRO ×13	Plac + CT ×4 → S → Plac ×13	25.2	EFS	0.58 (0.46–0.72)	L
											OS	0.73 (0.54–0.99)

^†, odds ratio; ^‡, relative risk; ^§, secondary end point. ADK, adenocarcinoma; ALK, anaplastic lymphoma kinase; ATEZO, atezolizumab; BSC, best supportive care; CI, confidence interval; Ctr, control; CT, chemotherapy; DFS, disease-free survival; DURVA, durvalumab; EFS, event-free survival; EGFR, epidermal growth factor receptor; Exp, experimental; HR, hazard ratio; L, low; M, moderate; NIVO, nivolumab; PD-L1, programmed death-ligand 1; PCR, pathological complete response; PEMBRO, pembrolizumab; plac, placebo; S, surgery; SCC, squamous cell carcinoma; TORIPA, toripalimab.

Summary of network geometry

The abundance of trials and patient representation in this NMA reflects the increasing interest in using ICIs for operable NSCLC. Across eight RCTs, a total of 3,659 patients were included, making this a robust analysis of perioperative, neoadjuvant, and adjuvant strategies. Notably, perioperative ICIs had the strongest presence in the network, reflecting the growing belief that administering immunotherapy both before and after surgery might offer superior benefits. However, adjuvant and neoadjuvant treatments had fewer studies, raising questions about whether these strategies are receiving enough clinical attention. Patient demographics also varied, with a significant number of studies enrolling predominantly Asian or White populations, leading to potential concerns about generalizability across different ethnic groups.

The treatment network structure allowed for meaningful comparisons across different ICI-based approaches, even in the absence of direct head-to-head trials. Bayesian modeling facilitated indirect comparisons, ranking the effectiveness of various strategies based on DFS and EFS. Among the standout findings, perioperative toripalimab combined with chemotherapy emerged as the most effective treatment, achieving the highest probability of ranking first, followed closely by perioperative nivolumab. These perioperative regimens outperformed neoadjuvant and adjuvant strategies, suggesting that ICIs might be most effective when administered before and after surgery rather than in just one phase of treatment. Neoadjuvant nivolumab with chemotherapy did demonstrate improved EFS over surgery alone, but its overall impact was still lower compared to perioperative approaches. The rankings indicate that the timing of ICI administration plays a crucial role in maximizing treatment benefit, though the exact mechanisms behind these differences remain a topic of ongoing investigation.

Despite the strengths of this analysis, some important gaps in evidence remain. A major limitation is the absence of direct trials comparing perioperative ICIs with neoadjuvant or adjuvant approaches, making most insights reliant on indirect comparisons. This lack of head-to-head data means that while perioperative ICIs appear superior, definitive proof is still lacking. Additionally, inconsistencies in control arms across different studies add complexity—some trials used surgery alone as the baseline, while others included chemotherapy, leading to variations in treatment effects. Another critical gap is the limited availability of long-term OS data, as most trials focused primarily on DFS or EFS. Without clear evidence that these benefits translate into extended patient survival, the long-term impact of different strategies remains uncertain. Furthermore, the role of biomarkers such as PD-L1 expression is not fully resolved, as different trials had varying inclusion criteria regarding PD-L1 positivity, making it difficult to determine how much this factor influences treatment outcomes.

The treatment network structure also introduces some potential biases that should be acknowledged. One concern is publication bias—negative trials that show little or no benefit of ICIs may be underreported, leading to an inflated perception of their effectiveness. Selection bias is another issue, as many trials focused on patients with good performance status (ECOG 0–1), limiting the applicability of these findings to real-world populations where comorbidities and frailty are more common. Study heterogeneity, particularly differences in follow-up duration, chemotherapy regimens, and treatment cycles, further complicates the interpretation of results. In some cases, discrepancies in endpoint definitions—such as variations in how DFS and EFS were measured—add another layer of complexity, making it challenging to directly compare outcomes across studies. These factors highlight the need for future research that standardizes methodology and directly compares different ICI strategies in well-controlled trials.

EFS/RFS results of NMA

The analysis revealed that perioperative ICIs, particularly when combined with chemotherapy, demonstrated the highest efficacy in DFS and EFS. Among the evaluated regimens, perioperative toripalimab emerged as the top-performing treatment, with a SUCRA value of 0.99 and a probability rank of 87% for being the most effective (Table S1; Figure 2; Figures S1,S2). Nivolumab-chemotherapy in the perioperative setting followed closely, achieving a SUCRA value of 0.94 and a second-best ranking probability of 41%. Both treatments significantly outperformed surgery alone or surgery combined with chemotherapy.

Figure 2.

Comparison of the included interventions: hazard ratio (95% CrI). Each cell gives the effect of the column-defining intervention relative to the row-defining intervention. ATEZO, atezolizumab; CrI, credible interval; CT, chemotherapy; DURVA, durvalumab; HR, hazard ratio; NIVO, nivolumab; PEMBRO, pembrolizumab; S, surgery; TORI, toripalimab.

Direct and indirect comparisons confirmed the superiority of perioperative ICIs over neoadjuvant and adjuvant approaches. For instance, perioperative toripalimab demonstrated a HR of 0.64 (95% CrI: 0.53–0.75) when compared to adjuvant atezolizumab, reflecting a marked improvement in survival outcomes. Similarly, perioperative nivolumab showed a HR of 0.72 (95% CrI: 0.62–0.85) against the same comparator, underscoring its effectiveness in reducing the risk of recurrence.

Neoadjuvant ICIs, while offering benefits such as tumor downstaging and earlier intervention for micrometastatic disease, exhibited relatively lower efficacy compared to perioperative strategies. For example, neoadjuvant nivolumab (CheckMate 816) achieved a HR of 0.79 (95% CrI: 0.67–0.93) relative to surgery alone. Similarly, adjuvant ICIs, despite demonstrating survival advantages over traditional treatments, ranked lower in the SUCRA hierarchy. Adjuvant atezolizumab (IMpower010) had a HR of 0.85 (95% CrI: 0.74–0.98), indicating limited impact compared to perioperative regimens.

Model fit comparison (consistency vs. inconsistency models)

The investigation of inconsistency in the NMA is an essential step in ensuring that direct and indirect comparisons across treatment arms align well, maintaining the validity of the model’s conclusions. One of the primary ways this is assessed is by evaluating the overall model fit, which can be done by comparing the consistency and inconsistency models. In this case, the DIC was reported as 12.4, a slightly improved value compared to the previous model (DIC: 15.2). Since a lower DIC suggests a better fit, this finding supports the assumption that the consistency model remains appropriate. Additionally, the residual deviance value of 6.4 is close to the number of data points, indicating that the model is well-calibrated and there is no significant overfitting or underfitting. Given that inconsistency models tend to show lower residual deviance when there is significant network inconsistency, the results here suggest that applying an inconsistency model would not provide a substantially better fit.

Further analysis of the leverage versus residual deviance plot helps identify whether any specific study exerts an undue influence on the model. In this case, the distribution of studies within expected ranges indicates that none are acting as extreme outliers. The study by Provencio (4), for example, had the lowest residual deviance (0.814) and leverage (0.809), suggesting that it fit the model well with minimal influence on the overall results. Meanwhile, studies such as Heymach (7) and Cascone (1) displayed slightly higher residual deviance values (1.018 and 1.012, respectively), but these remained within an acceptable range. This even distribution of leverage values further reinforces the assumption of consistency across the network and suggests that no single study is disproportionately affecting the model’s estimates.

A key aspect of inconsistency analysis is examining residual deviance estimates across the treatment network. If certain studies showed particularly high residual deviance values compared to others, this could suggest localized inconsistency in the model. However, in this case, the residual deviance values range from 0.814 to 1.018, a relatively narrow distribution, further supporting the idea that the network is consistent. In a scenario where major inconsistencies existed, one would expect to see substantial deviations in these values across different comparisons, indicating that direct and indirect evidence do not align well. Since no such deviations were observed, this provides further confidence in the validity of the results.

Sensitivity analyses reinforced the robustness of the results. Excluding studies with higher risks of bias or applying random-effects models did not substantially alter the treatment rankings, confirming the stability of the conclusions. Additionally, the results remained consistent when individual studies were removed, demonstrating that any single trial did not unduly influence the overall findings.

Discussion

This NMA provides critical insights into the efficacy of ICIs across various therapeutic contexts—neoadjuvant, adjuvant, and perioperative—for operable NSCLC (10,11). By synthesizing data from RCTs, this study elucidates the optimal timing and regimen for ICIs in early-stage NSCLC, thereby addressing a significant gap in contemporary clinical knowledge. The ensuing discussion interprets these results, considering their broader implications for clinical practice and future research.

The analysis highlights the clear superiority of perioperative ICIs, particularly when administered in conjunction with chemotherapy. Among the treatment strategies evaluated, perioperative toripalimab and nivolumab emerged as the most effective, with toripalimab exhibiting a probability rank of 87% and a SUCRA value of 0.99. These findings indicate a robust confidence in its efficacy. Nivolumab combined with chemotherapy also achieved a high SUCRA value of 0.94, underscoring its reliability and potential as a preferred treatment option. The remarkable performance of perioperative strategies is likely attributable to their capacity to target micrometastatic disease and enhance the immune response to residual tumor cells, thus reducing the risk of recurrence and improving survival outcomes.

In contrast, neoadjuvant and adjuvant strategies ranked lower in terms of efficacy. Although neoadjuvant ICIs theoretically provide benefits such as tumor downstaging and early intervention for micrometastatic disease, their effectiveness may be compromised by the immunosuppressive tumor microenvironment characteristic of untreated NSCLC. Conversely, adjuvant ICIs encounter challenges, including the potential loss of antigenic targets following tumor resection, which diminishes their impact on residual disease. These observations support the growing consensus favoring perioperative ICIs as the most effective approach for operable NSCLC.

Our findings align with and extend prior work by He et al., who performed an NMA restricted to neoadjuvant immunochemotherapy versus perioperative approaches, reporting a potential benefit for perioperative regimens (12). In contrast, our analysis includes a broader scope—evaluating all three strategies including adjuvant ICIs—and integrates additional recent trials, such as CheckMate 77T and KEYNOTE-671, offering a more comprehensive overview of current treatment paradigms. The perioperative period offers a unique opportunity for immunotherapy, wherein surgery-induced immune activation can synergistically interact with ICIs to eradicate residual tumor cells and micrometastases, thereby enhancing long-term survival outcomes. This hypothesis is strongly supported by the superior performance of perioperative toripalimab and nivolumab observed in the analysis.

These findings have significant implications for clinical practice, indicating a need for a paradigm shift in the management of operable NSCLC. However, the implementation of perioperative ICIs in real-world settings necessitates careful consideration of logistical and clinical challenges. First, patient selection is critical to ensure that candidates are suitable for the combined systemic and surgical therapies, which require optimal physical and physiological reserves. Second, regulatory and accessibility issues must be addressed. For example, while toripalimab demonstrated exceptional efficacy, its limited approval for lung cancer in Europe and North America restricts its immediate application in these regions. Expanding regulatory approvals and addressing cost-effectiveness are essential for broader adoption.

The diversity of patient populations represented in the included trials also merits attention. Most studies predominantly involved Asian populations, raising questions regarding the generalizability of the findings to other demographic groups. While NSCLC biology may exhibit similarities across populations, further research is necessary to confirm the universal applicability of perioperative ICIs and to ensure equitable benefits across diverse patient cohorts.

Despite its strengths, this analysis also acknowledges certain limitations. The heterogeneity among included studies concerning design, patient characteristics, and follow-up durations introduces potential biases that could influence the results. Furthermore, the limited representation of neoadjuvant and adjuvant strategies restricts definitive conclusions regarding their relative efficacy. Additionally, the focus on DFS and EFS as primary endpoints, although valuable, does not provide a comprehensive assessment of OS—a critical metric for evaluating long-term treatment efficacy.

Looking forward, this NMA identifies several critical areas for future research. Direct head-to-head trials comparing perioperative, neoadjuvant, and adjuvant approaches are essential for validating these findings and informing clinical decision-making. Moreover, identifying predictive biomarkers for ICI response could facilitate personalized treatment strategies, optimize outcomes while minimizing unnecessary toxicity. Investigating novel combinations of ICIs with radiotherapy, targeted agents, or other immunotherapies may further enhance therapeutic efficacy. Real-world evidence from post-approval studies and registries will play an equally pivotal role in assessing the efficacy and safety of perioperative ICIs in broader clinical practice. Such data will provide valuable insights into their performance outside controlled trial settings, particularly in patient populations with varying baseline characteristics and comorbidities (13,14).

Conclusions

In conclusion, this NMA underscores the transformative potential of perioperative ICIs, particularly toripalimab and nivolumab, in improving outcomes for patients with operable NSCLC. These findings represent a significant advancement, offering new hope for enhancing survival and reducing recurrence risks in this challenging disease. However, translating these findings into clinical practice necessitates addressing logistical, regulatory, and research challenges to ensure that these promising therapies reach the patients who stand to benefit most. As the field continues to evolve, perioperative ICIs are poised to redefine the treatment paradigm for early-stage NSCLC, marking a pivotal advancement in the pursuit of improved patient outcomes.

Supplementary

The article’s supplementary files as

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.