Efficacy and safety of perioperative, adjuvant and neoadjuvant chemoimmunotherapy stratified by clinical stage and PD-L1 expression in resectable non-small cell lung cancer: a systematic review and network meta-analysis

Introduction

Lung cancer is one of the most common malignancies worldwide, and non-small cell lung cancer (NSCLC) accounts for approximately 85% of cases (1). Approximately 30% cases of NSCLC are diagnosed in the early stage and receive surgical resection with curative intent, but the 5-year survival rates remain unsatisfactory, ranging from 36% for stage IIIA to 68% for stage IB (2-4). Additionally, conventional preoperative chemotherapy confers only a 5% absolute benefit on 5-year survival over surgery alone (5).

Immune-checkpoint inhibitors (ICIs), mainly monoclonal antibodies blocking programmed cell death protein 1 (PD-1) or programmed cell death ligand 1 (PD-L1), have revolutionized the treatment landscape for patients with resectable NSCLC. Several randomized controlled trials (RCTs) proved the clinical benefit with perioperative (6-11), adjuvant (12,13), and neoadjuvant (14) chemoimmunotherapies compared to chemotherapies. However, it is still unclear which one of these three strategies is more favorable. In addition, whether clinical stage and PD-L1 expression influence treatment decision-making also remains a matter of debate. As no clinical trials have compared these three strategies head-to-head, a well-designed meta-analysis might be available to answer these questions.

Several meta-analyses have evaluated chemoimmunotherapy strategies in resectable NSCLC. Three meta-analyses have compared neoadjuvant chemoimmunotherapy to neoadjuvant chemotherapy (15-17), one compared both neoadjuvant and adjuvant chemoimmunotherapy to neoadjuvant and adjuvant chemotherapy (18), and one compared the perioperative chemoimmunotherapy to neoadjuvant chemoimmunotherapy (19). To date, there is still lacking network meta-analysis to compare the efficacy and safety among perioperative, adjuvant, and neoadjuvant chemoimmunotherapies. Additionally, several RCTs have updated efficacy and safety data with long-term follow-up (7-9,11,13,14,20). Therefore, we conducted this systematic review and network meta-analysis to assess the efficacy and safety of these three chemoimmunotherapy strategies in resectable NSCLC with latest follow-up data. Moreover, we conducted subgroup analyses by disease stage (21), PD-L1 expression (22) and Eastern Cooperative Oncology Group (ECOG) performance status (PS) to better inform personalized treatment decision-making. We present this article in accordance with the PRISMA NMA reporting checklist (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-426/rc).

Methods

Registration of the study protocol was completed in PROSPERO (Prospective Register of Systematic Reviews; No. CRD42023493137).

Data sources and search strategies

We conducted a comprehensive search for studies until April 1, 2025. Literature from PubMed, the Cochrane Central Register of Controlled Trials, and registered studies in ClinicalTrials.gov were systematically searched to identify eligible studies. Our search terms included adjuvant, neoadjuvant, perioperative, immunotherapy, NSCLC, and RCTs in various combinations (Table S1). Additionally, we expanded our search by reviewing publications from several major conference proceedings, such as the European Society of Medical Oncology (ESMO), the American Society of Clinical Oncology (ASCO), the American Association for Cancer Research (AACR), the European Lung Cancer Congress (ELCC) and the World Conference on Lung Cancer (WCLC). The final selection was further validated by manual search of the reference lists of the retrieved studies and all existing reviews. In cases where multiple reports describing the same population were available, we only included the latest and comprehensive report.

Study selection

The studies included in the analysis encompassed published and unpublished phase II/III randomized controlled clinical trials that met the inclusion criteria: (I) randomized controlled phase II or III clinical trials; (II) studies that enrolled patients diagnosed with surgically treatable NSCLC, and confirmed through either histological or cytological examination; (III) studies assessing chemotherapy or immune-checkpoint inhibition in the adjuvant, neoadjuvant, or perioperative setting; (IV) studies that reported one or more of the following detailed outcomes or adverse events including: event-free survival (EFS), overall survival (OS), treatment-related adverse events (TRAEs) of grade 3 or higher. Studies failing to meet these criteria were excluded. Other exclusion criteria were: (I) single arm experiment; (II) unresectable advanced NSCLC; (III) RCTs without platinum-based doublet chemotherapy.

We systematically screened titles and abstracts, followed by a sequential assessment of the full text of articles that potentially met the inclusion criteria for final incorporation.

Data extraction and quality assessment

Two researchers (Y.W. and Lin Zhang) independently extracted relevant data using a pre-designed form. Disputed data were discussed with a third researcher (D.S.). Data were extracted on study ID, publication year, study design, number of patients enrolled, type of treatment, and patient characteristics. Survival data were collected as the hazard ratios (HRs) with 95% confidence interval (CI) of EFS, DFS, and OS in the intention-to-treat (ITT) population and assessed by the independent review committee, the investigator-assessed outcomes were included only if independent review was unavailable. Dichotomous data such as grade ≥3 adverse events (grade ≥3 AEs) were collected as the number of events in both the experimental and control arm. We favored the utilization of adverse events associated with treatment. According to the National Comprehensive Cancer Network guidelines (version 2.2023), platinum-based chemotherapy is recommended for resectable NSCLC (23). Therefore, we only included platinum-based doublet chemotherapy in the network meta-analysis to ensure better consistency.

The Cochrane Risk of Bias tool for randomized trials (RoB2) was used to determine the study’s methodological quality by assessing the risk of bias. The RoB2 tool evaluated bias in domains such as randomization process, deviations from intended interventions, outcome data measurement and reported result selection (24,25). Each domain was scored as “low”, “high”, or “some concerns” by two reviewers with adequately prepared empirical evidence. Every researcher conducted study selection and data extraction separately. Two investigators (Y.W. and Lin Zhang) individually evaluated the risk of bias in each study. Any disagreements were resolved through consensus and arbitration by members of the arbitration panel (Y.W., Lin Zhang, D.S., and H.Z.).

To ensure the comprehensiveness and reliability of the meta-analysis, it is critical to consider the role of publication bias. Thus, statistical funnel plots and Egger’s test were employed to assess for publication bias of the included studies. A symmetrical distribution of studies in the funnel plot was considered indicative of no substantial publication bias.

Statistical analysis

To compare the efficacy and safety of different chemoimmunotherapy approaches, we incorporated both direct and indirect evidence in the network meta-analysis. Since data on OS were not mature and might cause heterogeneity, we set EFS as primary outcome to assess each treatment’s realistic effect. The secondary outcomes included OS and occurrences of grade 3 or higher adverse events. In terms of the subgroup analysis of clinical stage, PD-L1 expression level and ECOG PS, EFS was used as the clinical endpoint to compare the different clinical benefit of the three models.

In Stata (version 17.0), the network diagrams were created to visually represent the structures of various outcomes and specify the direct or indirect comparisons included in this study (26). The pooled estimates for survival outcomes EFS and OS were reported as HRs with corresponding 95% credible intervals (CrI) and P values calculated using the inverse-variance-weighted method, while the measures for dichotomous data (grade ≥3 adverse events) were pooled with the odds ratios (ORs), with corresponding 95% CrIs and P values calculated using the Mantel-Haenszel method. We used the Cochrane’s Q statistic and calculated the I² statistic to assess the percentage of total variation across studies due to heterogeneity (27). The statistical significance threshold was set at a P value of 0.05. Heterogeneity levels were classified as low, moderate, or high depending on estimated I² values below 25%, between 25% and 50%, and above 50%, respectively. The random effect models were chosen if obvious heterogeneity was present (I²>25%). Otherwise, the fixed effect models were applied (24).

We performed the network meta-analysis within a Bayesian framework using rjags and gemtc packages in R software (28,29). The deviation information criterion (DIC) evaluates model fit while considering the number of effective parameters, thus indicating the most suitable model depending on the lowest DIC value (30). For EFS and OS effects, 50,000 sample iterations were generated with 20,000 burn-ins and a thinning interval of 1. For toxic effects, 100,000 sample iterations were generated with 50,000 burn-ins and a thinning interval of 1.

Four different sets of initial values were used in the Markov Chain Monte Carlo (MCMC) simulation. We assessed the convergence of iterations by visually examining the four chains to ensure consistent parameter estimates and following the Brooks-Gelman-Rubin diagnostic (Figure S1). Credible intervals, which were displayed as CrI in the following context, can be understood similarly to conventional confidence interval in the presence of minimally informative priors. Moreover, a Bayesian framework meta-analysis provided a rank probability to assign probabilities to each treatment’s rank by calculating the surface under the cumulative ranking curve. A value of 1 indicates high certainty that a treatment is superior, while a value of 0 suggests high certainty that it is inferior (31).

Transitivity and consistency

Transitivity was evaluated by comparing baseline characteristics between experimental and control groups, including sample size, age, sex, ECOG PS, histological classification, smoking status, disease stage, and PD-L1 expression. Moreover, the global inconsistency was assessed by comparing the fit of consistency and inconsistency models, and local inconsistency was assessed by using Bayesian P values by the node splitting analysis. Pairwise meta-analyses were conducted within a Bayesian framework and complemented by frequentist framework results to compare the pooled HR or OR from network meta-analyses for assessing detailed local inconsistency.

Given the variability in study design, execution, and patient characteristics across the included trials, which could introduce substantial heterogeneity and potential bias in the meta-analysis outcomes. For instance, Timothy’s study (32) includes only stage I–II NSCLC patients, whereas NADIM II, NEOTORCH, Wei’s study (33) and Rosell’s study (34) focus on stage III patients. The meta-regression analyses were performed on EFS, OS and grade ≥3 TRAEs to evaluate the potential influence caused by confounding variables including the percentage of patients with stage III, the treatment cycles, male percentage, median age and sample size, providing a clearer understanding of how these factors impact the efficacy and safety of chemoimmunotherapy.

Sensitivity analysis

To quantify the impact of study heterogeneity on the results and assess the robustness of the conclusions, a sensitivity analysis was conducted to assess the impact of included studies identified as high-risk on the overall meta-analysis findings. Furthermore, different from the neoadjuvant chemotherapy of other perioperative studies, the NEOTORCH trial added a cycle of adjuvant chemotherapy in both groups, which was obvious different in study design. To avoid the potential risk of bias, we excluded the NEOTORCH trial and performed the second sensitivity analysis in the overall population and subgroups.

Results

Study selection and characteristics

We identified 1,136 records during the initial screening of titles and abstracts, and subsequently examined 214 full-text reports. Twenty-one randomized controlled clinical trials comprising a total of 12,258 patients were included in the network meta-analysis (Figure 1).

Figure 1 PRISMA 2020 flow diagram of the study selection process. The diagram illustrates the flow of information through the different phases of this systematic review. PRISMA, Preferred Reporting Items for Systematic Reviews and Meta-analyses.

Overall, six approaches of treatment were investigated, including perioperative chemoimmunotherapy, adjuvant chemoimmunotherapy, neoadjuvant chemoimmunotherapy, adjuvant chemotherapy, neoadjuvant chemotherapy and surgery alone (Figure 2A). Characteristics of the nine trials testing immune-checkpoint inhibition are summarized in Table 1. Since all patients in the IMpower010 study and the vast majority of patients in the KEYNOTE-091 study (1,010/1,177) had received adjuvant chemotherapy before randomization. Both studies were considered as adjuvant chemoimmunotherapy versus adjuvant chemotherapy. The characteristics of the other trials comparing chemotherapy with surgery alone are summarized in Table S3. The detailed results of the RoB2 assessments are shown in Figure S2. The assumption of transitivity was accepted since no variability was observed in the baselines of the study and population (Figure S3).

Figure 2 Efficacy and safety of chemoimmunotherapy strategies in resectable NSCLC: a network meta-analysis. The figure presents the results of a Bayesian network meta-analysis comparing six treatment strategies for resectable NSCLC. (A) Network diagram of included randomized controlled trials. Each node represents a treatment strategy, with its size proportional to the total number of patients randomized to that arm. The lines (edges) connect treatments that have been directly compared in trials, with the width of the line proportional to the number of studies for that comparison, as indicated by the numerical key 1–6. The network structure for EFS, OS, and grade ≥3 AEs is shown. (B) League table of pooled HRs for efficacy outcomes. The upper triangle displays HRs for EFS, and the lower triangle displays HRs for OS. An HR <1.0 indicates a lower risk of an event (progression or death) and favors the row-defining treatment over the column-defining treatment. (C) League table of pooled ORs for safety outcomes. The table displays ORs for grade ≥3 treatment-related AEs. An OR <1.0 indicates lower odds of severe AEs and favors the row-defining treatment. All estimates are presented with their 95% CrIs. (D) Ranking probability plots. These curves show the probability of each treatment achieving a specific rank for EFS (orange), OS (blue), and grade ≥3 AEs (red). For efficacy outcomes (EFS and OS), a high probability at a low rank indicates a more favorable treatment. For toxicity (AEs), a high probability at a low rank indicates a more toxic treatment. These curves correspond to the Bayesian ranking results presented in Table S2. Adj Chemo, adjuvant chemotherapy; Adj ChemoIO, adjuvant chemoimmunotherapy; AEs, adverse events; EFS, event-free survival; HRs, hazard ratios; Neo Chemo, neoadjuvant chemotherapy; Neo ChemoIO, neoadjuvant chemoimmunotherapy; NSCLC, non-small cell lung cancer; ORs, odds ratios; OS, overall survival; Peri ChemoIO, perioperative chemoimmunotherapy.

Network meta-analysis and rank probabilities

The network meta-analysis included six treatment approaches, assessing HRs for EFS and OS, and ORs for TRAEs of grade 3 or higher. All extracted EFS and OS data from the included immunochemotherapy cohorts were based on the ITT population assessed by the independent review committee. The fixed-effect consistency model was chosen with the smallest DIC among all treatment effects (Table S4).

In terms of EFS, the combination of ICIs and chemotherapy significantly prolonged EFS compared with chemotherapy (Figure 2B). Moreover, perioperative chemoimmunotherapy was associated with improved EFS versus adjuvant chemoimmunotherapy (HR 0.75, 95% CI: 0.61 to 0.94; P=0.01), but comparable to neoadjuvant chemoimmunotherapy (HR 0.84, 95% CI: 0.60 to 1.17; P=0.31). Bayesian ranking was in line with the pooled analyses above. Perioperative chemoimmunotherapy had the highest probability of providing the best EFS (84.73%), followed by neoadjuvant and adjuvant chemoimmunotherapy (Table S4).

In terms of OS, perioperative chemoimmunotherapy was also associated with improved OS versus adjuvant chemoimmunotherapy (HR 1.34, 95% CI: 1.01 to 1.78; P=0.042), but comparable to neoadjuvant chemoimmunotherapy (HR 1.07, 95% CI: 0.68 to 1.69; P=0.77). Perioperative chemoimmunotherapy was most likely to be ranked first (60.43%) in prolonging OS, followed by neoadjuvant and adjuvant chemoimmunotherapy.

In terms of grade ≥3 AEs, perioperative and adjuvant chemoimmunotherapy was noted with more toxicities than neoadjuvant (OR 1.30, 95% CI: 1.12 to 1.51; P<0.001) and adjuvant (OR 1.74, 95% CI: 1.42 to 2.13; P<0.001) chemotherapy, respectively. However, neoadjuvant chemoimmunotherapy did not increase toxicity versus neoadjuvant chemotherapy (OR 0.80, 95% CI: 0.51 to 1.24; P=0.33). Furthermore, perioperative chemoimmunotherapy showed greater toxicity than neoadjuvant chemoimmunotherapy (OR 1.63, 95% CI: 1.02 to 2.59; P=0.041; Figure 2C). Bayesian ranking indicated that perioperative (53.20%) and adjuvant (45.69%) chemoimmunotherapy were more likely to cause grade ≥3 AEs (Figure 2D). Regarding immune-related adverse event (irAE) profiles, prolonged immune checkpoint inhibitor (ICI) exposure in the perioperative and adjuvant chemoimmunotherapy settings led to increased risks of rash, hypothyroidism, pneumonitis, and hyperthyroidism compared to chemotherapy alone. In contrast, neoadjuvant chemoimmunotherapy notably increased the risk of rash and demonstrated a relatively safer toxicity (Figure S4). Further irAE comparisons between perioperative and neoadjuvant chemoimmunotherapy revealed that hypothyroidism or thyroiditis occurred more frequently with perioperative chemoimmunotherapy than with neoadjuvant chemoimmunotherapy (OR 5.53, 95% CI: 1.69 to 16.61; P=0.003) (Figure 3).

Figure 3 Incidence and comparative risk of irAEs. This figure illustrates the safety profile of different chemoimmunotherapy strategies with a focus on common irAEs of any grade. (A) Incidence of common irAEs. This bar chart shows the absolute incidence rates (%) of specific irAEs (rash, hypothyroidism, pneumonitis, and hyperthyroidism) for perioperative, adjuvant, and neoadjuvant chemoimmunotherapy, as well as for chemotherapy alone. (B) Forest plot comparing irAEs between perioperative and neoadjuvant chemoimmunotherapy. This plot displays the ORs and 95% CrIs for the risk of specific irAEs. The squares represent the point estimate of the OR, and the horizontal lines represent the 95% CrI. The vertical dashed red line at OR =1.0 indicates no difference in risk between the two strategies. An OR >1.0 suggests a higher risk of the event with perioperative chemoimmunotherapy compared to neoadjuvant chemoimmunotherapy. Adj Chemo, adjuvant chemotherapy; Adj ChemoIO, adjuvant chemoimmunotherapy; CrI, credible interval; irAE, immune-related adverse event; Neo Chemo, neoadjuvant chemotherapy; Neo ChemoIO, neoadjuvant chemoimmunotherapy; OR, odds ratio; Peri ChemoIO, perioperative chemoimmunotherapy.

Meta-regression analyses

In meta-regression analyses, the coefficient reflects the relationship between a covariate and the effect size across studies. Specifically, it quantifies how much the effect size changes with each unit increase in the covariate, with the effect size expressed as the natural logarithm of the HR or OR in our study.

In the network meta-regression analysis of EFS, the proportion of patients with stage III (coefficient =−0.0042, P=0.001) and the sample size (coefficient =0.00041, P=0.02) were potential covariates influencing the evaluated outcomes. In the network meta-analysis of OS, sample size was the only covariate potentially modifying treatment effects (coefficient =0.00047, P=0.03). Notably, although the findings were statistically significant, their impact on the evaluated outcomes was limited. Moreover, no covariate showed significant influence on adverse events of grade 3 or higher. These findings indicate the robustness of the overall network meta-analysis and provide a reference for further subgroup analyses (Figure 4, Table S5).

Figure 4 Network meta-regression analysis of potential effect modifiers. This figure displays bubble plots from the network meta-regression, performed to examine whether study-level covariates systematically modify treatment effects on (A) EFS, (B) OS and (C) grade ≥3 adverse events. Each bubble represents the estimated meta-regression coefficient for a given covariate. Its horizontal position corresponds to the coefficient calculated on the natural-log scale of the hazard ratio or odds ratio. The size of each bubble is inversely proportional to the standard error of the estimate, meaning larger bubbles represent more precise estimates. The color gradient corresponds to the statistical significance of the association, where red indicates more significant associations, and blue indicates less significant ones. Stars indicate statistical significance: *, P<0.05. AEs, adverse events; EFS, event-free survival; OS, overall survival.

Subgroup analyses by disease stage, PD-L1 expression and ECOG PS

A total of 8,346 patients were included in a disease stage-stratified network meta-analysis, with participants divided into stage I–II and stage III subgroups for separate comparisons. Additionally, 11,927 patients were incorporated into a parallel network meta-analysis stratified by PD-L1 expression levels, categorized into three subgroups based on tumour proportion score (TPS): TPS <1%, TPS 1–49%, and TPS ≥50%. Furthermore, 5,155 patients were included in the subgroup network meta-analysis stratified by ECOG PS. The patient counts and corresponding network diagrams for each subgroup are detailed in Figure S5. The subgroup analysis was performed by a Bayesian fixed-effect consistency model (Figure 5 and Table S6). The EFS benefit of ICIs differed significantly among patients with various clinical stage, PD-L1 expression and ECOG PS.

Figure 5 Subgroup network meta-analysis of EFS. This figure displays the results of the network meta-analysis for EFS stratified by key clinical and biomarker subgroups. Each panel shows a league table of pooled HRs with 95% CrIs and the corresponding cumulative ranking probability curves for the three chemoimmunotherapy strategies. An HR <1.0 favors the row-defining treatment. In the ranking curves, a curve shifted to the upper-left indicates a higher probability of being the most effective strategy for EFS. The subgroups are: (A) patients with stage IB–II NSCLC; (B) patients with stage IIIA–IIIB NSCLC; (C) patients with PD-L1 TPS <1%; (D) patients with PD-L1 TPS 1–49%; (E) patients with PD-L1 TPS ≥50%; (F) patients with ECOG PS =0; (G) patients with ECOG PS ≥1. Adj ChemoIO, adjuvant chemoimmunotherapy; CrI, credible interval; ECOG PS, Eastern Cooperative Oncology Group performance status; EFS, event-free survival; HR, hazard ratio; Neo ChemoIO, neoadjuvant chemoimmunotherapy; NSCLC, non-small-cell lung cancer; Peri ChemoIO, perioperative chemoimmunotherapy; TPS, tumor proportion score.

In patients with stage IB to II NSCLC, both perioperative chemoimmunotherapy (HR 0.68, 95% CI: 0.49 to 0.96; P=0.03) and adjuvant (HR 0.73, 95% CI: 0.57 to 0.94; P=0.01) chemoimmunotherapy yielded significant survival benefit versus chemotherapy. Furthermore, perioperative chemoimmunotherapy was comparable with adjuvant chemoimmunotherapy (HR 1.06, 95% CI: 0.64 to 1.77; P=0.82) and neoadjuvant chemoimmunotherapy (HR 0.78, 95% CI: 0.37 to 1.62; P=0.51) in terms of EFS. Bayesian ranking revealed that adjuvant chemoimmunotherapy had the highest probability of providing the best EFS (51.79%), followed by perioperative chemoimmunotherapy (33.21%) (Figure 5A).

In patients with stage IIIA to IIIB NSCLC, both perioperative (HR 0.54, 95% CI: 0.47 to 0.63; P<0.001) and neoadjuvant chemoimmunotherapy (HR 0.54, 95% CI: 0.37 to 0.79; P=0.001) yielded significant survival benefit versus neoadjuvant chemotherapy. Of note, perioperative chemoimmunotherapy (HR 0.32, 95% CI: 0.16 to 0.64; P=0.001) and neoadjuvant chemoimmunotherapy (HR 0.31, 95% CI: 0.14 to 0.69; P=0.004) were associated with significantly improved EFS compared to adjuvant chemoimmunotherapy. Neoadjuvant chemoimmunotherapy had the highest probability of providing the best EFS (51.40%), followed by perioperative chemoimmunotherapy (48.65%) (Figure 5B).

In the subgroup analysis of PD-L1 expression, we noticed a correlation between the benefit of chemoimmunotherapy and PD-L1 expression levels. In patients with PD-L1 TPS ≥50%, all three chemoimmunotherapy approaches demonstrated significant superiority over chemotherapy. In patients with PD-L1 TPS of 1–49%, both perioperative (HR 0.56, 95% CI: 0.45 to 0.70; P<0.001) and adjuvant chemoimmunotherapy (HR 0.75, 95% CI: 0.59 to 0.96; P=0.02) showed significant benefit versus chemotherapy. In patients with PD-L1 TPS <1%, only perioperative chemoimmunotherapy significantly improved EFS versus chemotherapy (HR 0.74, 95% CI: 0.60 to 0.91; P=0.005). However, no significant difference was found among the three chemoimmunotherapy approaches, regardless of PD-L1 expression (Figure 5C-5E).

Furthermore, subgroup analysis indicated that the efficacy of chemoimmunotherapy was associated with the patient’s ECOG PS. While all three chemoimmunotherapy regimens significantly outperformed chemotherapy in patients with an ECOG PS of 0, only perioperative chemoimmunotherapy led to a significant improvement in EFS for those with an ECOG PS of 1 or higher (HR 0.55, 95% CI: 0.43 to 0.72; P<0.001) (Figure 5F,5G).

Heterogeneity and bias assessment

Heterogeneity analysis was shown in the forest plots of pairwise comparisons (Figures S6-S9). In the analysis of EFS, all comparisons showed low heterogeneity (I²<25%). In subgroup analyses of disease stage and PD-L1 status, most comparisons between chemoimmunotherapy and chemotherapy showed low heterogeneity, except for perioperative chemoimmunotherapy versus neoadjuvant chemotherapy in PD-L1 TPS of 1–49% (55.5%) and TPS ≥50% (30.7%) population, and adjuvant chemoimmunotherapy versus chemotherapy in PD-L1 TPS ≥50% (78.4%) population. Additionally, results from pairwise and network meta-analyses were highly consistent (Figure S10). In the node splitting analysis, no significant difference was observed in comparisons of direct, indirect, and network meta-analysis (Figure S11). Funnel plots were used to evaluate publication bias, and the absence of marked asymmetry was taken to indicate no evidence of such bias (Figure 6).

Figure 6 Funnel plots for assessment of publication bias. The figure presents funnel plots used to visually assess the potential for publication bias in the network meta-analysis for the outcomes of (A) EFS, (B) OS, and (C) grade ≥3 adverse events. Each point represents an individual study. The x-axis shows the treatment effect estimate (log hazard ratio or log odds ratio), and the y-axis represents the precision of the estimate (standard error). In the absence of publication bias, studies are expected to be distributed symmetrically around the summary effect estimate (the vertical line). The observed symmetrical distribution of studies in these plots suggests a low likelihood of substantial publication bias affecting the analysis results. Adj Chemo, adjuvant chemotherapy; Adj ChemoIO, adjuvant chemoimmunotherapy; EFS, event-free survival; Neo Chemo, neoadjuvant chemotherapy; Neo ChemoIO, neoadjuvant chemoimmunotherapy; OS, overall survival; Peri ChemoIO, perioperative chemoimmunotherapy.

Sensitivity analysis

Four studies were identified as high-risk evaluated by the RoB2 tool, including Depierre’s study (35), ANITA, NATCH and Wei’s study (33,35-37). In the first sensitivity analysis, we excluded these high-risk studies from our analysis of the overall and subgroups. Following the concordant methodological process, the results were consistent with the original network meta-analysis (Figures S12,S13 and Tables S7-S9). In the overall cohort of patients with resectable NSCLC, perioperative and neoadjuvant chemoimmunotherapy demonstrated comparable EFS (HR 0.84, 95% CI: 0.60 to 1.17; P=0.31) and OS (HR 0.93, 95% CI: 0.59 to 1.47; P=0.75). However, perioperative chemoimmunotherapy was potentially to be more effective than the adjuvant chemoimmunotherapy [EFS (HR 0.76, 95% CI: 0.59 to 0.96; P=0.03), OS (HR 0.69, 95% CI: 0.56 to 0.94; P<0.001)]. Grade ≥3 adverse events occurred most frequently with adjuvant (45.69%) and perioperative (53.20%) chemoimmunotherapy. Among patients with stage IB–II disease, both perioperative chemoimmunotherapy (HR 0.68, 95% CI: 0.51 to 0.90; P=0.009) and adjuvant chemoimmunotherapy (HR 0.72, 95% CI: 0.60 to 0.87; P<0.001) significantly improved EFS compared to chemotherapy. No significant differences in EFS were observed among perioperative, adjuvant, and neoadjuvant chemoimmunotherapy. In contrast, for stage IIIA–IIIB NSCLC, perioperative chemoimmunotherapy (HR 0.54, 95% CI: 0.44 to 0.66; P<0.001) and neoadjuvant chemoimmunotherapy (HR 0.54, 95% CI: 0.32 to 0.91; P=0.02) provided superior outcomes than neoadjuvant chemotherapy. Perioperative chemoimmunotherapy (HR 0.28, 95% CI: 0.11 to 0.69; P=0.007) and neoadjuvant chemoimmunotherapy (HR 0.28, 95% CI: 0.10 to 0.78; P=0.01) were associated with significantly improved EFS compared to adjuvant chemoimmunotherapy. Notably, in PD-L1-negative patients, perioperative chemoimmunotherapy was the only strategy associated with a significant EFS benefit over chemotherapy (HR 0.74, 95% CI: 0.60 to 0.90; P=0.005).

In the second sensitivity analysis, we excluded the NEOTORCH trial which assigned three cycles of neoadjuvant chemotherapy and one cycle of adjuvant chemotherapy. The results remained consistent with the original network meta-analysis (Figures S14,S15 and Tables S10,S11). For the overall population, perioperative and neoadjuvant chemoimmunotherapy remained comparable for patients with resectable NSCLC in EFS (HR 0.88, 95% CI: 0.63 to 1.24; P=0.45) and OS (HR 0.95, 95% CI: 0.60 to 1.51; P=0.83). Perioperative chemoimmunotherapy was associated with longer EFS than the adjuvant chemoimmunotherapy (HR 0.79, 0.63 to 0.98; P=0.04). Perioperative (50.57%) chemoimmunotherapy and adjuvant (48.02%) were most likely to cause adverse events of grade 3 or higher. In stage IB–II NSCLC, both perioperative chemoimmunotherapy (HR 0.68, 95% CI: 0.49 to 0.96; P=0.02) and adjuvant chemoimmunotherapy (HR 0.73, 95% CI: 0.57 to 0.94; P=0.01) significantly improved EFS compared to chemotherapy, and three chemoimmunotherapy strategies still showed comparable EFS. In stage IIIA–IIIB NSCLC, perioperative chemoimmunotherapy (HR 0.58, 95% CI: 0.50 to 0.68; P<0.001) and neoadjuvant chemoimmunotherapy (HR 0.54, 95% CI: 0.37 to 0.69; P=0.001) appeared to result in longer EFS than neoadjuvant chemotherapy. Moreover, perioperative chemoimmunotherapy (HR 0.31, 95% CI: 0.14 to 0.69; P=0.004) and neoadjuvant (HR 0.34, 95% CI: 0.17 to 0.68; P=0.002) were highly associated with EFS improvement over adjuvant chemoimmunotherapy. For PD-L1-negative patients, perioperative chemoimmunotherapy was the only approach that significantly improved EFS versus chemotherapy (HR 0.74, 95% CI: 0.60 to 0.91; P=0.005).

Discontinuation rates and surgery

Treatment discontinuation and surgical delay are two noteworthy considerations in the context of adjuvant and neoadjuvant chemoimmunotherapy. In the adjuvant setting (Table S12), among 1,973 patients receiving adjuvant chemoimmunotherapy across six trials, approximately 37.98% of patients experienced treatment interruption during the adjuvant maintenance of ICIs, and 16.22% discontinued adjuvant ICIs due to adverse events. In the neoadjuvant setting (Table S13), 82.3% of patients completed neoadjuvant therapy, with 81.5% proceeding to definitive surgery and 74.0% achieving R0 resection.

The pooled analysis indicated inherent limitations of different chemoimmunotherapy regimens in clinical practice. To summarize, adjuvant chemoimmunotherapy may be associated with a risk of treatment discontinuation, while neoadjuvant chemoimmunotherapy may potentially lead to delays or cancellations of surgery due to the time required for its administration. These observations highlight the need for a balanced and individualized treatment approach.

Discussion

To date, ICIs have reshaped the treatment landscape for patients with resectable NSCLC. Perioperative, adjuvant and neoadjuvant chemoimmunotherapy all have demonstrated clinical benefits in this setting. However, whether clinical stage, PD-L1 expression and ECOG PS should influence treatment decision-making for patients with resectable NSCLC remains elusive. To address this issue, we conducted the first network meta-analysis to compare the efficacy and safety of different chemoimmunotherapy strategies across subgroups defined by clinical stage, PD-L1 expression and ECOG PS.

Our subgroup analyses demonstrated notable variations in optimal treatment strategies across patients with different clinical stage, PD-L1 expression and ECOG PS. Among stage IB-II patients, perioperative yielded comparable EFS benefits with adjuvant chemoimmunotherapy (HR 1.06, 95% CI: 0.64 to 1.77; P=0.82) and neoadjuvant chemoimmunotherapy (HR 0.78, 95% CI: 0.37 to 1.62; P=0.51). For patients with stage III NSCLC, perioperative and neoadjuvant chemoimmunotherapy provided significantly better outcomes versus adjuvant chemoimmunotherapy. Consistent with our findings, a phase II head-to-head RCT involving patients with resectable stage IIIB-IVC melanoma proved that perioperative pembrolizumab significantly prolonged EFS more than adjuvant pembrolizumab (72% versus 49% at 2 years, P=0.004) (38). Several explanations might account for the enhanced benefit of neoadjuvant ICIs in patients with stage III disease. Firstly, neoadjuvant treatment could reduce the size of the primary tumor and facilitate tumor downstaging, increasing the chance of complete resection for patients with relatively late-stage disease (39,40). Furthermore, patients have more intact vasculature surrounding the tumor bed before surgery, which may enhance the infiltration of anti-tumor immune cells and thereby theoretically optimize the efficacy of chemoimmunotherapy (41-43). Compared to stage I–II NSCLC, stage III disease is generally more aggressive and associated with a higher risk of recurrence, rendering early initiation of immunotherapy particularly beneficial for this patient population.

In terms of the PD-L1 expression subgroup analysis, three chemoimmunotherapy approaches showed superior EFS versus chemotherapy in PD-L1-positive patients. In contrast, only perioperative chemoimmunotherapy significantly improved EFS in PD-L1-negative patients (HR 0.74, 95% CI: 0.60 to 0.91; P=0.005), suggesting PD-L1 expression as a potential biomarker to guide the choice of chemoimmunotherapy approaches. These findings are consistent with individual participant data (IPD) meta-analysis from the CheckMate-816 and CheckMate-77T trials, while both treatment strategies yielded comparable efficacy in patients with PD-L1 expression ≥1%, perioperative chemoimmunotherapy was associated with improved EFS in the PD-L1 <1% subgroup (20). Collectively, these results suggest that patients with low PD-L1 expression (1%) may require more intensified immunotherapeutic approaches to derive more clinical benefit. Additionally, we also found that perioperative chemoimmunotherapy was the only approach that led to significant improvement in EFS for patients with ECOG PS ≥1. Nevertheless, as the statistical power of this subgroup analysis may be limited and influenced by other confounding factors, further prospective research is warranted to confirm these observations.

From a safety perspective, perioperative chemoimmunotherapy was associated with a significantly higher incidence of grade ≥3 TRAEs compared to neoadjuvant chemoimmunotherapy (OR 1.63, 95% CI: 1.02 to 2.59; P=0.041). In contrast, the incidence of grade ≥3 TRAEs was comparable between perioperative and adjuvant chemoimmunotherapy (OR 1.03, 95% CI: 0.55 to 1.94; P=0.93). Given that the perioperative regimen includes both neoadjuvant and adjuvant phases, we speculate that the increased toxicity observed in the perioperative setting may be primarily attributable to the adjuvant component. Consistent with our analysis, a recent meta-analysis including 28 clinical trials also revealed that adjuvant treatment with ICIs significantly increased treatment-related death (OR 4.02, 95% CI: 1.04 to 15.63; P=0.04) and adverse events of grade 3 to 4 (5.31, 3.08 to 9.15; P<0.001), whereas neoadjuvant ICIs were not associated with significantly increased toxicity (44,45). Based on both efficacy and toxicity analyses, we suggest that only PD-L1-negative patients require the perioperative chemoimmunotherapy approach. For PD-L1-positive patients, neoadjuvant chemoimmunotherapy may be more favorable in order to avoid unnecessary toxicity (14,46). Treatment for this group may be further refined by incorporating minimal residual disease (MRD) detection (47-51). Recent studies have shown that the detection of circulating tumor DNA (ctDNA) after surgery is associated with shortened recurrence-free survival in lung cancer and other solid tumors (51-53). The LUNGCA trial found that adjuvant ICIs improved recurrence-free survival in MRD-positive resectable NSCLC patients, but not in MRD-negative patients (49). Therefore, for patients with positive ctDNA detected after surgery, adjuvant ICIs might also be required even for those with positive PD-L1 expression. However, the benefit of prolonged adjuvant ICIs in resectable NSCLC with PD-L1 and MRD-positive needs further investigation in RCTs.

There are several limitations in this meta-analysis. First, only one reported RCT, the CheckMate816 trial, directly compared neoadjuvant chemoimmunotherapy with chemotherapy, limiting the available data on this specific comparison. In addition, all patients receiving adjuvant chemoimmunotherapy underwent surgery, but a small fraction of the patients receiving neoadjuvant and perioperative treatment did not, potentially introducing a slight prognostic disadvantage and contributing to population heterogeneity. Furthermore, OS data were not yet fully available, further analysis is warranted to assess long-term benefit of chemoimmunotherapy in resectable NSCLC.

Conclusions

Perioperative, adjuvant and neoadjuvant chemoimmunotherapy were all suitable for patients with stage IB–II NSCLC, while perioperative and neoadjuvant chemoimmunotherapy may represent more favorable options for stage III disease. In patients with PD-L1-negative expression or ECOG PS ≥1, only perioperative chemoimmunotherapy was associated with improved EFS versus chemotherapy. Our findings might benefit future individualized immunochemotherapy for resectable NSCLC.

Acknowledgments

None.

Footnote

Reporting Checklist: The authors have completed the PRISMA NMA reporting checklist. Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-426/rc

Peer Review File: Available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-426/prf

Funding: This study was supported by the Noncommunicable Chronic Diseases-National Science and Technology Major Project (No. 2024ZD0519700), and the National Natural Science Foundation of China (Nos. 82241232 and 82272789).

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tlcr.amegroups.com/article/view/10.21037/tlcr-2025-426/coif). The authors have no conflicts of interest to declare.

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.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

1. Sung H, Ferlay J, Siegel RL, et al. Global Cancer Statistics 2020: GLOBOCAN Estimates of Incidence and Mortality Worldwide for 36 Cancers in 185 Countries. CA Cancer J Clin 2021;71:209-49. [Crossref] [PubMed]
2. Goldstraw P, Chansky K, Crowley J, et al. The IASLC Lung Cancer Staging Project: Proposals for Revision of the TNM Stage Groupings in the Forthcoming (Eighth) Edition of the TNM Classification for Lung Cancer. J Thorac Oncol 2016;11:39-51. [Crossref] [PubMed]
3. Zeng H, Ran X, An L, et al. Disparities in stage at diagnosis for five common cancers in China: a multicentre, hospital-based, observational study. Lancet Public Health 2021;6:e877-87. [Crossref] [PubMed]
4. Araghi M, Fidler-Benaoudia M, Arnold M, et al. International differences in lung cancer survival by sex, histological type and stage at diagnosis: an ICBP SURVMARK-2 Study. Thorax 2022;77:378-90. [Crossref] [PubMed]
5. Preoperative chemotherapy for non-small-cell lung cancer: a systematic review and meta-analysis of individual participant data. Lancet 2014;383:1561-71. [Crossref] [PubMed]
6. Provencio M, Nadal E, González-Larriba JL, et al. Perioperative Nivolumab and Chemotherapy in Stage III Non-Small-Cell Lung Cancer. N Engl J Med 2023;389:504-13. [Crossref] [PubMed]
7. Yue D, Wang W, Liu H, et al. VP1-2024: RATIONALE-315: Event-free survival (EFS) and overall survival (OS) of neoadjuvant tislelizumab (TIS) plus chemotherapy (CT) with adjuvant TIS in resectable non-small cell lung cancer (NSCLC). Ann Oncol 2024;35:332-3.
8. Cascone T, Awad MM, Spicer JD, et al. Perioperative Nivolumab in Resectable Lung Cancer. N Engl J Med 2024;390:1756-69. [Crossref] [PubMed]
9. Lu S, Zhang W, Wu L, et al. Perioperative Toripalimab Plus Chemotherapy for Patients With Resectable Non-Small Cell Lung Cancer: The Neotorch Randomized Clinical Trial. JAMA 2024;331:201-11. [Crossref] [PubMed]
10. Heymach JV, Harpole D, Mitsudomi T, et al. Perioperative Durvalumab for Resectable Non-Small-Cell Lung Cancer. N Engl J Med 2023;389:1672-84. [Crossref] [PubMed]
11. Spicer JD, Garassino MC, Wakelee H, et al. Neoadjuvant pembrolizumab plus chemotherapy followed by adjuvant pembrolizumab compared with neoadjuvant chemotherapy alone in patients with early-stage non-small-cell lung cancer (KEYNOTE-671): a randomised, double-blind, placebo-controlled, phase 3 trial. Lancet 2024;404:1240-52. [Crossref] [PubMed]
12. O'Brien M, Paz-Ares L, Marreaud S, et al. Pembrolizumab versus placebo as adjuvant therapy for completely resected stage IB-IIIA non-small-cell lung cancer (PEARLS/KEYNOTE-091): an interim analysis of a randomised, triple-blind, phase 3 trial. Lancet Oncol 2022;23:1274-86. [Crossref] [PubMed]
13. Wakelee HA, Altorki NK, Zhou C, et al. IMpower010: Final disease-free survival (DFS) and second overall survival (OS) interim results after ≥5 years of follow up of a phase III study of adjuvant atezolizumab vs best supportive care in resected stage IB-IIIA non-small cell lung cancer (NSCLC). J Clin Oncol 2024;42:LBA8035.
14. Spicer J, Girard N, Provencio M, et al. Neoadjuvant nivolumab (NIVO) + chemotherapy (chemo) vs chemo in patients (pts) with resectable NSCLC: 4-year update from CheckMate 816. J Clin Oncol 2024;42:LBA8010.
15. Wu Y, Verma V, Gay CM, et al. Neoadjuvant immunotherapy for advanced, resectable non-small cell lung cancer: A systematic review and meta-analysis. Cancer 2023;129:1969-85. [Crossref] [PubMed]
16. Jia XH, Xu H, Geng LY, et al. Efficacy and safety of neoadjuvant immunotherapy in resectable nonsmall cell lung cancer: A meta-analysis. Lung Cancer 2020;147:143-53. [Crossref] [PubMed]
17. Liu W, Zhang T, Zhang Q, et al. A systematic review and meta-analysis of neoadjuvant chemoimmunotherapy in stage III non-small cell lung cancer. BMC Pulm Med 2022;22:490. [Crossref] [PubMed]
18. Pasqualotto E, Moraes FCA, Chavez MP, et al. PD-1/PD-L1 Inhibitors plus Chemotherapy Versus Chemotherapy Alone for Resectable Non-Small Cell Lung Cancer: A Systematic Review and Meta-Analysis of Randomized Controlled Trials. Cancers (Basel) 2023;15:5143. [Crossref] [PubMed]
19. Zhou Y, Li A, Yu H, et al. Neoadjuvant-Adjuvant vs Neoadjuvant-Only PD-1 and PD-L1 Inhibitors for Patients With Resectable NSCLC: An Indirect Meta-Analysis. JAMA Netw Open 2024;7:e241285. [Crossref] [PubMed]
20. Forde PM, Peters S, Donington J, et al., editors. Perioperative vs Neoadjuvant Nivolumab for Resectable NSCLC: Patient-Level Data Analysis of CheckMate 77T vs CheckMate 816. The IASLC World Conference on Lung Cancer; 2024 Sunday, 8 September 2024; San Diego, CA, USA. 2024. doi: 10.1016/j.jtho.2024.09.014.
21. Banna GL, Hassan MA, Signori A, et al. Neoadjuvant Chemo-Immunotherapy for Early-Stage Non-Small Cell Lung Cancer: A Systematic Review and Meta-Analysis. JAMA Netw Open 2024;7:e246837. [Crossref] [PubMed]
22. Zhang SL, Tian Y, Yu J, et al. Is neoadjuvant immunotherapy necessary in patients with programmed death ligand 1 expression-negative resectable non-small cell lung cancer? A systematic review and meta-analysis. Lung Cancer 2024;191:107799. [Crossref] [PubMed]
23. Ettinger DS, Wood DE, Aisner DL, et al. NCCN Guidelines® Insights: Non-Small Cell Lung Cancer, Version 2.2023. J Natl Compr Canc Netw 2023;21:340-50. [Crossref] [PubMed]
24. Cochrane Handbook for Systematic Reviews of Interventions. Wiley; 2019/9/20.
25. Higgins JP, Altman DG, Gøtzsche PC, et al. The Cochrane Collaboration's tool for assessing risk of bias in randomised trials. BMJ 2011;343:d5928. [Crossref] [PubMed]
26. Chaimani A, Higgins JP, Mavridis D, et al. Graphical tools for network meta-analysis in STATA. PLoS One 2013;8:e76654. [Crossref] [PubMed]
27. Higgins JP, Thompson SG, Deeks JJ, et al. Measuring inconsistency in meta-analyses. BMJ 2003;327:557-60. [Crossref] [PubMed]
28. Plummer M, others, editors. JAGS: A program for analysis of Bayesian graphical models using Gibbs sampling 2003. Available online: http://www.ci.tuwien.ac.at/Conferences/DSC-2003/
29. Valkenhoef GV, Kuiper J. gemtc: Network meta-analysis using bayesian methods. R Packag version 08-2.
30. Spiegelhalter DJ, Best NG, Carlin BP, et al. Bayesian Measures of Model Complexity and Fit. J R Stat Soc Series B Stat Methodol 2002;64:583-639.
31. Salanti G, Ades AE, Ioannidis JP. Graphical methods and numerical summaries for presenting results from multiple-treatment meta-analysis: an overview and tutorial. J Clin Epidemiol 2011;64:163-71. [Crossref] [PubMed]
32. Winton T, Livingston R, Johnson DNational Cancer Institute of Canada Clinical Trials Group, et al. National Cancer Institute of the United States Intergroup JBR.10 Trial Investigators. Vinorelbine plus cisplatin vs. observation in resected non-small-cell lung cancer. N Engl J Med 2005;352:2589-97. [Crossref] [PubMed]
33. Ou W, Sun HB, Ye X, et al. Adjuvant carboplatin-based chemotherapy in resected stage IIIA-N2 non-small cell lung cancer. J Thorac Oncol 2010;5:1033-41. [Crossref] [PubMed]
34. Rosell R, Gómez-Codina J, Camps C, et al. Preresectional chemotherapy in stage IIIA non-small-cell lung cancer: a 7-year assessment of a randomized controlled trial. Lung Cancer 1999;26:7-14. [Crossref] [PubMed]
35. Depierre A, Milleron B, Moro-Sibilot DFrench Thoracic Cooperative Group, et al. Preoperative chemotherapy followed by surgery compared with primary surgery in resectable stage I (except T1N0), II, and IIIa non-small-cell lung cancer. J Clin Oncol 2002;20:247-53. [Crossref] [PubMed]
36. Douillard JY, Rosell R, De Lena M, et al. Adjuvant vinorelbine plus cisplatin versus observation in patients with completely resected stage IB-IIIA non-small-cell lung cancer (Adjuvant Navelbine International Trialist Association [ANITA]): a randomised controlled trial. Lancet Oncol 2006;7:719-27.
37. Felip E, Rosell R, Maestre JASpanish Lung Cancer Group, et al. Preoperative chemotherapy plus surgery versus surgery plus adjuvant chemotherapy versus surgery alone in early-stage non-small-cell lung cancer. J Clin Oncol 2010;28:3138-45. [Crossref] [PubMed]
38. Patel SP, Othus M, Chen Y, et al. Neoadjuvant-Adjuvant or Adjuvant-Only Pembrolizumab in Advanced Melanoma. N Engl J Med 2023;388:813-23. [Crossref] [PubMed]
39. Mountzios G, Remon J, Hendriks LEL, et al. Immune-checkpoint inhibition for resectable non-small-cell lung cancer - opportunities and challenges. Nat Rev Clin Oncol 2023;20:664-77. [Crossref] [PubMed]
40. Kang J, Zhang C, Zhong WZ. Neoadjuvant immunotherapy for non-small cell lung cancer: State of the art. Cancer Commun (Lond) 2021;41:287-302. [Crossref] [PubMed]
41. Thommen DS, Koelzer VH, Herzig P, et al. A transcriptionally and functionally distinct PD-1(+) CD8(+) T cell pool with predictive potential in non-small-cell lung cancer treated with PD-1 blockade. Nat Med 2018;24:994-1004. [Crossref] [PubMed]
42. Sade-Feldman M, Yizhak K, Bjorgaard SL, et al. Defining T Cell States Associated with Response to Checkpoint Immunotherapy in Melanoma. Cell 2018;175:998-1013.e20. [Crossref] [PubMed]
43. Zhang Y, Zhang Z. The history and advances in cancer immunotherapy: understanding the characteristics of tumor-infiltrating immune cells and their therapeutic implications. Cell Mol Immunol 2020;17:807-21. [Crossref] [PubMed]
44. Delyon J, Michielin O. Adjuvant or neoadjuvant treatment with immune checkpoint inhibitors: re-assessing the risk-benefit ratio. Lancet Oncol 2024;25:3-5.
45. Fujiwara Y, Horita N, Adib E, et al. Treatment-related adverse events, including fatal toxicities, in patients with solid tumours receiving neoadjuvant and adjuvant immune checkpoint blockade: a systematic review and meta-analysis of randomised controlled trials. Lancet Oncol 2024;25:62-75. [Crossref] [PubMed]
46. White B, Harris M, Villacorta R, et al. Cost-effectiveness of perioperative nivolumab + neoadjuvant platinum doublet chemotherapy as treatment for resectable non-small cell lung cancer in the United States. J Med Econ 2025;28:625-37. [Crossref] [PubMed]
47. Abbosh C, Birkbak NJ, Wilson GA, et al. Phylogenetic ctDNA analysis depicts early-stage lung cancer evolution. Nature 2017;545:446-51. [Crossref] [PubMed]
48. Chaudhuri AA, Chabon JJ, Lovejoy AF, et al. Early Detection of Molecular Residual Disease in Localized Lung Cancer by Circulating Tumor DNA Profiling. Cancer Discov 2017;7:1394-403. [Crossref] [PubMed]
49. Xia L, Mei J, Kang R, et al. Perioperative ctDNA-Based Molecular Residual Disease Detection for Non-Small Cell Lung Cancer: A Prospective Multicenter Cohort Study (LUNGCA-1). Clin Cancer Res 2022;28:3308-17. [Crossref] [PubMed]
50. Zhou C, Das Thakur M, Srivastava MK, et al. IMpower010: Biomarkers of disease-free survival (DFS) in a phase III study of atezolizumab (atezo) vs best supportive care (BSC) after adjuvant chemotherapy in stage IB-IIIA NSCLC. Ann Oncol 2021;32:S1374.
51. Gale D, Heider K, Ruiz-Valdepenas A, et al. Residual ctDNA after treatment predicts early relapse in patients with early-stage non-small cell lung cancer. Ann Oncol 2022;33:500-10. [Crossref] [PubMed]
52. Zhang JT, Liu SY, Gao W, et al. Longitudinal Undetectable Molecular Residual Disease Defines Potentially Cured Population in Localized Non-Small Cell Lung Cancer. Cancer Discov 2022;12:1690-701. [Crossref] [PubMed]
53. Chen K, Zhao H, Shi Y, et al. Perioperative Dynamic Changes in Circulating Tumor DNA in Patients with Lung Cancer (DYNAMIC). Clin Cancer Res 2019;25:7058-67. [Crossref] [PubMed]