Optimal duration of adjuvant immunotherapy after neoadjuvant chemoimmunotherapy in resectable non-small cell lung cancer: a systematic review and meta-analysis.

Introduction
Non-small cell lung cancer (NSCLC) remains a potentially curable disease for a subset of patients who present with localized tumors amenable to surgical resection. However, long-term disease control after surgery alone is frequently compromised by postoperative recurrence, underscoring the need for effective perioperative systematic strategies (1,2). Although platinum-based chemotherapy has been widely incorporated into neoadjuvant or adjuvant settings, its contribution to long-term survival improvement has been limited, translating into only marginal gains (5–6%) in 5-year overall survival (OS) (3,4).
The advent of immune checkpoint inhibitors (ICIs) has reshaped therapeutic paradigms. In the adjuvant setting, landmark trials such as IMpower010 and KEYNOTE-091 demonstrated that adding ICIs following platinum-based chemotherapy either brought numeric benefits to disease-free survival (DFS), or significantly prolonged the DFS in stage II–IIIA NSCLC [hazard ratio (HR) =0.66–0.76; P<0.01] (5-8). Concurrently, neoadjuvant ICI-chemotherapy combinations showed groundbreaking efficacy. The CheckMate 816 trial first established the efficacy of neoadjuvant nivolumab plus chemotherapy, demonstrating significant improvements in pathologic complete response rates (24% vs. 2%), event-free survival (EFS) [HR =0.68, 95% confidence interval (CI): 0.51–0.91] and OS (HR =0.72, 95% CI: 0.523–0.998) in stage IB–IIIA resectable disease, thereby setting a new standard of care (9,10).
Building on these findings, the strategy of perioperative immunotherapy (combining neoadjuvant and adjuvant ICIs) has been evaluated in several phase 3 trials, such as AEGEAN (11), Neotorch (12), KEYNOTE-671 (13,14), CheckMate 77T (15), and RATIONALE 315 (16,17). Furthermore, recent recommendations from the International Association for the Study of Lung Cancer (IASLC) suggest that adjuvant immunotherapy can be considered following neoadjuvant chemoimmunotherapy (18). The National Comprehensive Cancer Network (NCCN) guidelines also recommend the use of adjuvant ICIs following neoadjuvant ICIs plus chemotherapy as a Category 1 standard-of-care regimen.
A critical issue that remains unresolved in clinical practice is the optimal duration of adjuvant therapy. Currently, within the NCCN-recommended adjuvant ICIs following neoadjuvant therapy, there are two common treatment durations: 12–13 and 6 cycles. However, only 24% to 44% of patients are able to complete the full 12-cycle regimen, despite a median progression-free survival (PFS) exceeding 2 years (11-13,15,16). This raises the important question: is it necessary to maintain 12–13 cycles of treatment, and does more cycles offer superior efficacy compared to a fewer one?
Importantly, no randomized trial to date has been specifically designed to compare different durations of adjuvant immunotherapy following neoadjuvant chemoimmunotherapy and surgery. Given the absence of direct comparative evidence and the limited feasibility of conducting such trials in the near future, we performed a systematic review and meta-analysis using indirect comparison methods to evaluate whether extended adjuvant immunotherapy (12/13 cycles) provides additional survival benefit over shorter regimens (6/8 cycles) in patients with resectable NSCLC treated with perioperative chemoimmunotherapy. We present this article in accordance with the PRISMA reporting checklist (available at https://jtd.amegroups.com/article/view/10.21037/jtd-2025-1389/rc).

Methods
The study was registered on the International Platform of Registered Systematic Review and Meta-analysis Protocols (CRD42025617836). No separate protocol was prepared beyond the PROSPERO (International Prospective Register of Systematic Reviews) registration. No amendments were made to the original registration after initial submission. No patients or members of the public were involved in the design, conduct and reporting of this research.

Data sources and search strategy
We conducted a comprehensive literature search to identify randomized controlled trials (RCTs) evaluating perioperative immunotherapy in resectable NSCLC. Searches were performed in PubMed, Embase, and the Cochrane Library, complemented by manual screening of abstracts from major international oncology conferences. All databases were searched from inception through July 31, 2024.
The search strategy was designed to capture trials incorporating both neoadjuvant and adjuvant components of programmed cell death 1/programmed death-ligand 1 (PD-1/PD-L1) inhibitor therapy and included terms related to immune checkpoint inhibition, perioperative or neoadjuvant treatment, adjuvant therapy, NSCLC, and randomized trial design. The complete electronic search strategy is provided in Appendix 1.

Inclusion criteria
Studies were included if they met the following criteria: (I) involved patients with resectable NSCLC; (II) administered PD-1/PD-L1 inhibitors as adjuvant therapy following surgery; (III) included a neoadjuvant phase combining PD-1/PD-L1 inhibitor with platinum-based chemotherapy; (IV) compared an investigational arm (ICI + chemotherapy) against a control arm receiving chemotherapy alone; (V) reported outcomes for EFS or PFS; and (VI) phase 2/3 RCTs published in peer-reviewed journals.

Data extraction
Z.C. and M.S.C.L. independently extracted relevant data and resolved any discrepancies through consensus. The extracted data included EFS, OS, and subgroup-specific outcomes. Additional details were recorded using a predefined information sheet. The methodological quality of each included study was assessed using the Cochrane Risk of Bias Tool (19). The data extracted can be accessed upon reasonable requests.

Statistical analysis
Three treatment strategies were prespecified for analysis: neoadjuvant chemoimmunotherapy followed by adjuvant PD-1/PD-L1 inhibitor therapy for 12/13 cycles, neoadjuvant chemoimmunotherapy followed by adjuvant therapy for 6/8 cycles, and neoadjuvant chemotherapy alone. EFS was the primary endpoint.
For each adjuvant strategy, HRs for EFS were pooled against the chemotherapy control using inverse-variance weighting. Heterogeneity across studies was quantified using Cochran’s Q and I2 statistics, which informed the choice between fixed- and random-effects models (20). Funnel plots and Egger’s test were subsequently applied to evaluate potential publication bias.
Indirect comparisons between the 12/13-cycle and 6/8-cycle adjuvant regimens were conducted using a common comparator approach, with neoadjuvant chemotherapy serving as the reference group. Log-transformed HRs and their standard errors (SEs) were used to estimate the relative treatment effect. The formula used was (21): log HRAB = log HRAC − log HRBC, and its SE for the log HR was SE (log HRAB) = [SE (log HRAC)2 + SE (log HRBC)2]1/2.
Based on the primary analysis, the study further conducted prespecified subgroup and sensitivity evaluations. Each included trial provided subgroup-level data. The analysis pooled these data using direct meta-analytic methods. The same indirect comparison framework used in the overall analysis was also applied to the subgroup analyses. The study additionally explored variability in treatment effects across predefined subgroups. Meta-regression analysis was used. The models were fitted using the restricted maximum likelihood method. These analyses together allowed a structured assessment of heterogeneity and supported the robustness of the main findings.
P values for interaction terms in subgroups were determined from this analysis, using the meta and gemtc packages in R (version 4.2.1). Statistical significance was set at P<0.05 for all tests, with two-tailed analysis used throughout.

Results
A total of six trials [AEGEAN (11), Neotorch (12), KEYNOTE-671 (13,14), CheckMate 77T (15), RATIONALE 315 (16,17) and NADIM-II (22)], encompassing 3,003 patients, were included in the analysis (Figure 1).

Trial characteristics and outcomes
All included studies were rigorously designed RCTs, demonstrating robust methodological quality. Random sequence generation were appropriately implemented across all trials, and incomplete outcome data were adequately addressed. No selective reporting biases were identified, though one trial (RATIONALE-315) utilized conference data, which may have limited the transparency of some details. Overall, the studies exhibited high methodological quality (Table S1).
The characteristics and outcomes of the trials are summarized in Table 1. All studies compared the combination of neoadjuvant immunotherapy and chemotherapy followed by adjuvant immunotherapy with neoadjuvant chemotherapy alone. Patients enrolled in these studies were diagnosed with stage II–III or III resectable NSCLC, as per the American Joint Committee on Cancer (AJCC) Staging Manual 8th edition. All trials involved 3 to 4 cycles of neoadjuvant platinum-doublet chemotherapy in combination with either a PD-1/PD-L1 inhibitor or placebo. Four trials (KEYNOTE-671, Neotorch, AEGEAN, and CheckMate 77T) utilized 12/13 cycles of adjuvant immunotherapy, while the remaining two trials (NADIM-II and RATIONALE-315) employed 6/8 cycles.
In the four studies with 12/13-cycle regimens, 77.6–82.1% of patients underwent definitive surgery, while in the studies with 6/8 cycles of adjuvant therapy, 84.1–93% of patients underwent definitive surgery. For the 12/13-cycle groups, 62% to 73.2% of patients initiated adjuvant immunotherapy, with a median of 12/13 cycles of PD-1 monoclonal antibody treatment. In the 6/8-cycle regimens, 74% of patients began adjuvant immunotherapy, with a median duration of 6/8 cycles.
The median follow-up period ranged from 18.3 to 36.6 months. Except for the KEYNOTE-671 study, which reported a median EFS of 47.2 months, no study reported a definitive median EFS for the experimental groups. In available control arms, the median EFS ranged from 15.1 to 30.0 months. Across all studies, the 2-year EFS rates for both the 12/13-cycle (62.4–65.0%) and 6/8-cycle (67.2–68.3%) adjuvant immunotherapy groups was similar, with EFS rates significantly higher in the experimental arms compared to the control. Additional characteristics of the enrolled trials are available in Table S2.

Direct meta-analysis for EFS
The analysis compared EFS between perioperative chemoimmunotherapy regimens and neoadjuvant chemotherapy alone. Specifically, pooled HRs for EFS were estimated for patients receiving neoadjuvant chemoimmunotherapy followed by adjuvant PD-1/PD-L1 inhibitor therapy for either 12/13 cycles or 6/8 cycles. These comparisons were performed for the overall population as well as for predefined subgroups (Figures S1,S2). Given the low level of between-study heterogeneity, fixed-effects models were used. The corresponding results are presented in Figure 2.
Chemoimmunotherapy resulted in superior EFS outcomes compared with chemotherapy alone in both the 12/13-cycle and 6/8-cycle adjuvant groups [HR, 0.58 (95% CI: 0.50–0.66), I2=47.3%; HR, 0.55 (95% CI: 0.41–0.73), I2=0%] (Figures S1,S2). Funnel plots showed no asymmetry (Egger’s test P=0.19 for all six studies, and P=0.11 for 12/13-cycle studies, Figure S3). This benefit in EFS was consistent across subgroups, except for patients with a history of never smoking in the 12/13-cycle adjuvant immunotherapy group (Figure S1). For the 6/8-cycle regimen, no significant advantage was observed in the PD-L1 <1% subgroups (Figure S2).

Indirect meta-analysis for EFS
Figure 2 summarizes the indirect comparison between adjuvant immunotherapy delivered for 12/13 cycles and 6/8 cycles. The analysis showed no meaningful difference in EFS between the two strategies (HR, 1.05; 95% CI: 0.77–1.45). This pattern remained stable across most prespecified subgroups. These subgroups included age, sex, Eastern Cooperative Oncology Group (ECOG) performance status, disease stage, histologic characteristics, smoking status, and pathologic response. In the PD-L1 subgroups, both the <1% and ≥1% groups showed comparable efficacy between the two regimens. However, when subdividing the ≥1% group further into ≥50%, a notable difference emerged (P for interaction =0.02), indicating improved EFS with the longer adjuvant treatment for patients with PD-L1 expression ≥50% (Figure 2).
Considering that the RATIONALE-315 study, although using 8 cycles of adjuvant treatment, actually administered tislelizumab every 6 weeks, resulting in a longer maintenance duration of approximately 1 year, we combined the RATIONALE-315 study with the 12/13 cycle studies and referred to this group as the “1-year group” for comparison with the 6-cycle group (NADIM II study, half-a-year group). The results showed that, in the comparable subgroups, this classification method still did not reveal any differences in EFS [HR, 1.23 (95% CI: 0.65–2.34); Figure S4].
We further conducted targeted sensitivity analyses based on study-level covariates including definitive surgery rate, adjuvant initiation rate, and follow-up duration (Table S3). When restricting analysis to studies with definitive surgery rates >70%, the HR for EFS was 1.07 (95% CI: 0.73–1.54). Similarly, when including only studies with adjuvant initiation rates >70%, the HR was 0.95 (95% CI: 0.67–1.34). Finally, limiting analysis to studies with median follow-up exceeding 24 months yielded an HR of 1.30 (95% CI: 0.68–2.48).

Meta-analysis results for OS
Regarding OS, both extended (12/13 cycles) and shortened (6/8 cycles) adjuvant immunotherapy regimens were associated with a significantly reduced risk of death compared with chemotherapy alone. The HR for 12/13 cycles was 0.78 (95% CI: 0.66–0.92), while that for 6/8 cycles was 0.57 (95% CI: 0.38–0.85). Indirect comparison between the two adjuvant immunotherapy regimens revealed no statistically significant difference in OS [HR, 1.37 (95% CI: 0.89–2.12)].

Discussion
The findings of this meta-analysis suggest that extending adjuvant PD-1 inhibitor therapy from 6–8 cycles to 12–13 cycles does not significantly improve EFS in patients with resectable NSCLC who have received neoadjuvant chemoimmunotherapy and surgery. A potential trend toward improved EFS with increased cycles of adjuvant immunotherapy was observed in the PD-L1 expression ≥50% subgroup. However, insufficient data, including the limited HR values based on pathological response status, hindered further validation and limited subgroup analysis. Nonetheless, results imply that fewer adjuvant cycles [6–8] may suffice for most patients, whereas extended cycles might benefit PD-L1-high populations, pending prospective validation.
Perioperative immunotherapy, neoadjuvant chemoimmunotherapy followed by adjuvant immunotherapy, has emerged as a standard of care for resectable NSCLC (NCCN guidelines v2024). This paradigm, supported by landmark trials such as AEGEAN and KEYNOTE-671, has demonstrated superior outcomes compared to neoadjuvant chemotherapy alone, primarily through enhanced pathologic responses and delayed recurrence.
However, in current clinical practice, numerous clinical questions remain urgently awaiting answers. These include patient selection for adjuvant immunotherapy, whether patients who have achieved a pathological complete response (pCR) still require continued adjuvant treatment, and the choices regarding the adjuvant treatment regimen and its duration. Evidence from the CheckMate 816 trial indicates that even among patients who attained pCR after nivolumab-plus-chemotherapy, 7% experienced disease recurrence within 5 years, suggesting that pCR at the primary site does not guarantee eradication of distant micrometastases and that some patients remain at risk of recurrence. Thus, adjuvant therapy may still hold clinical value in mitigating residual risk.
As the “neoadjuvant chemoimmunotherapy + adjuvant immunotherapy” strategy gains broader clinical adoption, determining the optimal duration of adjuvant therapy has emerged as a pressing practical issue. Zhou et al. reported no significant survival benefit from adding adjuvant PD-1 inhibitors after neoadjuvant chemoimmunotherapy in their meta-analysis (23), this conclusion was limited by heterogeneity in trial designs, as only one included study focusing on pure neoadjuvant therapy, while four studies included both neoadjuvant and adjuvant treatment. Our analysis advances this discourse by specifically evaluating duration effects within perioperative regimens, controlling for prior neoadjuvant exposure.
Importantly, the efficacy of adjuvant immunotherapy should not be conflated with its necessity in perioperative strategies. Trials such as the IMpower010 and PEARLS/KEYNOTE-091 trials have demonstrated the efficacy of adjuvant-only immunotherapy in patients with completely resected NSCLC (5,6,8). However, when integrated into a perioperative approach, where patients already receive neoadjuvant PD-1 inhibitors, the incremental value of prolonged adjuvant cycles must be re-evaluated. Our findings challenge the default use of 12–13 cycles by demonstrating comparable efficacy with 6–8 cycles regimens in the overall population.
Although our results suggest that fewer cycles may optimize therapeutic value without compromising outcomes, there remains a critical need to identify patient populations that may benefit from more cycles of adjuvant immunotherapy. To achieve this, high-dimensional multi-omics studies within the neoadjuvant treatment framework will be necessary. In this study, we have addressed this issue preliminarily through subgroup analysis.
Our indirect comparison suggested a potential differential benefit of more cycles of adjuvant immunotherapy in patients with PD-L1 expression ≥50% (P for interaction =0.019). However, this observation was driven predominantly by the RATIONALE-315 trial, the sole study in the 6/8-cycle arm which reported inferior efficacy for the 8-cycle adjuvant treatment versus chemotherapy alone in this subgroup [HR, 0.71 (95% CI: 0.38–1.34)]. The trend aligns with biological plausibility: PD-L1-high tumors often exhibit an inflamed tumor microenvironment enriched with cytotoxic T cells and interferon-γ signaling, which may enhance sensitivity to prolonged immune checkpoint inhibition by sustaining antigen exposure and immune activation, thus suppressing residual micrometastases. This aligns with findings from the IMpower010 trial, where PD-L1 ≥50% patients derived significant DFS and OS benefits from adjuvant atezolizumab, whereas those with PD-L1 <1% showed no improvement (5,6). Similarly, the IASLC guidelines conditionally recommend adjuvant immunotherapy for resected NSCLC with PD-L1 ≥50% (18), reflecting accumulating (yet limited) evidence for biomarker-driven treatment strategies.
Within the non-pCR population, the comparison between the 12–13-cycle and 6–8-cycle adjuvant immunotherapy groups did not demonstrate a clear difference in EFS, and the estimated treatment effect was limited. In contrast, the number of patients achieving a pCR was insufficient to support a reliable subgroup comparison. These findings indicate persistent uncertainty regarding the optimal post-operative management of patients who do not achieve pCR after neoadjuvant chemoimmunotherapy. Accordingly, prospective clinical trials are warranted to directly evaluate continued adjuvant immunotherapy versus observation in this specific patient population (24,25).

Limitations
There are several limitations in this study. First, the conclusions rely on indirect comparisons between trial arms rather than direct randomized evidence, which may increase susceptibility to residual confounding from differences in trial designs, patient characteristics, and treatment protocols. Second, survival data remain immature, with median follow-up durations ranging from 18.3 to 36.6 months and only one trial reporting median EFS. Third, clinical and methodological heterogeneities across studies, including variable surgical resection rates (77.6–93%) and differential adjuvant treatment initiation rates (62–74%), may compromise comparability. Particularly, the RATIONALE-315 trial utilized 6-week tislelizumab dosing intervals rather than standard 3-week cycles, creating challenges in interpreting cycle-based duration comparisons. Fourth, the sample sizes in certain subgroups were small (e.g., PD-L1 ≥50%), potentially underpowering statistical comparisons and increasing risks of type II errors in detecting clinically meaningful differences.

Conclusions
To our knowledge, our meta-analysis is the first to compare outcomes between 12–13 cycles and 6–8 cycles of adjuvant anti-PD-1 therapy in resectable NSCLC patients receiving perioperative chemoimmunotherapy. These findings indicate no significant improvement in EFS with extended adjuvant immunotherapy, suggesting that regimens with fewer cycles (6–8 cycles) may achieve comparable efficacy while alleviating the financial and psychological burden on patients. While the data suggesting that the PD-L1 expression ≥50% subgroup might benefit from longer therapy, this signal requires validation in prospective biomarker-stratified trials. These results highlight the need to optimize adjuvant duration based on individual risk profiles rather than fixed protocols. Future studies should prioritize head-to-head comparisons of treatment durations and integrate biomarkers to guide personalized strategies.

Supplementary
The article’s supplementary files as