Circulating tumor DNA (ctDNA) has emerged as a robust biomarker for tumor burden and survival outcome prediction, but its role in neoadjuvant immunotherapy for non-small cell lung cancer (NSCLC) remains unclear.
MethodsPubMed, Web of Science, and Embase databases, as well as major international meetings (ASCO and ESMO), were systematically searched through February 2026. Clinical trials reporting patients with stage IA–IIIB NSCLC receiving neoadjuvant immunotherapy and documenting ctDNA levels together with outcome data, including pathologic complete response (pCR), major pathologic response (mPR), event-free survival (EFS), and overall survival (OS), were included.
ResultsTwelve studies comprising 10 clinical trials with 381 patients were included. Patients with ctDNA clearance had improved pCR (OR, 0.14; 95%CI, 0.05–0.40; I2=37.3%) and mPR (OR, 0.25; 95%CI, 0.07–0.85; I2=57%). This association was also observed in the subgroup receiving neoadjuvant chemoimmunotherapy for pCR (OR, 0.16; 95%CI, 0.05–0.49). ctDNA clearance was significantly associated with improved EFS (HR, 0.31; 95%CI, 0.21–0.46) and OS (HR, 0.31; 95%CI, 0.13–0.74). Pooled sensitivity for pCR prediction was 0.92 (95%CI, 0.79–1.00), specificity was 0.52 (95%CI, 0.44–0.59), positive predictive value was 0.49 (95%CI, 0.40–0.58), and negative predictive value was 0.92 (95%CI, 0.76–1.00). Significant heterogeneity was observed in diagnostic performance.
ConclusionsctDNA clearance before surgery correlates with pathologic response and survival outcomes in patients with NSCLC undergoing neoadjuvant immunotherapy. However, the prognostic utility of ctDNA clearance is limited by low specificity and substantial heterogeneity.
Lung cancer remains the leading cause of cancer-related mortality worldwide [1]. Non-small cell lung cancer (NSCLC) accounts for approximately 85% of all lung cancer cases [2]. Surgical resection is a feasible treatment option for patients with early-stage NSCLC; however, only approximately one-quarter of patients are eligible for surgery, and recurrence risk remains high [3]. Recent clinical trials have evaluated the efficacy of neoadjuvant immunotherapy in improving outcomes in early-stage disease [4]. Neoadjuvant immunotherapy has substantially altered the treatment paradigm for resectable NSCLC [5–7]. This led to approval by the US Food and Drug Administration of neoadjuvant pembrolizumab for patients with resectable NSCLC regardless of programmed death ligand-1 (PD-L1) expression level. However, the European Medicines Agency recommends neoadjuvant immunotherapy only for patients at high risk of recurrence with PD-L1-positive tumors. These discrepant approvals highlight the limitations of PD-L1 as a precise biomarker for patient stratification. Therefore, identification of novel and sensitive biomarkers for accurate recurrence-risk stratification in patients with NSCLC receiving neoadjuvant immunotherapy is urgently needed.
Circulating tumor DNA can reflect somatic alterations present in tumor tissue [8]. ctDNA has emerged as a useful noninvasive biomarker for predicting therapeutic response. For example, in colon cancer, ctDNA-guided treatment decisions can reduce unnecessary adjuvant chemotherapy use while maintaining recurrence-free survival outcomes [9]. In NSCLC, preoperative ctDNA has been identified as a phylogenetic biomarker associated with relapse risk [10]. Moreover, ctDNA may reflect mechanisms of resistance to targeted therapies in advanced-stage NSCLC [11]. Several studies evaluating neoadjuvant immunotherapy have suggested that ctDNA detection may predict pathologic response [12–14]. Prospective cohort studies have also indicated that ctDNA may predict survival outcomes in patients with NSCLC treated with neoadjuvant immunotherapy [15,16]. However, these studies are limited by small sample sizes and inconsistent findings. Consequently, it remains unclear whether ctDNA is a reliable circulating biomarker for predicting pathologic response and survival outcomes in early-stage NSCLC treated with neoadjuvant immunotherapy. Furthermore, the optimal timing for ctDNA assessment during preoperative immunotherapy remains uncertain. These limitations highlight a substantial gap in understanding the predictive value of ctDNA detection and dynamic ctDNA changes in patients with NSCLC receiving neoadjuvant immunotherapy.
In this meta-analysis, we aimed to evaluate the predictive and prognostic value of ctDNA levels in patients with early-stage NSCLC receiving neoadjuvant immunotherapy. We also sought to determine the optimal time points (baseline vs post-neoadjuvant treatment) for ctDNA assessment during neoadjuvant immunotherapy.
This study complies with the TITAN Guidelines 2025 (doi:10.70389/PJS.100082) [17].
MethodsStudy protocolThis systematic review and meta-analysis was conducted in accordance with the PRISMA guidelines [18]. The study protocol was registered in PROSPERO (CRD420251020476).
EndpointsThe primary endpoint of this study was to evaluate the predictive value of ctDNA levels for pathologic response (pCR and mPR) in patients with NSCLC receiving neoadjuvant immunotherapy. Sensitivity and specificity of ctDNA clearance for predicting pCR, together with positive predictive value (PPV) and negative predictive value (NPV), were also evaluated. Both baseline ctDNA levels and ctDNA clearance were assessed. ctDNA clearance was defined as disappearance of previously detectable ctDNA (measured at baseline) after completion of neoadjuvant immunotherapy and before surgery. Pathologic complete response was defined as complete eradication of viable tumor cells (0% residual viable tumor [RVT]) in surgical specimens from both the primary tumor and associated lymph nodes. Major pathologic response was defined as ≤10% RVT in the primary tumor and/or lymph nodes [19]. The predictive value of ctDNA clearance was also evaluated in the subgroup receiving neoadjuvant chemoimmunotherapy.
The secondary endpoint was to evaluate the prognostic value of ctDNA for EFS and OS in patients with NSCLC treated with neoadjuvant immunotherapy.
Eligibility criteriaStudies meeting the following eligibility criteria were included: (1) phase 1–3 registered clinical trials; (2) human studies; (3) published as full manuscripts or conference abstracts; (4) patients with stage IA-IIIB NSCLC receiving neoadjuvant immunotherapy with or without additional treatments; and (5) studies clearly documenting ctDNA levels together with outcome data, including pCR, mPR, EFS, and OS. All ctDNA detection and analytical methods were eligible.
Exclusion criteria included: (1) retrospective studies, reviews, editorials, protocols, comments, and letters to the editor; (2) metastatic or advanced NSCLC without surgical eligibility; and (3) studies from which outcome data could not be extracted.
Search strategy and study identificationPubMed, Web of Science, and Embase databases were searched in February 2026. Additional sources, including the ASCO website, ESMO website, preprint servers, and ClinicalTrials.gov, were also screened in February 2026. According to predefined eligibility criteria, potentially eligible full-text articles were independently reviewed by 2 authors (QS and SJS). Discrepancies were resolved by a third reviewer (ZJZ). Detailed search strategies for each database are provided in Table S1.
Data extractionThe following variables were extracted from eligible clinical trials: first author, publication year, clinical trial registration number, study phase, treatment arm, disease stage, immune checkpoint inhibitor agents and treatment cycles, combination therapies, quality assessment scores, total number of patients, number of patients receiving immunotherapy, number of patients undergoing ctDNA analysis, ctDNA collection time points, pathologic response outcomes (pCR and mPR), survival outcomes (EFS and OS), and ctDNA analytical methods.
In randomized studies comparing immunotherapy with nonimmunotherapy regimens, only patients receiving immune checkpoint inhibitors were included. Data extraction was independently performed by 2 reviewers (SLM and XPL), with disagreements resolved by a third reviewer (ZJZ).
Quality assessmentMethodological quality assessment was performed independently by 2 reviewers (QS and SJS) using the Reporting Recommendations for Tumor Marker Prognostic Studies (REMARK) checklist [20]. Disagreements were resolved by a third reviewer (SLM). Risk of bias in randomized controlled trials was evaluated using the Cochrane Risk of Bias 2 (ROB2) tool [21]. The ROBINS-I tool was used for nonrandomized studies [22].
Statistical analysisFor this meta-analysis, ctDNA status was analyzed as a binary variable (clearance vs detectable ctDNA or negative vs positive ctDNA). Overall effects for pCR and mPR analyses were evaluated using ORs together with corresponding 95%CIs. Pooled ORs were calculated by combining ORs derived from the included studies and analyzed using the METAN module in STATA. Pooled sensitivity, specificity, PPV, and NPV were calculated to evaluate diagnostic performance for pCR prediction.
For EFS and OS analyses, HRs and corresponding 95%CIs were obtained directly or calculated from reported data. Pooled HRs were estimated using the METAN module in STATA. Data synthesis required at least 2 eligible studies.
Heterogeneity was assessed using the I2 statistic and Cochran Q test in all analyses. When substantial heterogeneity was identified (I2>50%), random-effects models were used to account for heterogeneity. Otherwise, fixed-effects models were applied. Funnel plots together with Egger test were used to assess publication bias. All statistical analyses were performed using STATA version 18.0 (StataCorp LLC). Two-sided P<.05 was considered statistically significant.
This study was reported in accordance with the AMSTAR (Assessing the Methodological Quality of Systematic Reviews) guidelines [23].
ResultsStudy selection and characteristicsA total of 1893 records were initially identified through electronic database searches, including 939 from Embase, 213 from PubMed, and 741 from the Cochrane Library, without language restrictions. An additional 201 records were identified through other sources, including ASCO, ESMO, preprint servers, and ClinicalTrials.gov. After removal of 400 duplicate records, 1587 studies were excluded during title and abstract screening because they were unrelated to the study topic (n=1412), reviews or protocols (n=161), case reports (n=10), or editorials (n=4).
Following full-text review, an additional 95 studies were excluded for the following reasons: retrospective design (n=15), absence of ctDNA data (n=77), or nonclinical trial design (n=3). Ultimately, 10 published studies [5,12–16,24–27] and 2 conference abstracts [28,29] representing 10 clinical trials were included in the final analysis (Fig. 1).
The principal characteristics of the 12 included studies are summarized in Table 1. The clinical trials included 3 phase 3 studies [5,26,28,29] and 7 phase 2 studies [12–16,24,25,27]. Ten studies used anti-programmed cell death protein 1 (PD-1) antibodies (nivolumab, camrelizumab, sintilimab, and tislelizumab) for neoadjuvant immunotherapy [5,13–16,24–27,29], whereas 2 studies evaluated anti–PD-L1 therapy (durvalumab) [12,28]. In 8 trials, neoadjuvant immunotherapy was combined with chemotherapy [5,14–16,24–26,28,29], whereas 2 evaluated immunotherapy combined with novel therapeutic agents [12,13,27].
Study characteristics of the studies included in the systematic review and meta-analysis.
| Author | Year | Registered no. | Phase | Arm | Patient stage | ICI evaluated (No. of cycles) | ICIs plus other agents | Quality assessment |
|---|---|---|---|---|---|---|---|---|
| Patrick M. Forde | 2022 | NCT02998528 (CheckMate 816) | 3 | Double | IB–IIIA | Nivolumab (3) | Yes (platinum-doublet chemotherapy) | 30 |
| Mariano Provencio | 2022 | NCT03081689 (NADIM) | 2 | Single | IIIA | Nivolumab (3) | Yes (paclitaxel and carboplatin) | 31 |
| Si-Yang Liu | 2023 | NCT04015778 (CTONG1804) | 2 | Double | IIA–IIIB | Nivolumab (3) | Yes (nab-paclitaxel plus carboplatin) | 33 |
| Tina Cascone | 2023 | NCT03794544 (NeoCOAST) | 2 | Multiple | IA3 (>2cm)–IIIA | Durvalumab (1) | Yes (oleclumab, monalizumab, or danvatirsen) | 26 |
| Jun Zhao | 2023 | ChiCTR2000033588 | 2 | Single | IIA–IIIB | Camrelizumab (3) | Yes (apatinib) | 24 |
| Chao Zhang | 2024 | NCT05244213 (NEOTIDE/CTONG2104) | 2 | Single | IIB–IIIB | Sintilimab (3) | Yes (nab-paclitaxel plus carboplatin) | 25 |
| Jian-Zhen Shan | 2024 | NCT05024266 | 2 | Single | II–IIIB (N2) | Tislelizumab (4) | Yes (albumin-bound paclitaxel plus carboplatin) | 24 |
| Martin Reck | 2024 | NCT03800134 (AEGEAN) | 3 | Double | II–IIIB (N2) | Durvalumab (4) | Yes (platinum-based chemotherapy) | 21 |
| Long Xu | 2025 | NCT04422392 | 2 | Single | IIIA (N2) | Sintilimab (3) | Yes (chemotherapy) | 26 |
| Tina Cascone | 2025 | NCT04025879 (CheckMate 77T) | 3 | Double | IIA–IIIB (N2) | Nivolumab (4) | Yes (platinum-based chemotherapy) | 21 |
| Patrick M. Forde | 2025 | NCT02998528 (CheckMate 816) | 3 | Double | IB–IIIA | Nivolumab (3) | Yes (platinum-doublet chemotherapy) | 30 |
| Wei Guo | 2026 | ChiCTR2000033588 | 2 | Single | IIA–IIIB | Camrelizumab (3) | Yes (apatinib) | 24 |
Quality assessment using REMARK criteria yielded scores ranging from 21 to 33 (maximum score, 40) (Table S2). Overall, most included studies were considered of moderate methodological quality.
Among patients receiving perioperative immunotherapy, at least 381 underwent ctDNA analysis (Table 2). Eleven studies evaluated dynamic ctDNA changes [5,13–16,24–29], whereas 1 study assessed baseline ctDNA levels only [12]. The primary endpoint was pCR in 7 trials [5,13,15,16,24,25] and mPR in 6 trials [12–16,25]. The secondary endpoint was EFS in 5 trials [5,16,28] and OS in 3 trials [15,25]. Tumor-informed ctDNA approaches were used in 5 trials (50%) [12,14,16,28,30].
Number of patients included in the studies and ctDNA analysis with pathologic response and survival endpoint characteristics.
| Author | Year | Total patients | Patients receiving ICIs | ICI patients included in ctDNA analysis | ctDNA collection time points | ctDNA clearance detection time points | Pathologic responses (for ctDNA) | Survival outcomes (for ctDNA) | Method for ctDNA analysis |
|---|---|---|---|---|---|---|---|---|---|
| Patrick M. Forde | 2022 | 358 | 179 | 43 | Before cycle 1; before cycle 3 | Undetectable ctDNA before cycle 3 of neoadjuvant therapy | pCR | EFS | Tumor-guided personalized ctDNA panel |
| Mariano Provencio | 2022 | 46 | 46 | 43 | Baseline; post-neoadjuvant | Undetectable ctDNA after neoadjuvant therapy | pCR, mPR | OS, PFS | Oncomine Pan-Cancer Cell-Free Assay Kit |
| Si-Yang Liu | 2023 | 52 | 52 | 38 | Baseline; before cycle 3; post-neoadjuvant; after surgery | Undetectable ctDNA after neoadjuvant therapy | pCR, mPR | EFS | Tumor-informed ctDNA panel |
| Tina Cascone | 2023 | 83 | 83 | 33 | Baseline | NA | mPR | NA | Tumor-informed Signatera 16-plex assay |
| Jun Zhao | 2023 | 78 | 78 | 46 | Baseline; before cycle 2; before cycle 3; post-neoadjuvant | Undetectable ctDNA after neoadjuvant therapy and before surgery | pCR, mPR | NA | QIAamp Circulating Nucleic Acid Kit |
| Chao Zhang | 2024 | 18 | 18 | 12 | Baseline; before cycle 2; before cycle 3; post-neoadjuvant | NA | mPR | NA | Tumor-informed personalized ctDNA panel |
| Jian-Zhen Shan | 2024 | 35 | 35 | 26 | Baseline; post-neoadjuvant | Undetectable ctDNA after neoadjuvant therapy | pCR | NA | Personalized ctDNA panel PROPHET |
| Martin Reck | 2024 | 283 | 142 | NA | Baseline; before each cycle; post-neoadjuvant | NA | NA | EFS | Patient-specific tumor-informed assays |
| Long Xu | 2025 | 45 | 45 | 45 | Baseline; post-neoadjuvant | Undetectable ctDNA after neoadjuvant therapy and before surgery | pCR, mPR | OS | AmoyDx Blood and Leukocyte DNA Kit |
| Tina Cascone | 2025 | 461 | 229 | 76 | Baseline; post-neoadjuvant | Undetectable ctDNA after neoadjuvant therapy and before surgery | pCR | EFS | NA |
| Patrick M. Forde | 2025 | 358 | 179 | 43 | Before cycle 1; before cycle 3 | Undetectable ctDNA before cycle 3 of neoadjuvant therapy | pCR | OS | Tumor-guided personalized ctDNA panel |
| Wei Guo | 2026 | 78 | 78 | 65 | Baseline; before cycle 2; before cycle 3; post-neoadjuvant | Undetectable ctDNA after neoadjuvant therapy and before surgery | – | EFS | QIAamp Circulating Nucleic Acid Kit |
ctDNA, circulating tumor DNA; EFS, event-free survival; ICI, immune checkpoint inhibitor; mPR, major pathologic response; NA, not available; OS, overall survival; pCR, pathologic complete response; PFS, progression-free survival.
Three studies reported data regarding baseline ctDNA and pCR [15,16,24]. A total of 103 patients were included, of whom 83 (81%) had detectable ctDNA and 20 (19%) had undetectable ctDNA at baseline. Overall, baseline ctDNA detection was not associated with pCR in patients receiving neoadjuvant immunotherapy (OR, 1.02; 95%CI, 0.32–3.26; I2=0%; heterogeneity P=.46; Fig. 2A).
Seven studies reported data regarding ctDNA clearance and pCR [5,13,15,16,24,25,29]. Among 302 enrolled patients, 188 (62%) demonstrated ctDNA clearance and 114 (38%) had detectable ctDNA after neoadjuvant immunotherapy. Random-effects meta-analysis demonstrated that patients with ctDNA clearance had significantly improved pCR outcomes (OR, 0.14; 95%CI, 0.05–0.40; I2=37.3%; heterogeneity P=.14; Fig. 2B).
Subgroup analyses demonstrated that the association between ctDNA clearance and pCR was stronger when tumor-informed approaches were used (OR, 0.091; 95%CI, 0.025–0.329) compared with tumor-agnostic approaches (OR, 0.396; 95%CI, 0.137–1.140; Table 3). Sensitivity analysis confirmed the robustness of the meta-analysis (Fig. S1).
Subgroup meta-analysis of pCR and mPR in the ctDNA clearance cohort, stratified by detection method.
| Outcome | Included trials, no. | OR | 95%CI | p value |
|---|---|---|---|---|
| pCR stratified by ctDNA detection method | ||||
| Tumor-informed approach | 3 | 0.091 | 0.025–0.329 | <.0001 |
| Tumor-agnostic approach | 3 | 0.396 | 0.137–1.140 | .086 |
| mPR stratified by ctDNA detection method | ||||
| Tumor-informed approach | 3 | 0.533 | 0.194–1.465 | .176 |
| Tumor-agnostic approach | 2 | 0.146 | 0.047–0.457 | .001 |
The pooled sensitivity of ctDNA clearance for predicting pCR was 0.92 (95%CI, 0.79–1.00), with high heterogeneity (I2=66.98%; heterogeneity P=.01; Fig. S2A). The pooled specificity was 0.52 (95%CI, 0.44–0.59), with low heterogeneity (I2=0%; heterogeneity P=.46; Fig. S2B). The pooled PPV was 0.49 (95%CI, 0.40–0.58), with low heterogeneity (I2=28.14%; heterogeneity P=.21; Fig. S2C). The pooled NPV was 0.92 (95%CI, 0.76–1.00), with high heterogeneity (I2=78.03%; heterogeneity P<.001; Fig. S2D).
Association of ctDNA with mPRFour studies provided data regarding baseline ctDNA and mPR [12,15,16,25]. Among 155 patients, 113 (73%) had detectable ctDNA and 42 (27%) had undetectable ctDNA at baseline. Baseline ctDNA was not associated with mPR in patients receiving neoadjuvant immunotherapy (OR, 2.16; 95%CI, 0.89–5.27; I2=0%; heterogeneity P=.55; Fig. 3A).
Five studies were included in the analysis evaluating ctDNA clearance and mPR [13–16,25]. Among 170 patients, 112 (66%) demonstrated ctDNA clearance and 58 (34%) had detectable ctDNA after neoadjuvant immunotherapy. Overall, ctDNA clearance was significantly associated with improved mPR (OR, 0.25; 95%CI, 0.07–0.85; I2=57%; heterogeneity P=.05; Fig. 3B). Sensitivity analysis confirmed the robustness of the meta-analysis (Fig. S3).
The pooled sensitivity of ctDNA clearance for predicting mPR was 0.80 (95%CI, 0.69–0.89), with low heterogeneity (I2=23.01%; heterogeneity P=.27; Fig. S4A). The pooled specificity was 0.49 (95%CI, 0.23–0.76), with high heterogeneity (I2=77.68%; heterogeneity P<.001; Fig. S4B). The pooled PPV was 0.70 (95%CI, 0.51–0.86), with high heterogeneity (I2=74.75%; heterogeneity P<.001; Fig. S4C). The pooled NPV was 0.64 (95%CI, 0.34–0.91), with high heterogeneity (I2=70.87%; heterogeneity P=.01; Fig. S4D).
Association of ctDNA clearance with pCR and mPR during neoadjuvant chemoimmunotherapyBecause most clinical investigations have focused on chemoimmunotherapy combinations in the neoadjuvant setting, the associations between ctDNA clearance before surgery and pCR and mPR during neoadjuvant chemoimmunotherapy were further evaluated.
Association of ctDNA clearance with pCRSix studies were included in the analysis evaluating ctDNA clearance and pCR during neoadjuvant chemoimmunotherapy [5,15,16,24,25,29]. Among 256 patients, 155 (61%) demonstrated ctDNA clearance, whereas 101 (39%) had detectable ctDNA after neoadjuvant chemoimmunotherapy. Patients with ctDNA clearance had significantly improved pCR outcomes (OR, 0.16; 95%CI, 0.05–0.49; I2=41.8%; heterogeneity P=.13; Fig. 4A). Sensitivity analysis confirmed the robustness of the meta-analysis (Fig. S5A).
Association of ctDNA clearance with mPR
Four studies were included to evaluate the association between ctDNA clearance and mPR during neoadjuvant chemoimmunotherapy [14–16,25]. Among 158 patients, 101 (64%) demonstrated ctDNA clearance and 57 (36%) had detectable ctDNA after treatment. Overall, ctDNA clearance was not significantly associated with mPR outcomes (OR, 0.21; 95%CI, 0.04–1.17; I2=67.3%; heterogeneity P=.03; Fig. 4B). Sensitivity analysis confirmed the robustness of the meta-analysis (Fig. S6A).
Association of ctDNA with EFS and OS during neoadjuvant immunotherapyFive studies were included to investigate the prognostic effect of ctDNA clearance on EFS [5,16,27–29]. ctDNA clearance was significantly associated with improved EFS (HR, 0.31; 95%CI, 0.21–0.46; I2=0%; heterogeneity P=.44; Fig. 5A).
In addition, 3 studies were included in the analysis evaluating the association between ctDNA and OS [15,25,26]. Fixed-effects meta-analysis demonstrated that ctDNA clearance was significantly associated with improved OS outcomes (HR, 0.31; 95%CI, 0.13–0.74; I2=27.9%; heterogeneity P=.25; Fig. 5B).
Risk of biasRisk of bias assessment using the ROB2 tool for randomized controlled trials is presented in Fig. S7A. Among the included randomized controlled trials, 4 studies demonstrated low overall risk of bias, whereas the remaining studies raised some concerns, mainly related to selection of the reported results.
The ROBINS-I tool was additionally used to evaluate risk of bias in nonrandomized studies. Among the 7 nonrandomized studies, 5 (71.4%) demonstrated moderate risk of bias, whereas the remaining 2 studies demonstrated low risk of bias (Fig. S7B). The principal sources of bias in nonrandomized studies were participant selection (n=4) and outcome measurement (n=4).
Funnel plot analyses demonstrated no significant asymmetry for mPR and pCR analyses in the overall neoadjuvant immunotherapy cohort or in the neoadjuvant chemoimmunotherapy subgroup (Figs. S3B, S5B, and S6B), suggesting absence of substantial publication bias.
DiscussionTo our knowledge, this represents the first comprehensive meta-analysis evaluating the value of ctDNA for predicting pathologic response and prognosis in patients with NSCLC receiving neoadjuvant immunotherapy using prospective clinical trial data. Overall, we found that preoperative ctDNA dynamics were associated with pCR, mPR, EFS, and OS in patients with NSCLC receiving neoadjuvant immunotherapy. Moreover, ctDNA clearance correlated with pCR and mPR in patients undergoing neoadjuvant chemoimmunotherapy.
Pathologic complete response and mPR are considered reliable surrogate endpoints for prognosis following neoadjuvant immunotherapy in NSCLC [19,31]. Our meta-analysis demonstrated that ctDNA clearance after neoadjuvant immunotherapy was associated with improved pCR and mPR in patients with NSCLC. However, baseline ctDNA was not predictive of pCR or mPR, and wide confidence intervals indicated uncertainty regarding these findings. Therefore, additional clinical investigations evaluating baseline ctDNA are warranted to provide more robust evidence. These findings also emphasize the importance of ctDNA assessment after neoadjuvant immunotherapy for predicting tumor response before surgery.
Our findings are consistent with a previous meta-analysis demonstrating that ctDNA clearance was associated with pCR in patients with solid tumors receiving neoadjuvant immunotherapy [32]. Zhou et al. [33] also performed a narrative review concluding that ctDNA has strong potential in the neoadjuvant treatment setting for NSCLC. Collectively, current evidence suggests that ctDNA clearance, but not baseline ctDNA levels, correlates with pCR and mPR in patients with NSCLC undergoing neoadjuvant immunotherapy.
Neoadjuvant chemoimmunotherapy has emerged as a standard treatment strategy for patients with surgically resectable NSCLC. In addition, neoadjuvant chemoimmunotherapy has been shown to prolong EFS and significantly increase pCR rates in patients with NSCLC [34]. Sorin et al. [35] conducted a meta-analysis demonstrating superior outcomes with neoadjuvant chemoimmunotherapy compared with neoadjuvant chemotherapy alone. ctDNA monitoring during induction chemoimmunotherapy may predict treatment outcomes in patients with advanced NSCLC [36].
In a recent narrative review, Kaira et al. [37] concluded that pretreatment ctDNA levels and ctDNA clearance after neoadjuvant chemoimmunotherapy were strongly associated with survival outcomes. However, the relationship between ctDNA levels and pCR had not been conclusively established. Therefore, we further analyzed the association between ctDNA clearance and pathologic response during neoadjuvant chemoimmunotherapy in NSCLC. Our meta-analysis demonstrated that ctDNA clearance was significantly associated with improved pCR in patients receiving neoadjuvant chemoimmunotherapy, suggesting that ctDNA clearance may serve as a useful predictor of pCR in this setting.
Recent evidence has also highlighted the prognostic utility of ctDNA in NSCLC [38]. A clinical trial (NCT04367311) demonstrated that detectable ctDNA after surgery may represent a potential biomarker for survival outcomes in patients with NSCLC receiving adjuvant immunotherapy [39]. Nevertheless, the clinical utility of ctDNA for predicting survival outcomes in patients receiving neoadjuvant immunotherapy has not been systematically evaluated. Our meta-analysis demonstrated that persistent ctDNA detection before surgery was associated with significantly worse EFS and OS in patients undergoing neoadjuvant immunotherapy. Further clinical trials are warranted to clarify the prognostic role of ctDNA dynamics in this setting.
LimitationsThis study has several limitations. First, only 10 clinical trials were included in the meta-analysis. Several potentially eligible studies were excluded because of unavailable outcome data [40–42]. In addition, many studies reported incomplete information regarding primary and secondary endpoints.
Second, the overall sample size remained relatively small. Third, this meta-analysis was based on published clinical trial data without access to individual patient-level data.
Fourth, substantial heterogeneity existed regarding the timing of ctDNA assessment across studies. Consequently, only 2 major time points could be evaluated: baseline and post-neoadjuvant treatment before surgery (ctDNA clearance). The latter category was particularly heterogeneous, mainly because of differences in therapeutic regimens (PD-1 vs PD-L1 inhibitors; with or without chemotherapy) and treatment duration (range, 1–4 cycles).
Finally, marked heterogeneity existed in ctDNA analytical methodologies across the included studies. The influence of analytical variability, including tumor-informed versus tumor-agnostic assays, limits of detection, sequencing depth, and timing of blood sampling, remains unclear, underscoring the need for technical standardization of ctDNA detection methodologies in future studies.
ConclusionsThis meta-analysis demonstrated that ctDNA clearance before surgery, but not baseline ctDNA detection, significantly correlates with pCR and mPR in patients with NSCLC undergoing neoadjuvant immunotherapy. This association was also observed in the subgroup receiving neoadjuvant chemoimmunotherapy for pCR. Persistent preoperative ctDNA detection was associated with significantly worse EFS and OS in patients receiving neoadjuvant immunotherapy.
These findings highlight the growing importance of ctDNA in guiding therapeutic decision-making for early-stage NSCLC. They also provide a framework for future clinical trials and prospective studies evaluating ctDNA detection in neoadjuvant settings, with the potential to improve clinical outcomes further.
Because of its minimally invasive nature, ease of repeated assessment, and promising predictive performance, ctDNA clearance may represent a feasible surrogate endpoint for pathologic response assessment during neoadjuvant immunotherapy in NSCLC. With further validation, ctDNA clearance may potentially replace pathologic response as an early surrogate endpoint in neoadjuvant trials and facilitate approval of novel immunotherapies for resectable NSCLC.
Authors’ contributionsDr. Chen had full access to all study data and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Song, Yu, Chen. Acquisition, analysis, or interpretation of data: All authors. Drafting of the manuscript: Song, Shang, Zhai, Ma, Chen. Critical revision of the manuscript for important intellectual content: Song, Shang, Zhai, Ma, Li, Chen. Statistical analysis: Song, Shang, Zhai, Ma. Funding acquisition: Song, Yu, Chen. Administrative, technical, or material support: Song, Li, Mo, Zhang, Yu, Chen. Supervision: Yu, Chen.
Artificial intelligence involvementThe authors declare that no part of this manuscript was generated, partially or completely, using artificial intelligence software or tools.
FundingThis work was supported by grants from the National Natural Science Foundation of China (82030082, 82172676, 82373217, and 82403455), the Natural Science Foundation of Shandong (ZR2023ZD26 and ZR2024QH527), the Noncommunicable Chronic Diseases – National Science and Technology Major Project, the Distinguished Young Scholars of Shandong Provincial Science Fund (ZR2024JQ032), the Shandong Province International Science and Technology Cooperation Project (2025KJHZ012), and the China Postdoctoral Science Foundation (2024M761869 and 2025T180617).
Conflicts of interestNone declared.
Data availability statementThis manuscript includes tertiary use of data. Primary data sources have been previously published and are publicly available.
The followings are the supplementary data to this article:
