Circulating Tumor cells and multiple indicators combined to identify the risk of poorer prognosis in patients with resected non-small cell lung cancer Authors: Jinghan Song, Xiong Ye, Qianqian Peng, Xinnan Ying, Hui Xiao Published: 2024 ABSTRACT BackgroundSurgical resection is an important treatment option for patients with non-small cell lung cancer (NSCLC). However, recurrence and survival rates remain a cause of concern. To further improve prognosis, more studies have focused on liquid biopsy, which has significant value as a prognostic factor for defining the risk stratification of postoperative NSCLC patients. This study aimed to identify circulating tumor cells (CTCs) as biomarkers that indicate a poor prognosis, combined with multiple indicators to determine prognostic risks in advance and develop individualized treatment strategies.MethodsBetween November 2015 and August 2018, 65 radical resected patients with NSCLC were analyzed. Preoperative CTCs were collected, and follow-up lasted until August 2023. Overall survival (OS) and disease-free survival (DFS) were the primary outcomes.ResultsWith an 11 CTC unit threshold, the high preoperative CTC level group had worse OS and DFS than the low-level group, suggesting that preoperative CTC levels have prognostic value. Time-dependent receiver operating characteristic (ROC) curves also showed satisfactory predictive efficiency of CTCs. Univariate analysis revealed that preoperative CTC levels were significantly associated with increasing risks for OS and DFS. Moreover, we combined CTCs and multiple indicators to provide a reference for a group at high risk of adverse outcomes.ConclusionsCTCs serve as feasible biomarkers for predicting postoperative prognosis in NSCLC patients. The combination of hematological, radiological, and pathological features could be valuable tools to guide postoperative management and treatment decisions in these patients. A multimodal prognostic approach is important for the clinical evaluation of lung cancer. Introduction Lung cancer is a malignant disease that develops in the mucosa or glandular tissues of the trachea and bronchi. Lung cancer has the consistently highest mortality rate among all malignant tumors worldwide [1, 2]. Histologically, lung cancer is categorized into small cell lung cancer (SCLC) and non-small cell lung cancer (NSCLC), with NSCLC comprising 80–85% of cases [3, 4]. Surgery is considered the most effective curative approach for operable lung cancer [5] and offers the opportunity for a favorable prognosis. Surgical treatment has proven to be highly effective for early-stage lung cancer, significantly improving patient survival rates. For advanced stages, although surgery may not be the primary treatment, it can still provide potential benefits, particularly when used in conjunction with other therapies to improve outcomes and manage disease progression. However, despite curative resection, 30–55% of patients with NSCLC experience recurrence and ultimately succumb to the disease. Although the pathological tumor-node-metastasis (TNM) classification remains the most effective prognostic indicator for NSCLC, it has its limitations. Patients classified under the same TNM stage can have varying prognoses [6]. Even following radical cancer surgery, metastatic potential can persist, such as in the form of minimal residual disease (MRD). Therefore, identifying factors associated with survival and recurrence is essential for enhancing the prognosis of lung cancer. Currently, research efforts are concentrated on discovering novel, non-invasive methods for identifying patients at high risk of postoperative recurrence and metastasis ahead of time and providing suitable clinical interventions to track tumor progression and observe the therapeutic response in real time [7, 8]. In this context, liquid biopsy has emerged as a revolutionary tool in various aspects of oncology and precision tumor medicine. Liquid biopsy is a non-invasive diagnostic tool that detects proteins, circulating tumor cells (CTCs), circulating tumor RNA (ctRNA), circulating tumor DNA (ctDNA), and exosomes [9, 10]. Compared to needle biopsy, its non-invasive approach permits the continuous collection of patient blood samples, enabling real-time monitoring of disease progression for personalized treatment [11]. Moreover, liquid biopsy provides more comprehensive information [12, 13] since both primary and metastatic tumors can release CTC, ctDNA, extracellular vesicles (EVs), etc. Among them, circulating tumor cells (CTCs) are cancer cells that detach from a primary tumor lesion or metastatic deposit and enter the bloodstream which is the main target of liquid biopsy, as CTCs play a crucial role in the metastasis process [14]. Studies have shown that CTCs can be detected even before metastasis occurs [15]. Detecting CTCs in the blood allows for diagnosing earlier and controlling better of cancer, helping to prevent the adverse effects originating from metastasis. Additionally, CTCs can be utilized to assess patients’ prognosis and evaluate treatment outcomes in real time [16]. Although liquid biopsy has become a crucial alternative tool for cancer diagnosis and prognosis of poor survival [17] and serves as an independent predictor of poor prognosis, independent of traditional clinicopathological features such as tumor size and lymph node metastasis, its predictive efficacy for lung cancer prognosis and its role in guiding the follow-up treatment after surgery has not been clarified. Meanwhile, existing studies have employed various detection methods and have utilized different staining protocols at the same time [18, 19]. Thus, CTC detection hasn’t shown great use in a daily clinical routine yet. In this study, we conducted a real-world retrospective analysis of the prognosis of postoperative lung cancer patients, stratified by preoperative CTC levels. Furthermore, we examined the detailed relationships between preoperative CTC levels and lung cancer prognosis. The primary aim of this study was to evaluate the significance of preoperative CTC levels, as well as their integration with multiple prognostic indicators, in predicting outcomes for lung cancer patients undergoing surgical treatment. These findings may contribute to guiding follow-up management and adjuvant treatment strategies, ultimately improving postoperative survival outcomes for these patients. Our study offers novel insights into the practical applications of preoperative CTC levels in a real-world clinical setting. Materials and methods Patients This single-center retrospective cohort study included patients with pathologically confirmed NSCLC who were consecutively hospitalized at Shanghai General Hospital between November 2015 and August 2018. The inclusion criteria included: (1) patients aged 18 to 85 years; (2) individuals who had undergone radical surgery for NSCLC; (3) TNM stage of 0–3; (4) the blood tests, CTCs count measurements, and radiological examinations were conducted prior to the operation; (5) complete postoperative pathological parameters. The exclusion criteria were defined as follows: (1) patients who received preoperative radiotherapy, chemotherapy, or any other forms of treatment; (2) patients with other malignancies or diseases that significantly impacted their survival and prognosis were eliminated; (3) missing data on clinicopathological characteristics and follow-up. Disease stages were classified according to the 9th edition of the American Joint Committee on Cancer (AJCC) Staging manual. Under the guidance of three thoracic surgeons, the surgical procedures were executed at the thoracic center. The operations that were carried out included lobectomies and two-hole thoracoscopic wedge resections of lung nodules. A total of 65 patients were ultimately selected for the study, and their clinical and radiological data were obtained. The process of selecting patients is illustrated in Fig. 1. Overall survival time was defined as the duration from radical surgery to death due to lung cancer. Disease-free survival time was defined as the period from randomization to either the first recurrence or metastasis of the tumor or death from any cause. The last follow-up date was August 16, 2023. If the patient was still alive at the final censored follow-up, he/she was considered not to have met the study endpoint. The follow-up information came from outpatient follow-up reviews or telephone follow-ups. Fig. 1The flowchart of the patient selection process The flowchart of the patient selection process Study design This retrospective trial was conducted at our hospital’s clinical center within the Thoracic Surgery and Department of Respiratory and Critical Care Medicine. This study evaluated the concentration of tumor markers, the number of CTCs, radiological analyses, and pathological features. The primary inclusion criteria were lung cancer patients undergoing surgery and the availability of preoperative CTC unit data and complete prognostic information. Efforts were undertaken to mitigate potential sources of information bias. By establishing inclusion criteria, we ensured that all eligible patients were included. To minimize information bias, the detection of CTCs and tumor markers, along with radiological and pathological analyses, were conducted by qualified specialists specifically invited for this purpose. These specialists utilized pre-defined and validated assessment tools to ensure accurate results. The present retrospective study was approved by the Institutional Review Board at Shanghai General Hospital, Shanghai Jiao Tong University, and all experiments were conducted in accordance with the relevant guidelines of the Institutional Ethics Committee. This study was retrospective, utilizing previously collected data from the hospital system without involving tissue or other samples. Additionally, no personal information or private data were included. Therefore, informed consent from study participants was deemed not applicable and was waived by the Ethics Committee. Detection of CTCs Following the procedures detailed by Liu et al. and Ge et al. [20, 21], the detection of CTCs was conducted. In summary, the enrichment and identification of CTCs were carried out using the Cytelligen CTC Enrichment Kit and Human Tumor Cell Identification Kit (Cytelligen, San Diego, CA, USA). Before surgery, blood samples were drawn from the antecubital vein of all patients into 6-ml tubes containing ACD. The technicians, who were blinded to the sample identities, performed the CTC detection tests on de-identified samples. Following centrifugation, combining the tumor-associated antigen CK18 (indicating epithelial cells) and the CD45 antibody (indicating non-leukocyte cells) with the white buffy that was collected. The resultant cell pellet which has been magnetic separated from the mixture was resuspended in 100 µl of fixative before being applied to the CTC slides. Applying a Centromere Probe (CEP 8) Spectrum Orange from Vysis, Abbott Laboratories (Abbott Park, IL, USA) to the slides for denaturation and hybridization and incubating the slides in the dark for 2 h at room temperature. Following two washes, 5 µl of mounting medium (Ultra Cruz, Santa Cruz Biotechnology, CA, USA) containing 4’, 6-diamidino-2-phenylindole (DAPI, which stains the nuclei of captured cells) was added. A Carl Zeiss Axioplan 2 imaging fluorescence microscope (ZEISS Company, Germany) was used to perform cell counting. CTCs were identified based on the criteria DAPI+/CK18+/CD45−/CEP8 ≥ 3 spots. Counting each sample three times, and taking the average value as the final count for each participant. Measurement of tumor markers The serum of peripheral venous blood samples was separated after collecting from patients prior to surgery. Using an electrochemiluminescence analyzer from Roche Diagnostics (Bavaria, Germany), the concentration of serum tumor markers including CEA, CA125, CYFRA 21-1, NSE, and SCC was determined. Each test was conducted in strict accordance with the procedures outlined in the instrument’s operating manual. Imaging Chest computed tomography was initially conducted to the identification of lung nodules. Two certified radiologists evaluated the images, which contained various features including nodule diameter, margin, density, pleural traction, vascular bundle sign, and burr sign. Statistical analysis Receiver operating characteristic (ROC) curve analysis was used to determine the best cutoff values for predicting the prognosis of lung cancer patients. The area under the ROC curve (AUC) for each index was calculated. The Youden index was then utilized to determine the optimal cutoff point and evaluate the efficiency of the predictions. Survival curves were generated using the Kaplan–Meier method. A univariate Cox proportional hazards model was used to assess the associations between preoperative CTC level and other indicators and OS and DFS. Covariates found to be significant in the univariate analysis were then input into a multivariate model to generate adjusted hazard ratios (HRs). Stepwise regression was applied to adjust for multiple comparisons, creating a statistically significant multivariate model. All statistical significance tests were two-sided, with a p-value of less than 0.05 indicating statistical significance. The statistical analyses were conducted using SPSS software (version 25.0; SPSS Inc., Chicago, IL, USA), Prism 8.0 (GraphPad Software Inc., San Diego, CA, USA), R Foundation for Statistical Computing (Vienna), and MedCalc 22 software. Results Patient characteristics Between November 2015 and August 2018, we enrolled 87 patients in the study, of which 65 met the inclusion and exclusion criteria. The median duration of follow-up was 73 months (IQR, 63–77 months). The patient demographics are shown in Table 1. In our population, 81.54% of patients had TNM stage I, 90.77% had adenocarcinoma, and 36.92% received postoperative adjuvant chemotherapy. Table 1Clinicopathological variables of patients with NSCLCVariablesTotal (n = 65)Sex, n(%) Male25 (38.46) Female40 (61.54) Age, Median (range)61.00 (54.00, 68.00)TNM stage, n(%) 05 (7.69) I53 (81.54) II6 (9.23) III1 (1.54)Histology, n(%) ACIS5 (7.69) MIA22 (33.85) IAC32 (49.23) Others6 (9.23)Location, n(%) Others24 (36.92) Upper lobe41 (63.08) CTC, Median (range)7.00 (3.00, 13.00) NSE, Median (range)12.82 (11.44, 14.97) CEA, Median (range)1.98 (1.38, 2.94) CYFRA21-1, Median (range)2.18 (1.57, 3.08) CA125, Median (range)10.90 (7.00, 13.90) SCC, Median (range)0.70 (0.50, 1.10) MTD, Median (range)13.00 (8.00, 25.00)CTH, n(%) No41 (63.08) Yes24 (36.92)Spiculation, n(%) No51 (78.46) Yes14 (21.54)PI, n(%) No54 (83.08) Yes11 (16.92)MTD maximum tumor diameter; CTH chemotherapy; PI pleural invasion Table 1Clinicopathological variables of patients with NSCLCVariablesTotal (n = 65)Sex, n(%) Male25 (38.46) Female40 (61.54) Age, Median (range)61.00 (54.00, 68.00)TNM stage, n(%) 05 (7.69) I53 (81.54) II6 (9.23) III1 (1.54)Histology, n(%) ACIS5 (7.69) MIA22 (33.85) IAC32 (49.23) Others6 (9.23)Location, n(%) Others24 (36.92) Upper lobe41 (63.08) CTC, Median (range)7.00 (3.00, 13.00) NSE, Median (range)12.82 (11.44, 14.97) CEA, Median (range)1.98 (1.38, 2.94) CYFRA21-1, Median (range)2.18 (1.57, 3.08) CA125, Median (range)10.90 (7.00, 13.90) SCC, Median (range)0.70 (0.50, 1.10) MTD, Median (range)13.00 (8.00, 25.00)CTH, n(%) No41 (63.08) Yes24 (36.92)Spiculation, n(%) No51 (78.46) Yes14 (21.54)PI, n(%) No54 (83.08) Yes11 (16.92)MTD maximum tumor diameter; CTH chemotherapy; PI pleural invasion Clinicopathological variables of patients with NSCLC MTD maximum tumor diameter; CTH chemotherapy; PI pleural invasion Association of the preoperative CTC level and OS/DFS in patients with lung cancer After nearly eight years of follow-up, 10 lung cancer patients died and 17 lung cancer patients developed disease progression. The cut-off value (CTC = 11) of preoperative CTC units predicted in 5-year OS and 5-year DFS was calculated by ROC curve analysis, which was divided into low-level (CTC ≤ 11) and high-level groups (CTC > 11). Compared with patients with low preoperative CTC levels, patients with high preoperative CTC levels had worse overall survival (OS) (p = 0.014) and progression-free survival (DFS) (P = 0.023) (Fig. 2), indicating that preoperative CTC levels may have prognostic value. Fig. 2The Kaplan–Meier analysis curves for CTC levels on OS (A) and DFS (B). Patients in the higher CTC level group had poorer prognoses with shorter OS(p = 0.014) and DFS(p = 0.023) The Kaplan–Meier analysis curves for CTC levels on OS (A) and DFS (B). Patients in the higher CTC level group had poorer prognoses with shorter OS(p = 0.014) and DFS(p = 0.023) The prognostic ability of the preoperative CTCs We evaluated the predictive efficacy of preoperative CTC units using ROC analysis for different follow-up time points. As shown in Fig. 3, the area under the curve (AUCs) of preoperative CTC units for predicting OS were 0.857, 0.623, and 0.667 at 1, 3, and 5 years, respectively, and for predicting DFS were 0.857, 0.705, and 0.606 at 1, 3, and 5 years, respectively, indicating a satisfactory prognostic predictive efficiency for postoperative NSCLC patients. The efficiency varies with the follow-up time. 1-year OS(AUC = 0.857) and DFS(AUC = 0.857) have the highest predictive efficacy which shows great significance for early postoperative outcomes. Fig. 3Time-dependent ROC curve analyses of using the CTC threshold of 11 units/6 ml for the prediction of 1-year, 3-year, and 5-year OS (A) and DFS (B) Time-dependent ROC curve analyses of using the CTC threshold of 11 units/6 ml for the prediction of 1-year, 3-year, and 5-year OS (A) and DFS (B) The Cox analysis of hematological, radiological, and pathological factors According to the univariate Cox analysis, high preoperative CTC levels had significant influences and higher risks of OS(p = 0.024) and DFS(p = 0.030) than low preoperative CTC levels (Tables 2 and 3)(Fig. 4A, B). In multivariate Cox analysis, preoperative CTC levels correlated with poor prognosis second only to pleural invasion (Tables 2 and 3)(Fig. 4C, D) according to HR (Hazard Ratio) and remained a strong and independent predictor of OS. Preoperative CTC levels(HR = 2.82) demonstrated superior prognostic predictive ability than TNM staging(HR = 0.90). Meanwhile, univariate Cox analysis showed that TNM stage, CEA, CYFRA21-1, CA125, MTD, and PI were all correlated with OS and DFS, whereas Spiculation was only correlated with OS. In this cohort, the preoperative CTC level was an independent prognostic factor for OS. And MTD was an independent prognostic factor for both OS and DFS. Table 2Univariate and multivariate analysis of clinicopathological characteristics associated with overall survival (OS)VariablesUnivariate analysisMultivariate analysisPHR (95%CI)PHR (95%CI)CTC Low1.00 (Reference)1.00 (Reference) High0.0244.31 (1.21 ~ 15.28)0.0455.89 (1.04 ~ 33.36)TNM stage 0-I1.00 (Reference)1.00 (Reference) II-III< 0.00111.71 (3.36 ~ 40.76)0.5890.49 (0.04 ~ 6.58) NSE0.9821.00 (0.83 ~ 1.21) CEA0.0191.01 (1.01 ~ 1.02)0.2350.99 (0.97 ~ 1.01) CYFRA21-10.0041.52 (1.14 ~ 2.02)0.5801.20 (0.63 ~ 2.31) CA1250.0221.08 (1.01 ~ 1.16)0.3001.07 (0.94 ~ 1.22) SCC0.4541.25 (0.70 ~ 2.25) MTD< 0.0011.11 (1.06 ~ 1.17)0.0101.10 (1.02 ~ 1.19)CTH No1.00 (Reference) Yes0.3501.81 (0.52 ~ 6.24)Spiculation No1.00 (Reference)1.00 (Reference) Yes0.0293.98 (1.15 ~ 13.78)0.2543.05 (0.45 ~ 20.74)PI No1.00 (Reference)1.00 (Reference) Yes< 0.0019.55 (2.68 ~ 34.03)0.0576.50 (0.95 ~ 44.57)MTD maximum tumor diameter; CTH chemotherapy; PI pleural invasion Table 2Univariate and multivariate analysis of clinicopathological characteristics associated with overall survival (OS)VariablesUnivariate analysisMultivariate analysisPHR (95%CI)PHR (95%CI)CTC Low1.00 (Reference)1.00 (Reference) High0.0244.31 (1.21 ~ 15.28)0.0455.89 (1.04 ~ 33.36)TNM stage 0-I1.00 (Reference)1.00 (Reference) II-III< 0.00111.71 (3.36 ~ 40.76)0.5890.49 (0.04 ~ 6.58) NSE0.9821.00 (0.83 ~ 1.21) CEA0.0191.01 (1.01 ~ 1.02)0.2350.99 (0.97 ~ 1.01) CYFRA21-10.0041.52 (1.14 ~ 2.02)0.5801.20 (0.63 ~ 2.31) CA1250.0221.08 (1.01 ~ 1.16)0.3001.07 (0.94 ~ 1.22) SCC0.4541.25 (0.70 ~ 2.25) MTD< 0.0011.11 (1.06 ~ 1.17)0.0101.10 (1.02 ~ 1.19)CTH No1.00 (Reference) Yes0.3501.81 (0.52 ~ 6.24)Spiculation No1.00 (Reference)1.00 (Reference) Yes0.0293.98 (1.15 ~ 13.78)0.2543.05 (0.45 ~ 20.74)PI No1.00 (Reference)1.00 (Reference) Yes< 0.0019.55 (2.68 ~ 34.03)0.0576.50 (0.95 ~ 44.57)MTD maximum tumor diameter; CTH chemotherapy; PI pleural invasion Univariate and multivariate analysis of clinicopathological characteristics associated with overall survival (OS) MTD maximum tumor diameter; CTH chemotherapy; PI pleural invasion Table 3Univariate and multivariate analysis of clinicopathological characteristics associated with disease-free survival (DFS)VariablesUnivariate analysisMultivariate analysisPHR (95%CI)PHR (95%CI)CTC Low1.00 (Reference)1.00 (Reference) High0.0302.88 (1.11 ~ 7.49)0.0622.82 (0.95 ~ 8.35)TNM stage 0-I1.00 (Reference)1.00 (Reference) II-III0.0025.22 (1.83 ~ 14.93)0.9170.90 (0.14 ~ 6.05) NSE0.5411.04 (0.91 ~ 1.19) CEA0.0281.01 (1.01 ~ 1.01)0.3860.99 (0.98 ~ 1.01) CYFRA21-1< 0.0011.52 (1.21 ~ 1.91)0.1011.30 (0.95 ~ 1.77) CA1250.0411.06 (1.01 ~ 1.12)0.5511.02 (0.95 ~ 1.10) SCC0.6151.14 (0.69 ~ 1.88) MTD< 0.0011.07 (1.03 ~ 1.11)0.0341.05 (1.01 ~ 1.10)CTH No1.00 (Reference) Yes0.1032.21 (0.85 ~ 5.73)Spiculation No1.00 (Reference) Yes0.3571.63 (0.58 ~ 4.64)PI No1.00 (Reference)1.00 (Reference) Yes0.0123.57 (1.32 ~ 9.70)0.1403.12 (0.69 ~ 14.20)MTD maximum tumor diameter; CTH chemotherapy; PI pleural invasion Table 3Univariate and multivariate analysis of clinicopathological characteristics associated with disease-free survival (DFS)VariablesUnivariate analysisMultivariate analysisPHR (95%CI)PHR (95%CI)CTC Low1.00 (Reference)1.00 (Reference) High0.0302.88 (1.11 ~ 7.49)0.0622.82 (0.95 ~ 8.35)TNM stage 0-I1.00 (Reference)1.00 (Reference) II-III0.0025.22 (1.83 ~ 14.93)0.9170.90 (0.14 ~ 6.05) NSE0.5411.04 (0.91 ~ 1.19) CEA0.0281.01 (1.01 ~ 1.01)0.3860.99 (0.98 ~ 1.01) CYFRA21-1< 0.0011.52 (1.21 ~ 1.91)0.1011.30 (0.95 ~ 1.77) CA1250.0411.06 (1.01 ~ 1.12)0.5511.02 (0.95 ~ 1.10) SCC0.6151.14 (0.69 ~ 1.88) MTD< 0.0011.07 (1.03 ~ 1.11)0.0341.05 (1.01 ~ 1.10)CTH No1.00 (Reference) Yes0.1032.21 (0.85 ~ 5.73)Spiculation No1.00 (Reference) Yes0.3571.63 (0.58 ~ 4.64)PI No1.00 (Reference)1.00 (Reference) Yes0.0123.57 (1.32 ~ 9.70)0.1403.12 (0.69 ~ 14.20)MTD maximum tumor diameter; CTH chemotherapy; PI pleural invasion Univariate and multivariate analysis of clinicopathological characteristics associated with disease-free survival (DFS) MTD maximum tumor diameter; CTH chemotherapy; PI pleural invasion Fig. 4Independent prognostic value of the prognostic models. Forest plot of univariate Cox regression analysis (A) and multivariate Cox regression analysis (C) on OS. Forest plot of univariate Cox regression analysis (B) and multivariate Cox regression analysis (D) on DFS. The preoperative CTC level is an independent prognostic factor for OS(p = 0.045) Independent prognostic value of the prognostic models. Forest plot of univariate Cox regression analysis (A) and multivariate Cox regression analysis (C) on OS. Forest plot of univariate Cox regression analysis (B) and multivariate Cox regression analysis (D) on DFS. The preoperative CTC level is an independent prognostic factor for OS(p = 0.045) Subgroup analysis of the preoperative CTC level in lung cancer patients In our population, preoperative CTC levels didn’t interact with different groups of postoperative chemotherapy, TNM stage, sex, age, and tumor location (Fig. 5), indicating that, there was no statistical difference in the influence of preoperative CTC levels on OS/DFS among different subgroups of people, showing internal consistency. The predictive validity of CTC level as an independent prognostic indicator is not interfered with by those clinical variables. Fig. 5Subgroup analysis of the connection of CTC levels and clinicopathological variables on OS (A) and DFS (B). There was no statistical difference in the influence of preoperative CTC levels on OS/DFS among different subgroups of people Subgroup analysis of the connection of CTC levels and clinicopathological variables on OS (A) and DFS (B). There was no statistical difference in the influence of preoperative CTC levels on OS/DFS among different subgroups of people Prognostic value of multimetric combinations in NSCLC patients undergoing surgery Through the Cox stepwise regression analysis of factors affecting DFS, four clinical indicators were screened (Fig. 6A). The regression coefficients for all indicators were above zero, suggesting a positive relationship with the risk function, meaning an increased risk of poorer prognosis for the patient. We then created a standardized risk-scoring system based on regression coefficients and calculated the risk score for each patient. Preoperative CTC levels were converted to a binary format (low level = 0, high level = 1) and PI data were converted as follows: Yes = 1; No = 2. The risk score was calculated using the following formula: Risk score = (7.17 × CTC 0/1) + (1.88 × CYFRA21-1) + (0.34 ×MTD) + (10 × PI 0/1). The ROC experimental prognostic performance based on the risk scoring model for DFS in our group was 0.819 in 5 years, 0.867 in 3 years, and 0.937 in 1 year (Fig. 6B). Then patients were categorized into high-risk and low-risk groups based on the median risk score. A comparison of DFS rates in different risk groups was performed using Kaplan-Meier survival curves and log-rank tests to compare differences in DFS between risk groups (Fig. 6C). Patients identified with high-risk scores had less favorable outcomes than those with low-risk scores (P = 0.0052). These findings suggested that the risk score calculated based on preoperative CTC levels, CYFRA21-1, MTD, and PI demonstrated strong performance in prognostic prediction for NSCLC patients undergoing surgery. Multimodal prognostic prediction models combining hematology, imaging, and pathology have great potential for the guidance of clinical prognostic management. Fig. 6The standardized risk-scoring system based on multiple indicators The standardized risk-scoring system based on multiple indicators (A) Clinical indicators screened from Cox stepwise regression on DFS. (B) ROC diagnostic test. 1-year ROC curve has the highest AUC(0.937). (C) Kaplan-Meier survival curves on DFS. Patients were stratified into high-risk and low-risk groups based on the median risk score. The difference in DFS between the two groups was significant(p = 0.0052). Discussion NSCLC patients treated with radical intent should be closely monitored for treatment-related complications, detection of treatable relapse, or occurrence of metastasis [22]. The current routine means of postoperative follow-up review includes clinical consultations, physical examination, and chest CT scan [23]. However, these approaches have certain limitations. For instance, chest CT, which is often the preferred imaging modality for postoperative surveillance in lung cancer patients, raises concerns regarding the risks associated with repeated radiation exposure, especially during the first few years of intensive monitoring or in patients with advanced disease who require frequent scans [24, 25]. In contrast, tumor recurrence prediction using CTCs offers several advantages over traditional imaging. CTC-based monitoring can provide earlier warnings of recurrence compared to clinically observed imaging recurrence [26]. Additionally, CTC testing is minimally invasive, can be performed repeatedly and dynamically, and allows clinicians to continuously monitor tumor burden and treatment response. This facilitates the timely initiation of prospective adjuvant therapies, potentially improving patient outcomes and increasing the likelihood of a favorable prognosis. Despite the significant potential of CTCs in postoperative surveillance for NSCLC recurrence, research in this area remains limited. Most studies often rely on CTCs as a single biomarker, primarily demonstrating a correlation between CTC positivity and an elevated risk of postoperative recurrence [27]. However, relying solely on CTCs may not capture the full complexity of the disease. Integrating CTCs with additional biomarkers or clinical parameters could enhance predictive accuracy, enabling more precise risk stratification and paving the way for a more personalized treatment approach for NSCLC patients. Here, we firstly confirmed that there is a high correlation between preoperative CTC levels and postoperative prognosis. Previous clinical studies have shown that for all patients with clinically staged NSCLC, PFS, and OS were significantly higher in patients with a baseline CTC count of less than 5 than in patients with a baseline CTC count of greater than or equal to 5 [28]. In contrast, our study identified a higher threshold of 11 for preoperative CTC counts associated with poor prognosis, providing a more specific reference for populations with similar baseline characteristics to our study cohort. Using this cutoff point of 11 CTC units, we categorized patients into two groups based on preoperative CTC levels. Patients with high preoperative CTC levels showed worse 5-year OS(P = 0.014) and DFS(P = 0.023) rates, indicating that preoperative CTC levels may serve as a valuable prognostic marker. This finding aligns with the existing knowledge. For example, one study showed the same result that patients with elevated preoperative CTC levels had a significantly increased risk of postoperative recurrence and that the concentration of CTCs could be utilized as a preoperative serum tumor biomarker for detecting NSCLC metastasis [29]. The poorer 5-year OS and DFS observed in patients with high preoperative CTC levels could be attributed to the association of elevated CTC counts with increased tumor invasiveness and metastatic potential. CTCs enhance their metastatic capabilities through various mechanisms, such as homotypic clustering and heterotypic interactions with immune and stromal cells [30]. Consequently, patients with high CTC levels may harbor undetectable micrometastases, which could progress after surgery, leading to recurrence and metastasis. Furthermore, we evaluated the predictive efficacy of preoperative CTC units for different follow-up time points. The AUCs of preoperative CTC units for predicting OS were 0.857, 0.623, and 0.667 at 1, 3, and 5 years, respectively, and for predicting DFS were 0.857, 0.705, and 0.606 at 1, 3, and 5 years, respectively. Collectively, these findings suggest that preoperative CTCs exhibit satisfactory prognostic predictive efficiency for postoperative NSCLC patients, with variations in efficiency depending on the follow-up duration. Predictive efficacy was highest for 1-year OS and DFS but declined for the 3- and 5-year time points. This trend may be attributed to the fact that, during short-term follow-up, preoperative CTC levels effectively reflect recent tumor dynamics, facilitating early detection of recurrence and metastasis. However, with longer follow-ups, preoperative CTCs do not carry enough information to reflect the impact of multiple factors in the postoperative period, including treatment response, tumor heterogeneity, and the emergence of new metastases. Therefore, the predictive efficiency of CTCs may fluctuate during long-term follow-up and should be interpreted in conjunction with other biomarkers and imaging studies [31]. Accordingly, We constructed Cox regression models incorporating CTCs and other influences related to prognosis. Firstly, the importance of CTCs was reconfirmed through Cox analyses. In multivariate Cox analysis, preoperative CTC levels correlate with poor prognosis second only to pleural infiltration and higher than TNM staging according to HR (Hazard Ratio) which indicated that CTCs might make up for the shortcomings of the TNM stage as a powerful prognostic predictor in the future [32]. In this cohort, the preoperative CTC level was an independent prognostic factor for OS. This finding is in line with the results of a study by LiZ, et al. [29], who also highlighted preoperative CTC concentrations as sensitive and independent prognostic biomarkers for NSCLC patients. Moreover, their study demonstrated that a nomogram based on preoperative CTC concentrations is an effective and non-invasive tool for predicting NSCLC recurrence and metastasis. In clinical practice, factors such as postoperative chemotherapy, TNM stage, gender, age, and tumor location are usually considered to have a significant impact on patient prognosis [33]. However, our study found that these factors did not significantly affect the predictive ability of preoperative CTC levels for OS and DFS. This implies that the predictive validity of CTC level as an independent prognostic indicator is not interfered with by other clinical variables. Additionally, univariate Cox analysis revealed that TNM stage, CEA, CYFRA21-1, CA125, MTD, and pleural infiltration were all significantly associated with OS and DFS, providing further evidence that multiple factors contribute to poor prognosis. This highlights the potential to integrate related factors including hematological, radiological, and pathological data with preoperative CTC levels to enhance prognostic assessment in postoperative NSCLC patients. Diagnostic models incorporating both CTCs and serum biomarkers have been shown to enhance the efficiency of distinguishing between different histological types of lung cancer and benign space-occupying pulmonary diseases [34]. However, fewer studies have focused on the prognosis of lung cancer patients undergoing surgery. In this study, we developed a standardized risk-scoring system based on regression coefficients and calculated individual risk scores for each patient. The selection of indicators for model indicators also considered their findings of prognostic significance in the available literature for patients with different pathology types of NSCLC. Each included indicator contributed to enhancing the model’s predictive power when applied to this cohort. For example, CYFRA21-1, traditionally associated with squamous cell carcinoma, also demonstrated good predictive efficacy in our predominantly adenocarcinoma patient cohort. By integrating preoperative CTC levels, CYFRA21-1, PI, and measurement of MTD after surgery, the model compensated for the limitations of using a single indicator and provided a reference for identifying high-risk patients. The prognostic performance of the risk scores for DFS in our group was 0.819 in 5 years, 0.867 in 3 years, and 0.937 in 1 year. In addition, patients identified with high-risk scores had less favorable outcomes compared to those with low-risk scores significantly (P = 0.0052). The combined model demonstrated superior prognostic predictive efficacy and more pronounced prognostic differences compared to using preoperative CTC levels alone. These findings suggested that the risk score calculated derived from the integration of preoperative CTCs, tumor markers, imaging, and pathological features demonstrated strong performance in prognostic prediction for NSCLC patients undergoing surgery. Overall, multimodal prognostic prediction models combining hematology, imaging, and pathology have great potential for clinical applications. This multimodal approach allows for better risk stratification and can guide therapeutic decisions, potentially leading to improved patient outcomes. The limitations of this study include the relatively small sample size, which may have reduced its statistical power and limited the ability to detect certain associations, particularly in subgroup analyses. Additionally, the small sample size may affect the generalizability of the findings to broader populations. Therefore, while our findings provide valuable preliminary insights, further studies with larger sample sizes are needed to validate our results and improve their applicability to larger populations. This study was conducted as a single-center retrospective analysis and may not fully represent the diversity of NSCLC patients or treatment strategies present in different regions or healthcare systems. To improve the generalizability of these findings, future studies could employ a multicenter, prospective design, which would allow for a more diverse patient population and a more standardized data collection process. Another limitation is that single-center cross-sectional studies often have smaller cohort sizes and longer follow-up times, which can lead to patient dropout. Furthermore, this study didn’t assess the dynamic outcomes of CTCs at multiple follow-up time points after surgery. Dynamic monitoring of CTCs could enable more accurate determination of recurrence timing in NSCLC patients and facilitate earlier therapeutic intervention, thereby advancing the treatment time window. In addition to dynamic monitoring, future studies could also focus on genomic variants in CTCs. Genomic instability, a key driver of cancer progression and metastasis, represents an important area of investigation. By examining the impact of genomic instability carried by CTCs, potential mechanisms underlying tumor metastasis and recurrence may be uncovered. Despite these limitations, our study provides an important foundation for understanding the prognostic value of CTCs combined with multiple indicators in NSCLC patients. It highlights key areas for future research, including dynamic monitoring and genomic studies, to refine prognostic prediction and improve patient outcomes. Conclusions In conclusion, this study demonstrated that preoperative CTC levels serve as a feasible and effective biomarker for predicting postoperative prognosis in NSCLC patients. The integration of preoperative CTC levels with tumor markers, imaging data, and pathological features provides a powerful prognostic tool for predicting lung cancer outcomes. This integrated approach enhances the precision of prognosis, potentially leading to better patient management and personalized treatment strategies. Our findings underscore the importance of a multimodal prognostic approach in the clinical evaluation of lung cancer. Future clinical trials should focus on the genomic instability of CTCs to uncover the mechanisms driving tumor metastasis and recurrence dynamics, further enhancing the understanding of cancer progression and refining therapeutic approaches.