Integration of circulating biomarkers and clinical factors: construction and validation of a prediction model for lung cancer metastasis Authors: Wenlong Qi, Zhenyu Li, Jianan Xu, Tan Wang Published: 2025 ABSTRACT ObjectiveTo develop and validate a prediction model for metastasis risk in lung cancer patients based on circulating biomarkers and clinical factors, thereby facilitating early risk assessment.MethodsA total of 511 lung cancer patients who received treatment in the hospital from January 2020 to December 2024 were selected. Their clinical data and laboratory test indicators were collected and divided into a training set (n = 358) and a validation set (n = 153) at a ratio of 7:3. In the training set, risk factors were screened by univariate and multivariate Logistic regression to construct a nomogram model. The receiver operating characteristic curve (ROC) and calibration curve were drawn to evaluate the model’s efficacy, and the model was validated in the validation set. Decision curve analysis (DCA) was used to evaluate the clinical value.ResultsIn the training set, 143 cases (39.94%) had lung cancer metastasis, and in the validation set, 61 cases (39.87%) had lung cancer metastasis. Multivariate Logistic regression showed that lymph node status, Total Prostate-Specific Antigen (TPSA), Carcinoembryonic Antigen (CEA), tumor size, Carbohydrate Antigen 19 − 9(CA199) and Alpha-Fetoprotein were significantly associated with the risk of lung cancer metastasis (all P < 0.05). The nomogram demonstrated consistent performance across the training and validation sets, with C-indices of 0.714 and 0.710, and AUCs of 0.714 (95% CI: 0.649–0.778), with a sensitivity of 0.571 and a specificity of 0.745 and 0.710 (95% CI: 0.609–0.812), with a sensitivity of 0.622 and a specificity of 0.629, respectively. The P values of the Hosmer - Lemeshow test were 0.183 and 0.075, indicating a good model fit, respectively.ConclusionThe nomogram model constructed based on circulating biomarkers and clinical factors can effectively predict the metastasis risk of lung cancer patients and has certain clinical application value. However, multi - center and large - sample studies are still needed for further validation. Introduction Lung cancer was one of the most common and deadly malignancies globally. It was characterized by high invasiveness and poor prognosis, posing a serious threat to human health. Tumor metastasis was the primary cause of treatment failure and reduced survival rates in patients [1, 2].Statistics showed that approximately 30% − 50% of lung cancer patients have developed distant metastases at the time of diagnosis. At this stage, clinical treatment options are limited, and the 5-year survival rate decreases significantl [3, 4]. Therefore, early and accurate identification of high-risk populations for lung cancer metastasis was of great clinical significance for optimizing treatment strategies and improving patient prognosis. Currently, traditional clinical factors such as tumor size, clinical stage, and lymph node status, as well as imaging examinations (e.g., Computed Tomography (CT), Magnetic Resonance Imaging (MRI)), were commonly used in clinical practice to assess the risk of lung cancer metastasis [5, 6]. However, these methods have certain limitations. The predictive efficacy of traditional clinical factors alone was limited, and although imaging examinations can detect macroscopic metastases, they were difficult to provide early warnings of metastatic risks at the sub - clinical stage. In recent years, circulating biomarkers have attracted wide attention in the diagnosis, efficacy evaluation, and prognosis judgment of lung cancer due to their advantages of easy detection and dynamic monitoring [7–9]. Studies have shown that the abnormal expression of multiple circulating biomarkers is closely related to the progression and metastasis of lung cancer. However, the sensitivity and specificity of a single biomarker are insufficient to meet the clinical needs for accurate prediction. Many existing prediction models were limited by their reliance on a single type of indicator—either clinical imaging features [10] or individual circulating biomarkers [11], which limits their predictive accuracy. Furthermore, there is a lack of models that systematically integrate multi-dimensional information, particularly the combination of circulating biomarkers with key clinical factors, to leverage their synergistic predictive potential. To improve the accuracy of metastasis risk prediction, researchers are increasingly constructing models that integrate multi-dimensional information. The nomogram, as a visual risk assessment tool, can integrate multiple independent risk factors into individualized risk probabilities. It is intuitive and easy to use, and has been applied in the prognosis prediction of various malignancies. There was a growing recognition in oncology of the prognostic value inherent in multidimensional biomarker models. This trend was exemplified by recent research into inflammation- and nutrition-related indicators. Serum calcium ion levels have been shown to predict outcomes in colorectal cancer patients [12], while composite indicators such as the Onodera prognostic nutritional index, pan-immuno-inflammatory values, and the albumin-to-globulin ratio have demonstrated significant predictive value for prognosis and complications in gastrointestinal malignancies [13, 14]. Furthermore, novel prognostic scoring systems that combine markers like the platelet-albumin ratio with cancer inflammation indices were being developed to refine risk stratification in colorectal cancer cohorts [15, 16]. These advances highlighted a paradigm shift towards leveraging integrated, multi-parameter data for enhanced clinical prediction. Based on this, this study aims to retrospectively analyze the clinical data and circulating biomarker data of lung cancer patients, screen the independent risk factors affecting lung cancer metastasis, construct a nomogram prediction model integrating circulating biomarkers and clinical factors (such as tumor size, lymph node status, etc.), and evaluate its predictive efficacy through internal and external validation. It is expected to provide a reliable tool for the early clinical identification of high - risk patients with lung cancer metastasis and offer a reference for individualized treatment decisions. Materials and methods Study subjects A total of 511 lung cancer patients who received treatment at the hospital from January 2020 to December 2024 were selected as the research subjects. Inclusion criteria: meeting the diagnostic criteria for lung cancer (confirmed by pathological histology or cytology) [17, 18], aged between 18 and 80 years. Exclusion criteria: having other malignancies, severe liver or kidney insufficiency, incomplete clinical data, or being unable to cooperate with the follow - up. The patients were randomly divided into a training set (n = 358) and a validation set (n = 153) at a ratio of 7:3 using the complete randomization method via the ‘caret’ package in R software. All patients were followed up for at least 12 months to monitor for the occurrence of metastasis, thereby avoiding model overfitting caused by data distribution bias. This study was approved by the hospital’s ethics committee, and all patients signed informed consent forms. Data collection General clinical data of the patients were collected, including age, body mass index (BMI), gender, smoking history, tumor size, tumor differentiation degree, clinical stage and lymph node status. Fasting venous blood was collected from all patients within 24 h after admission to detect the levels of circulating biomarkers, including Carcinoembryonic Antigen (CEA) (ng/mL), Total Prostate-Specific Antigen (TPSA) (ng/mL), Free Prostate-Specific Antigen (ng/mL), Carbohydrate Antigen 19 − 9(CA199) (U/mL) and Alpha-Fetoprotein (ng/mL). The detection methods were strictly operated according to the instructions of the relevant kits. Diagnostic criteria The diagnosis of lung cancer metastasis was based on the criteria from the “NCCN Clinical Practice Guidelines in Oncology”. A metastatic lesion was definitively diagnosed by the presence of a new, histologically confirmed malignant lesion in a distant organ or non-regional lymph node, or, when biopsy was not feasible, by conclusive evidence on imaging (e.g., chest CT, abdominal CT, bone scan, head MRI) demonstrating typical characteristics of metastatic disease that was confirmed by multidisciplinary team discussion and radiographic progression on follow-up [19]. Statistical analysis Statistical analysis was performed using SPSS 26.0 and R 4.3.1 software. Normally distributed measurement data are presented as mean ± standard deviation and compared between groups using the independent samples t-test. M (Q1, Q3) is used when it does not conform to the normal distribution, and Mann-Whitney U test is used for comparison between groups. The counting data were compared by χ2 test or Fisher exact method. And count data were expressed as the number of cases. Multivariate Logistic regression analysis was used to screen risk factors, and a P value < 0.05 was considered statistically significant, the results were expressed as odds ratios (ORs) and 95% confidence intervals (CIs). The research subjects were randomly divided into the training set and validation set at a 7:3 ratio using the ‘caret’ package of R software. The “rms” package in R language was used to construct the nomogram model. The Bootstrap method (with 1000 repeated samplings) was used for internal validation of the model, and a calibration curve of the predicted results and the actual results was drawn. The concordance index (C - index) of the model was calculated, and the Hosmer - Lemeshow test was used to evaluate the goodness - of - fit of the prediction model. The discriminatory power was assessed by analyzing the receiver operating characteristic (ROC) curve and calculating the area under the curve (AUC). Decision curve analysis (DCA) was used to evaluate the clinical application value of the model. Significance testing was performed with the use of a two-sided alpha level of 0.05. Results Baseline characteristics in the training set and the validation set In the training set, 143 cases (39.94%) had lung cancer metastasis, and in the validation set, 61 cases (39.87%) had lung cancer metastasis. There were no statistically significant differences in general clinical characteristics such as age, gender, BMI, and most laboratory indicators between the two groups (all P > 0.05) (Table 1). Table 1Comparison of general clinical characteristics between patients in the training set and the validation setIndicatorsTraining set (n = 358)Validation set (n = 153)t/χ² P Age (years)56.33 ± 10.4957.16 ± 10.730.8140.416BMI21.67 ± 2.2021.38 ± 2.311.3440.179SexMale187(52.23)81(52.94)0.0220.883Female171(47.77)72(47.06)Smoking historyYes239(66.76)92(60.13)2.0640.151No119(33.24)61(39.87)Histological typeAdenocarcinoma216(60.34)88(57.52)0.5210.771Squamous cell carcinoma101(28.21)48(31.37)Small cell lung cancer41(11.45)17(11.11)Tumor differentiation degreeHigh126(35.20)50(32.68)1.6330.442Moderate146(40.76)58(37.91)Low86(24.02)45(29.41)Clinical stageⅠ-Ⅱ135(37.71)61(39.87)0.2120.646Ⅲ-Ⅳ223(62.29)92(60.13)Lymph node statusPositive155(54.47)74(48.37)1.1140.291Negative203(45.53)79(51.63)Metastasis statusWith metastasis143(39.94)61(39.87)0.0010.986Without metastasis215(60.06)92(60.13)TPSA(ng/mL)13.83 ± 5.2914.02 ± 5.150.3750.708CEA(ng/mL)5.69 ± 2.105.42 ± 2.151.3220.187Tumor size (cm)2.36 ± 0.812.43 ± 0.850.8820.379FPSA(ng/mL)2.14 ± 0.282.16 ± 0.310.6210.535CA199(U/mL)227.20 ± 71.05231.62 ± 72.160.6410.522AFP(ng/mL)119.07 ± 31.36122.15 ± 30.721.0230.307%fPSA14.85 ± 2.1815.02 ± 2.620.6420.521 Table 1Comparison of general clinical characteristics between patients in the training set and the validation setIndicatorsTraining set (n = 358)Validation set (n = 153)t/χ² P Age (years)56.33 ± 10.4957.16 ± 10.730.8140.416BMI21.67 ± 2.2021.38 ± 2.311.3440.179SexMale187(52.23)81(52.94)0.0220.883Female171(47.77)72(47.06)Smoking historyYes239(66.76)92(60.13)2.0640.151No119(33.24)61(39.87)Histological typeAdenocarcinoma216(60.34)88(57.52)0.5210.771Squamous cell carcinoma101(28.21)48(31.37)Small cell lung cancer41(11.45)17(11.11)Tumor differentiation degreeHigh126(35.20)50(32.68)1.6330.442Moderate146(40.76)58(37.91)Low86(24.02)45(29.41)Clinical stageⅠ-Ⅱ135(37.71)61(39.87)0.2120.646Ⅲ-Ⅳ223(62.29)92(60.13)Lymph node statusPositive155(54.47)74(48.37)1.1140.291Negative203(45.53)79(51.63)Metastasis statusWith metastasis143(39.94)61(39.87)0.0010.986Without metastasis215(60.06)92(60.13)TPSA(ng/mL)13.83 ± 5.2914.02 ± 5.150.3750.708CEA(ng/mL)5.69 ± 2.105.42 ± 2.151.3220.187Tumor size (cm)2.36 ± 0.812.43 ± 0.850.8820.379FPSA(ng/mL)2.14 ± 0.282.16 ± 0.310.6210.535CA199(U/mL)227.20 ± 71.05231.62 ± 72.160.6410.522AFP(ng/mL)119.07 ± 31.36122.15 ± 30.721.0230.307%fPSA14.85 ± 2.1815.02 ± 2.620.6420.521 Comparison of general clinical characteristics between patients in the training set and the validation set Univariate analysis in the training set The results of univariate analysis showed that there were statistically significant differences in indicators such as lymph node status, TPSA, CEA, Tumor size, CA199 and AFP between patients with metastasis and those without metastasis (all P < 0.05) (Table 2). Table 2Univariate analysis of influencing factors for metastasis in lung cancer patients in the training setIndicatorsWith metastasis (n = 143)Without metastasis (n = 215)t/χ² P Age (years)57.15 ± 10.6255.79 ± 10.381.2030.230BMI21.45 ± 2.1521.82 ± 2.241.5470.123SexMale73(51.05)114(53.02)1.1560.282Female70(48.95)101(46.98)Smoking historyYes91(63.64)148(68.84)0.7850.376No52(36.36)67(31.16)Tumor differentiation degreeHigh46(32.17)80(37.21)5.9400.051Moderate53(37.06)93(43.26)Low44(30.77)42(19.53)Clinical stageⅠ-Ⅱ46(32.17)89(41.40)3.1130.078Ⅲ-Ⅳ97(67.83)126(58.60)Lymph node statusPositive84(58.74)71(33.02)23.1370.001Negative59(41.26)144(66.98)TPSA(ng/mL)14.85 ± 5.4713.15 ± 5.083.0070.003CEA (ng/mL)6.02 ± 2.135.46 ± 2.072.4780.014Tumor size (cm)2.51 ± 0.812.26 ± 0.802.8820.004FPSA(ng/mL)2.15 ± 0.272.14 ± 0.260.3510.726CA199(U/mL)238.58 ± 73.18219.62 ± 68.732.4910.013AFP(ng/mL)124.76 ± 32.47115.29 ± 30.082.8260.005%fPSA14.58 ± 2.1815.08 ± 2.571.9130.057 Table 2Univariate analysis of influencing factors for metastasis in lung cancer patients in the training setIndicatorsWith metastasis (n = 143)Without metastasis (n = 215)t/χ² P Age (years)57.15 ± 10.6255.79 ± 10.381.2030.230BMI21.45 ± 2.1521.82 ± 2.241.5470.123SexMale73(51.05)114(53.02)1.1560.282Female70(48.95)101(46.98)Smoking historyYes91(63.64)148(68.84)0.7850.376No52(36.36)67(31.16)Tumor differentiation degreeHigh46(32.17)80(37.21)5.9400.051Moderate53(37.06)93(43.26)Low44(30.77)42(19.53)Clinical stageⅠ-Ⅱ46(32.17)89(41.40)3.1130.078Ⅲ-Ⅳ97(67.83)126(58.60)Lymph node statusPositive84(58.74)71(33.02)23.1370.001Negative59(41.26)144(66.98)TPSA(ng/mL)14.85 ± 5.4713.15 ± 5.083.0070.003CEA (ng/mL)6.02 ± 2.135.46 ± 2.072.4780.014Tumor size (cm)2.51 ± 0.812.26 ± 0.802.8820.004FPSA(ng/mL)2.15 ± 0.272.14 ± 0.260.3510.726CA199(U/mL)238.58 ± 73.18219.62 ± 68.732.4910.013AFP(ng/mL)124.76 ± 32.47115.29 ± 30.082.8260.005%fPSA14.58 ± 2.1815.08 ± 2.571.9130.057 Univariate analysis of influencing factors for metastasis in lung cancer patients in the training set Multivariate logistic regression analysis in the training set Multivariate Logistic regression analysis was performed with metastasis occurrence as the dependent variable and the factors with P < 0.05 in the univariate analysis as independent variables. The results showed that lymph node status, TPSA, CEA, Tumor size, CA199 and AFP were significantly associated with the risk of lung cancer metastasis (all P < 0.05) (Table 3). Table 3Multivariate logistic regression analysisIndicatorsβSEWald P OR95%CILymph node status1.1520.24122.7780.0013.1651.972–5.080TPSA0.0460.0233.9630.0471.0471.001–1.095CEA0.1330.0565.5980.0181.1431.023–1.276Tumor size0.3540.1505.5330.0191.4251.061–1.913CA1990.0040.0025.5170.0191.0041.001–1.007AFP0.0100.0046.9850.0081.0101.003–1.018 Table 3Multivariate logistic regression analysisIndicatorsβSEWald P OR95%CILymph node status1.1520.24122.7780.0013.1651.972–5.080TPSA0.0460.0233.9630.0471.0471.001–1.095CEA0.1330.0565.5980.0181.1431.023–1.276Tumor size0.3540.1505.5330.0191.4251.061–1.913CA1990.0040.0025.5170.0191.0041.001–1.007AFP0.0100.0046.9850.0081.0101.003–1.018 Multivariate logistic regression analysis Construction of the nomogram prediction model Based on the independent risk factors screened by the multivariate Logistic regression analysis, a nomogram prediction model for lung cancer metastasis was constructed (Fig. 1). Using this nomogram, first, assign corresponding scores to each indicator based on their specific test values, then sum up the scores of all indicators to obtain a total score, and finally determine the risk probability of lung cancer metastasis through the corresponding total score. It provides an intuitive and quantitative tool for clinical evaluation of the possibility of lung cancer metastasi. Fig. 1Nomogram prediction model for lung cancer metastasis Nomogram prediction model for lung cancer metastasis Evaluation and validation of the nomogram prediction model In the training set, the C - index of the nomogram model was 0.714, and the P value of the Hosmer - Lemeshow test was 0.183, indicating a good fit of the model (Fig. 2). The ROC curve showed that the AUC for predicting lung cancer metastasis was 0.714 (95%CI: 0.649–0.778), with a sensitivity of 0.571 and a specificity of 0.745 (Fig. 3). In the validation set, the C - index of the model was 0.710, the P value of the Hosmer - Lemeshow test was 0.075, the AUC was 0.710 (95%CI: 0.609–0.812), with a sensitivity of 0.622 and a specificity of 0.629. Fig. 2Calibration curves of nomogram model for predicting lung cancer metastasis (A: the training set, B: the validation set. The 45°dotted line represents the ideal prediction, and the solid line indicates the nomogram’s performance. Closer fit to the diagonal suggests better calibration) Calibration curves of nomogram model for predicting lung cancer metastasis (A: the training set, B: the validation set. The 45°dotted line represents the ideal prediction, and the solid line indicates the nomogram’s performance. Closer fit to the diagonal suggests better calibration) Fig. 3Receiver operating characteristic curves of the nomogram model for predicting lung cancer metastasis (A: the training set, B: the validation set. AUC: Area under the curve, CI: Confidence interval) Receiver operating characteristic curves of the nomogram model for predicting lung cancer metastasis (A: the training set, B: the validation set. AUC: Area under the curve, CI: Confidence interval) Decision curve analysis of the nomogram prediction model he DCA results showed that when the threshold probability was between 0.15 and 0.85, the net benefit of applying the nomogram model constructed in this study to predict lung cancer metastasis was higher than that of the assumption that all patients had metastasis or no metastasis, indicating that the model had certain clinical application value (Fig. 4). Fig. 4Decision curves of the nomogram model for predicting lung cancer metastasis (A: the training set, B: the validation set. The solid blue line represents the nomogram model, the grey line represents the ‘treat-all’ strategy, the solid black line represents the ‘treat-none’ strategy, the x-axis represents the threshold probability, while the y-axis represents the net clinical benefit) Decision curves of the nomogram model for predicting lung cancer metastasis (A: the training set, B: the validation set. The solid blue line represents the nomogram model, the grey line represents the ‘treat-all’ strategy, the solid black line represents the ‘treat-none’ strategy, the x-axis represents the threshold probability, while the y-axis represents the net clinical benefit) Discussion In this study, a nomogram prediction model integrating lymph node status and circulating biomarkers such as lymph node status, TPSA, CEA, Tumor size, CA199 and AFP was constructed through a retrospective analysis of the clinical data of 511 lung cancer patients. The results showed that the C - indices of the model in the training set and the validation set were 0.714 and 0.710 respectively, and the AUC values were 0.714 (95%CI: 0.649–0.778) and 0.710 (95%CI: 0.609–0.812) respectively. The Hosmer - Lemeshow test indicated a good fit of the model, and the decision curve analysis confirmed its net benefit advantage within the threshold probability range of 0.15–0.85. Multivariate Logistic regression identified lymph node status, TPSA, CEA, Tumor size, CA199 and AFP as independent risk factors for lung cancer metastasis, suggesting that the above - mentioned indicators may affect disease progression through synergistic effects during the tumor metastasis process, and providing a reliable basis for the early clinical identification of metastasis risk. In this study, lymph node status, as the core indicator of clinicopathological features, showed the highest OR value (3.165, 95%CI: 1.972–5.080) in the multivariate analysis, indicating its particularly prominent predictive value for lung cancer metastasis. Mechanistically, lymph nodes are the primary target organs for tumor cells to spread through the lymphatic circulation. A positive lymph node status not only reflects that tumor cells have broken through the local barrier but also suggests their invasive metastasis potential [20]. Clinical studies have confirmed that the number of lymph node metastases in lung cancer patients is positively correlated with the risk of distant metastasis, and the degree of lymph node involvement is an important basis for stage upgrading, which is consistent with the result of lymph node status as an independent risk factor in this study [21]. In addition, the early detection of lymph node micrometastasis can significantly improve the accuracy of metastasis risk assessment. By including lymph node status in the model, this study further strengthened the association between clinicopathological features and metastasis risk, providing a key reference for preoperative staging and treatment plan formulation. Although TPSA is conventionally associated with prostate cancer, it demonstrated significant predictive value in our study (OR = 1.047), suggesting a potential role in lung cancer through participation in tumor microenvironment regulation. TPSA may promote tumor cell invasion into surrounding tissues by influencing the activity of extracellular matrix-degrading enzymes, while potentially enhancing the survival capacity of circulating tumor cells [22, 23]. In this study, elevated TPSA levels correlated with an increased risk of lung cancer metastasis, indicating its potential as a latent alert indicator for metastasis. It is particularly suitable for use in combination with other biomarkers to improve predictive power, thereby offering a novel detection target for clinical practice. As a broad - spectrum tumor marker, CEA showed significant predictive value in this study (OR = 1.143, P = 0.018) [24]. Its mechanism may be related to the epithelial - mesenchymal transition (EMT) process of tumor cells: CEA promotes tumor cells to detach from the primary tumor and invade the surrounding tissues by participating in the regulation of intercellular adhesion molecules, and at the same time enhances their survival ability in the circulatory system [25, 26]. Previous studies have shown that an elevated CEA level is closely related to pleural metastasis and liver metastasis in lung cancer patients, and dynamic monitoring of CEA changes can predict treatment response and recurrence risk [27, 28]. The independent predictive role of CEA in this study further supports its clinical significance in the risk stratification of lung cancer metastasis, especially for patients who cannot undergo invasive examinations. Tumor size, a readily measurable clinical characteristic, was an independent risk factor (OR = 1.425). Larger tumors are associated with increased cellular proliferation and local invasion, which can breach anatomical barriers and create the foundational conditions for metastasis [29]. Additionally, larger tumors typically carry higher mutational burdens, potentially accumulating more metastasis-associated genetic alterations (e.g., activation of invasion-related genes) [30]. Clinical data consistently demonstrate a positive correlation between primary lung tumor size and risks of lymph node/distant metastasis. Our findings align with this established pattern, indicating that clinical assessment should incorporate tumor size to adjust follow-up frequency—intensifying metastasis surveillance for patients with larger tumors to enable early intervention. CA199 demonstrated predictive significance in this study (OR = 1.004, P = 0.019). Although primarily associated with gastrointestinal tumors, CA199 may influence lung cancer metastasis through tumor metabolic regulation: its elevated levels may reflect aberrant glucose metabolism in cancer cells. This metabolic dysregulation can promote matrix metalloproteinase activation by altering the tumor microenvironment (e.g., acidification), thereby enhancing invasive capacity [31–33]. Research suggests lung cancer patients with elevated CA199 are more prone to liver metastasis—potentially attributable to CA199’s direct impact on the hepatic microenvironment via systemic circulation [34]. Incorporating CA199 into our prediction model supplements the metabolic assessment dimension, providing additional evidence for identifying high-risk metastasis patients. AFP, traditionally associated with hepatocellular carcinoma, was identified as an independent risk factor in this study (OR = 1.010, P = 0.008). Its involvement in lung cancer metastasis may relate to tumor cell dedifferentiation: malignant cells can develop aberrant phenotypes during progression, gaining AFP-secreting capabilities [35, 36]. Furthermore, AFP may facilitate “escape” of circulating tumor cells by suppressing host immune responses—notably through inhibiting T-cell activity—creating permissive conditions for metastasis [37]. Our results linking elevated AFP with metastatic risk suggest a distinct molecular phenotype in certain lung cancer patients. AFP thus serves as an alert indicator for metastasis risk within this subgroup, with improved predictive comprehensiveness when combined with other biomarkers. Beyond the biological plausibility of individual factors, the primary translational value of this integrated nomogram model lies in its potential for clinical risk stratification and personalized management. By providing a quantifiable probability of metastasis, the model can help clinicians identify high-risk patients who may benefit from more intensive surveillance protocols (e.g., shorter interval imaging or inclusion of advanced imaging techniques) even in the absence of overt clinical signs. Conversely, patients identified as low-risk could be spared unnecessary procedures, optimizing healthcare resource allocation. Furthermore, this model could serve as a decision-support tool when considering adjuvant therapy, where estimating the risk of occult metastasis is crucial. Its visual and quantitative nature makes it particularly suitable for application in primary care and oncology clinics to facilitate shared decision-making between doctors and patients. This study has several advantages. Firstly, it is the first to systematically integrate lymph node status with biomarkers into a prediction model, thereby combining clinical accessibility with predictive accuracy. Compared with single indicators or traditional clinical factors, multi - dimensional information significantly improves the discriminatory efficacy of the model. Secondly, a nomogram visualization tool is used to convert the complex regression equation into an individualized risk probability. Clinicians can quickly obtain the patient’s metastasis risk by simply referring to the table, which is especially suitable for primary medical institutions. In addition, the model showed good consistency in the training set and the validation set (C - index close to 0.71), confirming its stability and providing a basis for clinical promotion. However, the study still has some limitations. Firstly, the single - center retrospective design may introduce selection bias. The homogeneity of patient baseline characteristics (such as treatment plans and follow - up strategies) may affect the extrapolation of the results. Secondly, external validation was not performed, mainly because the research time span was relatively short, the data sharing mechanism for multi - center collaboration was not yet perfect, and there were differences in detection methods among different centers, which may affect the consistency of indicators. This is crucial for confirming the robustness and applicability of the model in different populations and clinical settings. To address this issue, we propose a clear future validation strategy: (1) Establish cooperative agreements with multiple independent medical centers and create an external validation cohort using a standardized data collection protocol. (2) Design a prospective study to evaluate the performance of the model in real - world clinical workflows, which will ultimately test its true clinical utility. Finally, the model did not include molecular biological indicators (such as EGFR and ALK mutation status) and imaging features (such as tumor enhancement patterns). These factors may affect the prediction efficacy by regulating metastasis - related pathways and need to be further supplemented in the future. In conclusion, the nomogram model constructed in this study provides a reliable tool for predicting the metastasis risk of lung cancer by integrating lymph node status and circulating biomarkers. The mechanism analysis of the six independent risk factors provides a theoretical basis for understanding the metastasis process. In the future, multi - center, large - sample studies are needed to verify the efficacy of the model, and more multi - dimensional indicators should be included to improve its accuracy, ultimately achieving early warning and individualized intervention for lung cancer metastasis.