LASSO-based nomograms predict early death in small cell lung cancer (SCLC) patients with brain metastasis

Introduction

Small cell lung cancer (SCLC) is characterized by a poor prognosis, early dissemination, and eventual metastasis. It is estimated that 2 million people are diagnosed with lung cancer annually, of whom 15% have SCLC (1). Concurrently, brain metastasis (BM) represents the most common malignancy-related condition in the central nervous system (2,3). The probability of synchronous BM at initial SCLC diagnosis is high (10–20%), and up to 50% of patients develop BM after initial diagnosis (4). Small cell lung cancer with brain metastasis (SCLC-BM) has a dismal prognosis characterized by elevated morbidity and mortality (5).

Research is increasingly focusing on therapeutic advances and survival outcomes in SCLC (6-8); however, few studies have investigated the clinical characteristics, prognoses, and relevant prognostic indicators of SCLC-BM specifically. Previous studies suggest that epidemiological and clinical features, including sex, race, age, tumor-node-metastasis (TNM) stage, surgery, chemotherapy, and radiotherapy, are correlated with SCLC prognosis (6-8). However, it remains unclear whether these findings apply to SCLC-BM patients.

Early death (ED) is defined as mortality within 3 months of diagnosis. This period represents the most critical prognostic window for SCLC-BM patients, necessitating the prompt identification of and implementation of immediate clinical intervention for high-risk patients. Understanding the factors that influence ED could inform aggressive treatment, clinical trials, and supportive care and elucidate relationships between tumor-related variables and early mortality (9,10). To date, few studies have examined ED (i.e., death within 3 months) in SCLC-BM patients, and have instead primarily focused on long-term prognostic factors. Thus, investigating the predictors of ED and developing nomograms for this population is critical to guide clinical management and follow-up.

In this study, we extracted data from the Surveillance, Epidemiology, and End Results (SEER) database (2010–2015) for patients diagnosed with SCLC-BM. Using univariate and least absolute shrinkage and selection operator (LASSO) regression to mitigate collinearity and overfitting, we identified the minimal prognostic factors. Using these factors, we constructed two nomograms to predict all-cause early death (ACED) and cancer-specific early death (CSED). We present this article in accordance with the Transparent Reporting of a multivariable prediction model for Individual Prognosis Or Diagnosis (TRIPOD) reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1269/rc).

Methods

Data gathering

As a population-based cancer registry covering approximately 30% of the United States population, SEER constitutes a population-based cohort in which each cancer case is treated as an independent observation in standard epidemiological analyses (11,12). Data from SEER served as the foundation for this retrospective analysis. This study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments. As SEER is an open public database, approval by an ethics committee was not required for this study.

Data collation

SCLC-BM patients diagnosed between 2010–2015 were selected to ensure the uniform application of the 7th edition of the American Joint Committee on Cancer (AJCC) TNM staging system and consistent treatment protocols during the pre-immunotherapy era. Patients were included in the study if they met the following inclusion criteria: (I) SCLC-BM; and (II) International Classification of Diseases for Oncology, Third Edition (ICD-O-3) pathological type 8002, 8041, 8042, 8043, 8044, or 8045, with site codes C34.0–C34.3, or C34.8–C34. Patients were excluded from the study if they met any of the following exclusion criteria: (I) Tx (primary tumor cannot be assessed) or Nx (regional lymph nodes cannot be assessed); (II) unknown race; (III) unknown tumor size; (IV) unknown bone metastasis, lung metastasis, or liver metastasis status; (V) unknown grade; (VI) unknown marital status at diagnosis; (VII) multiple primary tumors; and/or (VIII) missing or unknown cause of death.

The following clinical information was included: sex, age, chemotherapy, race, marital status, tumor size, T stage, N stage, radiotherapy, bone metastasis, liver metastasis, lung metastasis, median household income, primary site, and grade. Based on previous studies, the enrolled SCLC-BM patients were randomly divided (7:3) into training (n=443) and test cohorts (n=190) (13). It should be noted that this randomization was performed solely for model development purposes (training vs. test cohorts), and did not constitute intervention randomization as would be found in randomized controlled trials. The months of survival and the cause of death were provided as follow-up information.

ACED and CSED were the study outcomes; ED was defined as mortality within three months of the first diagnosis. It should be noted that cause-of-death information in SEER is derived from death certificates, which may contain inaccuracies in distinguishing cancer-specific from non-cancer deaths, particularly those related to treatment complications. This potential misclassification bias affects the reliability of CSED as an endpoint. The three-month cutoff was specifically chosen to identify patients facing an imminent mortality risk, enabling rapid clinical decision-making for urgent supportive care and palliative interventions. The screening process for the study subjects is shown in Figure 1.

Figure 1 Study flow chart. N, node; SCLC, small cell lung cancer; SEER, Surveillance, Epidemiology, and End Results; T, tumor.

A complete case analysis was employed to ensure model reliability and interpretability. The inclusion of patients with missing key prognostic variables (e.g., tumor size and metastasis status) might have compromised model performance and clinical applicability, as these have been established as strong prognostic factors in SCLC patients.

Statistical analysis

R version 4.2.3 (R Foundation for Statistical Computing) was used to analyze the data and SEERStat version 8.4.1.1 (National Cancer Institute) was used to extract the data. Consistent with the established methodology for SEER-based nomogram studies, each case was treated as an independent observation unit without applying sampling weights, as SEER represents a population-based cohort rather than a complex survey design. The data are presented as numbers and percentages. The training cohort’s risk and protective factors were determined by univariate and LASSO regression analyses. Based on the findings of the LASSO regression, two nomograms were created to predict the risk of ACED and CSED, respectively. The concordance index (C-index) and receiver operating characteristic (ROC) curves were used to assess the discrimination of the nomograms and calibration curves were used to determine the degree of agreement between the predicted probabilities and observed outcomes (14,15). Finally, a decision curve analysis (DCA) was conducted to evaluate the clinical utility of the predictive models at different thresholds (16,17). A two-tailed P value less than 0.05 was considered statistically significant.

Results

Characteristics of patients with SCLC-BM

A total of 633 SCLC-BM patients were included in the study and randomly allocated to the training (n=443) and test (n=190) cohorts at a ratio of 7:3. In the training cohort, 173 patients (39.1%) experienced ACED and 164 patients (37.0%) experienced CSED. In the test cohort, 79 patients (41.6%) experienced ACED and 77 patients (40.5%) experienced CSED. The process for patient selection is shown in Figure 1. The baseline characteristics of the SCLC-BM patients are provided in Table 1.

Of the patients, 54.7% were male and 45.3% were female. Most tumors were located in the upper lobe (50%), followed by the middle lobe (24%), overlapping sites (11%), lung locations not otherwise specified (NOS) (8%), bronchus (5%), and lower lobe (2%). Grade I–III lung cancer accounted for 41% of cases, while grade IV accounted for 59%. Of the patients, 27% had bone metastasis, 28% had liver metastasis, and 18% had lung metastasis. In terms of tumor size, 18% of the patients had tumors <30 mm, 26% had tumors between 30 and 49 mm, 23% had tumors between 50 and 69 mm, and 33% had tumors >69 mm. Of the patients, 72% received radiotherapy and 73% received chemotherapy. The majority of the patients were white, while only 10% were black and only 4% were of other races. Despite a nominal difference in age distribution (P=0.03), randomization ensured comparable baseline characteristics between the cohorts.

Identification of prognostic factors for ED

Table 2 sets out the results of the univariate analysis. The 15 variables were divided into 32 dichotomous variables for the LASSO regression. After the LASSO and multivariate logistic regression analyses, 7 variables, including primary site (lung, NOS), N0, age <60 years, tumor size <30 mm, radiotherapy, chemotherapy, and race (white), were selected as the predictors of ACED based on non-zero coefficients calculated from the LASSO regression (Figure 2A,2B). Additionally, age <60 years, tumor size <30 mm, radiotherapy, chemotherapy, and race (White) were selected as the predictors of CSED (Figure 2C,2D, and Table 3).

Figure 2 Variable selection using LASSO regression for early death prediction in SCLC-BM patients. (A) Coefficient paths for the ACED model: x-axis shows log(λ) values; y-axis shows standardized coefficients. Each colored line represents one predictor variable, with numbers indicating non-zero coefficients at each λ value. (B) Cross-validation curve for the ACED model: x-axis shows log(λ); y-axis shows binomial deviance ± standard error. Vertical dashed lines indicate λ.min (minimum deviance) and λ.1se. (C,D) Corresponding analyses for the CSED model. 1se, one standard error rule; ACED, all-cause early death; CSED, cancer-specific early death; LASSO, least absolute shrinkage and selection operator; SCLC-BM, small cell lung cancer with brain metastasis.

Nomogram establishment

The multivariate regression analysis was conducted using variables screened by the LASSO regression. Two nomograms were established to predict ACED and CSED, respectively, in patients with SCLC-BM. The ACED model served as our primary prognostic tool due to its objective nature, while the CSED findings should be interpreted with recognition of the potential death-coding limitations. The nomograms incorporated both diagnostic and treatment variables, and thus are most applicable after initial treatment decisions have been established. Clinicians can predict the prognosis of patients by inputting their clinical characteristics into the nomograms, enabling the identification of high-risk patients who may benefit from increased monitoring (Figure 3).

Figure 3 Nomogram prediction models for the SCLC-BM patients. (A) ACED prediction nomogram: each variable contributes points based on its value (top scale). Total points (bottom scale) correspond to the ED probability. (B) CSED prediction nomogram with similar interpretation. To use, locate the patient’s value for each variable, draw a vertical line to the points scale, sum all the points, and find the corresponding probability on the bottom scale. ACED, all-cause early death; CSED, cancer-specific early death; ED, early death; N, node; NOS, not otherwise specified; SCLC-BM, small cell lung cancer with brain metastasis.

Nomogram validation

The area under the curve (AUC) values of the ROC curves for ACED were 0.837 [95% confidence interval (CI): 0.797–0.876] in the training set and 0.813 (95% CI: 0.747–0.878) in the test set, while those for CSED were 0.794 (95% CI: 0.751–0.838) in the training set and 0.783 (95% CI: 0.708–0.858) in the test set, respectively. For the binary logistic regression models, the C-index values were mathematically equivalent to the AUC values reported above (Figure 4). The observed outcome and the anticipated probability were in high agreement as shown by the calibration curves (Figure 5). Additionally, the DCA demonstrated that the nomograms were clinically useful in predicting ACED and CSED (Figure 6).

Figure 4 ROC curves evaluating nomogram discriminative performance. (A,B) ACED model ROC curves: x-axis shows 1-specificity (false positive rate); y-axis shows sensitivity (true positive rate) in training and test cohorts. (C,D) CSED model ROC curves with same interpretation. AUC values indicate discriminative ability (0.5= random, 1.0= perfect discrimination). ACED, all-cause early death; AUC, area under the curve; CSED, cancer-specific early death; ROC, receiver operating characteristic.

Figure 5 Calibration plots assessing nomogram accuracy. X-axis shows predicted probability; y-axis shows observed probability. Perfect calibration follows diagonal reference line. (A,B) ACED model calibration in training and validation cohorts. (C,D) CSED model calibration. Closer alignment with diagonal line indicates better calibration. ACED, all-cause early death; CSED, cancer-specific early death.

Figure 6 DCA evaluating clinical utility. X-axis shows risk threshold; y-axis shows net benefit. Red line represents nomogram performance; green dashed line represents “treat all” strategy; blue dashed line represents “treat none” strategy. Higher net benefit indicates better clinical utility at given risk thresholds. (A,B) ACED model analysis. (C,D) CSED model analysis. ACED, all-cause early death; CSED, cancer-specific early death; DCA, decision curve analysis.

Discussion

BM is the most common intracranial tumor in adults, and is associated with significant morbidity and mortality. Lung cancer is a leading primary source of BM (18-20). Despite advances in anti-cancer therapeutics and the prolonged survival of patients with primary malignancies, the prevalence of BM continues to rise, currently reaching approximately 200,000 cases annually in the United States (18). The development of BM may induce neurological deficits, seizures, or even mental confusion, significantly reducing overall survival (21). Although significant advances have been made in the diagnosis and management of SCLC-BM patients, their prognosis remains unsatisfactory. Due to the poor prognosis associated with SCLC-BM, these patients are frequently left untreated, posing a significant global problem.

The blood-brain barrier (BBB) inhibits many systemic drugs from passing through and functioning at a therapeutic dose (22). As a result, BMs from SCLC have become a significant issue in neurosurgery, and the poor prognosis of patients with SCLC-BM continues to be a concern. To date, most research has focused on the ED of patients with SCLC, with little attention paid to the ED of patients with SCLC-BM specifically. An important methodological consideration is the inclusion of treatment variables as predictors, which introduces potential confounding by indication. Treatment decisions are influenced by unmeasured factors, including performance status and comorbidities. However, our models serve as prognostic tools for the early post-diagnosis period when treatment plans are established and can be used to identify patients at high risk despite intended standard care.

This study developed prognostic models for SCLC-BM. Although various studies have identified clinical-pathologic and therapeutic indicators for SCLC-BM survival, to the best of our knowledge, this was the first study to build nomogram prediction models for SCLC-BM ED. We found that primary site, N stage, age, race, tumor size, chemotherapy, and radiotherapy were independent predictors of ACED in patients with SCLC-BM, while tumor size, age, chemotherapy, radiotherapy, and race were significant predictors of CSED. The nomograms demonstrated good predictive performance for the clinical outcomes in both the training and test groups. An important methodological consideration is the inherent limitation of cause-of-death classification in the SEER data. Death certificates may not accurately distinguish between cancer-specific deaths and deaths from other causes, particularly those related to treatment complications such as chemotherapy-induced cardiovascular events. This misclassification bias directly undermines the specificity of CSED outcomes and supports the prioritizing of our more robust ACED model for primary clinical interpretation. In terms of their clinical applicability, our nomograms were designed for use after initial treatment planning rather than at diagnosis. The optimal timing is when treatment decisions have been established, enabling the identification of patients who remain at high risk despite planned standard treatment.

Age is undoubtedly related to the prognosis of patients with SCLC-BM (23). In our study, patients aged under 60 years displayed a decreased risk of all-cause and cancer-specific ED compared to those aged over 60 years. Previous studies have found that black patients with SCLC had a lower risk of death than white patients with SCLC, as black patients with SCLC were more likely to receive radiotherapy and treatments at academic/research centers (24). Our results are consistent with this, even though no statistically significant difference in mortality was observed between the black and white patients. Our findings should be interpreted cautiously, given the potential confounding by unmeasured socioeconomic and access-to-care factors not captured in the SEER data. Additionally, no statistically significant difference in mortality was observed between the patients of other races and the white patients, but this may be due to the small number of patients in the “other races” category. Thus, differences in ED between patients of other races and white patients need to be further explored.

Tumor size has been identified as a significant predictor of survival outcomes in SCLC patients. Larger tumor dimensions are correlated with a higher risk of ED (7). Patients with tumors <30 mm have a lower risk of early mortality than those with tumors >30 mm. Notably, while previous studies have reported that N stage has weak predictive value for ED in general SCLC populations (7), N stage demonstrated significantly greater prognostic significance in the SCLC-BM cohort in this study. Specifically, the patients with a N stage ≥2 exhibited an odds ratio of ≥2.33 (P<0.05) for ED, underscoring the heightened prognostic weight of lymph node metastasis in our SCLC-BM cohort. Notably, clinicopathological characteristics such as N stage and primary tumor site were selected as risk factors for ACED but not for CSED. This may be due to the smaller sample size of CSED cases relative to that of ACED cases in our study.

Consistent with previous studies (9), chemotherapy (P<0.0001) and radiotherapy (P<0.05) emerged as robust protective factors against ED. The pronounced efficacy of chemotherapy may be related to its ability to cross the BBB; for example, etoposide inhibits topoisomerase II to suppress DNA replication in BM (25,26). Notably, the significantly lower odds ratio for chemotherapy (0.05/0.10) versus radiotherapy (0.52/0.62) suggests that chemotherapy plays a dominant role in preventing ED in SCLC-BM. This divergence may be related to the high proliferative nature of SCLC, such that chemotherapy more effectively eradicates micrometastases, while radiotherapy better controls established brain lesions (27-29). Additionally, while radiotherapy reduces the local tumor burden, it may exacerbate BBB disruption, potentially increasing short-term neurotoxicity (30-31). In selecting treatment for SCLC-BM patients, a multidisciplinary approach is essential to maximize survival.

Another important methodological consideration is confounding by indication related to treatment variables. Treatment decisions are strongly influenced by unmeasured factors, including performance status and comorbidities, which our SEER-based models cannot capture. Thus, our nomograms can only serve as preliminary risk stratification tools based on objective tumor characteristics rather than standalone decision-making instruments. It is important to acknowledge the evolving therapeutic landscape of SCLC-BM management. While our study encompasses the pre-immunotherapy era (2010–2015), recent advances in immune checkpoint inhibitors have shown promise in the treatment of SCLC patients, including those with BM. Future prognostic models may need to incorporate immunotherapy variables to reflect the changing treatment paradigms and their effects on survival outcomes.

Previous research has shown that nomograms are more precise models for predicting ED in patients with lung cancer than traditional TNM staging alone (9,13). Heng et al. established a nomogram to predict the ED of lung cancer patients, and reported AUC values as high as 0.793 and 0.794 for both ACED and CSED (9). Yang et al. constructed two nomograms to predict ED in patients with non-SCLC, which had very high AUC values of 0.813 and 0.808 for both ACED and CSED (13). Using the factors identified as significant in the LASSO regression analysis, we developed nomograms for both ACED and CSED. Unlike previous studies using conventional univariate and multivariate analyses, our LASSO-based model effectively addressed the multicollinearity between variables (e.g., between TNM stages), yielding a superior AUC (0.837 vs. 0.793–0.813). This represents the first SEER-based LASSO nomogram models for ED prediction in SCLC-BM. Our model validation approach included a train-test split evaluation, which showed minimal performance degradation (AUC difference: 0.024 for ACED and 0.011 for CSED), suggesting adequate model stability. The LASSO regression with 20-fold cross-validation for parameter selection provides additional protection against overfitting through regularization, contributing to model robustness.

The internal examination of the nomograms revealed an excellent agreement between predicted and observed ED. Following initial treatment planning, these nomograms offer oncologists tools for mortality risk stratification, enabling the identification of patients at high risk despite planned treatment who may benefit from enhanced supportive care. Notably, our nomograms are not intended to suggest that chemotherapy should be withheld from high-risk patients, but rather to prompt more intensive monitoring and enhanced supportive care to ensure these patients can safely receive and potentially benefit from treatment. For patients with borderline performance status, in whom treatment feasibility is uncertain, nomogram scores could help determine whether enhanced supportive measures could enable the patient to undergo chemotherapy.

The prognostic factors in these models could be employed for stratification in clinical investigations that are randomized. Specifically, patients at high risk (>70%) of ED should receive earlier palliative care, and individualized treatment goals should be discussed with patients and their families. The treatment focus should shift from purely curative approaches to comprehensive supportive care to ensure quality of life and to develop individualized plans aligned with patients’ values while avoiding futile overtreatment. Conversely, moderate-to-low risk patients can proceed more confidently with standard chemotherapy and radiotherapy protocols, with objective risk assessments strengthening clinicians’ commitment to established treatment plans.

The present study had several limitations. Most significantly, the SEER database lacks critical clinical variables, including performance status and comorbidity indices, which are fundamental determinants of cancer prognosis. This absence creates potential confounding by indication, where the apparent protective effects of chemotherapy and radiotherapy may partially reflect patient selection rather than true treatment efficacy. As a retrospective cohort study, our findings are subject to the inherent limitations of observational research design, including potential selection bias and unmeasured confounding variables. Another significant study limitation is confounding by indication from treatment variables. The apparent protective effects of chemotherapy and radiotherapy may partially reflect patient selection rather than true treatment efficacy. This limits the models’ utility as a pure baseline prediction tool, making it most suitable for prognostic assessment after initial treatment planning. In addition, the SEER dataset contains no additional details on molecule-level pathological markers. Further, as so few SCLC-BM selected patients underwent surgery, the surgery information was excluded. The SEER database lacks detailed information on radiotherapy and chemotherapy, which is a limitation that should be addressed in further research. Further, despite the internal validation, no external validation was conducted. Therefore, studies with real-world cohorts need to be conducted to confirm the therapeutic applicability of the prognostic nomogram models. Additionally, our study used 2010–2015 data to maintain staging consistent with the 7th edition AJCC criteria. While this timeframe predated immunotherapy integration, the fundamental treatment approach (platinum-etoposide chemotherapy with radiotherapy) has remained largely unchanged, supporting the continued relevance of our predictive model. Another significant study limitation is potential misclassification bias in SEER cause-of-death coding, which may incorrectly attribute treatment-related deaths to cancer. This affects the reliability of our CSED model in particular, supporting the prioritization of our more robust ACED model for primary clinical conclusions. While our train-test validation and LASSO cross-validation provide some assurance against overfitting, we acknowledge that we did not perform additional stability assessments such as bootstrap resampling, which provides a more comprehensive validation of model robustness, particularly in smaller subgroups. Future studies incorporating bootstrap validation or other resampling methods could provide a more thorough assessment of model stability.

Conclusions

We developed predictive models for ED in SCLC-BM patients. The ACED model demonstrated robust predictive ability and clinical applicability, while the CSED model served as an exploratory analysis with recognized data limitations. These nomograms enable practical risk stratification that can guide individualized management strategies and optimize clinical decision-making in SCLC-BM care when combined with comprehensive clinical assessment.

Acknowledgments

The authors express their gratitude to all the patients, researchers, and institutions who contributed to the SEER database.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1269/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1269/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-2025-1269/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved. The study was conducted in accordance with the Declaration of Helsinki and its subsequent amendments.

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

References

1. Kim SY, Park HS, Chiang AC. Small Cell Lung Cancer: A Review. JAMA 2025;333:1906-17. [Crossref] [PubMed]
2. Xing X, Zhong J, Biermann J, et al. Pan-cancer human brain metastases atlas at single-cell resolution. Cancer Cell 2025;43:1242-1260.e9. [Crossref] [PubMed]
3. Wang S, Tang W, Jin F, et al. Comprehensive Analysis of Lung Cancer Metastasis: Sites, Rates, Survival, and Risk Factors-A Systematic Review and Meta-Analysis. Clin Respir J 2025;19:e70107. [Crossref] [PubMed]
4. Robinson SD, Wong CD, Mazarakis NK, et al. A Review of Focal Therapies for Small Cell Lung Cancer Brain Metastases. Clin Lung Cancer 2025;26:441-57. [Crossref] [PubMed]
5. Xie Z, Wang Y, Chen T, et al. Circulating tumor cells with increasing aneuploidy predict inferior prognosis and therapeutic resistance in small cell lung cancer. Drug Resist Updat 2024;76:101117. [Crossref] [PubMed]
6. Chen R, Liu Y, Tou F, et al. A practical nomogram for predicting early death in elderly small cell lung cancer patients: A SEER-based study. Medicine (Baltimore) 2024;103:e37759. [Crossref] [PubMed]
7. Li Z, Pang M, Liang X, et al. Risk factors of early mortality in patients with small cell lung cancer: a retrospective study in the SEER database. J Cancer Res Clin Oncol 2023;149:11193-205. [Crossref] [PubMed]
8. Zheng X, Liu K, Gao Z, et al. Predicting overall survival and prophylactic cranial irradiation benefit in small cell lung cancer patients: a multicenter cohort study. BMC Cancer 2024;24:1507. [Crossref] [PubMed]
9. Shen H, Deng G, Chen Q, et al. The incidence, risk factors and predictive nomograms for early death of lung cancer with synchronous brain metastasis: a retrospective study in the SEER database. BMC Cancer 2021;21:825. [Crossref] [PubMed]
10. Zhang L, Jing H, Tang S, et al. Benefits of Dynamic Nomogram Models for Elderly Diffuse Large B-Cell Lymphoma Patients' Early Death Prediction and Clinical Application. Mediators Inflamm 2023;2023:7874239. [Crossref] [PubMed]
11. Hankey BF, Ries LA, Edwards BK. The surveillance, epidemiology, and end results program: a national resource. Cancer Epidemiol Biomarkers Prev 1999;8:1117-21. [PubMed]
12. Hayat MJ, Howlader N, Reichman ME, et al. Cancer statistics, trends, and multiple primary cancer analyses from the Surveillance, Epidemiology, and End Results (SEER) Program. Oncologist 2007;12:20-37. [Crossref] [PubMed]
13. Yang F, Gao L, Wang Q, et al. Development and validation of nomograms to predict early death in non-small cell lung cancer patients with brain metastasis: a retrospective study in the SEER database. Transl Cancer Res 2023;12:473-89. [Crossref] [PubMed]
14. Janka EA, Szabó IL, Toka-Farkas T, et al. Development of Nomograms to Predict the Probability of Recurrence at Specific Sites in Patients with Cutaneous Melanoma. Cancers (Basel) 2025;17:3080. [Crossref] [PubMed]
15. Liu J, Lyu Y, He Y, et al. Competing risk nomogram and risk classification system for evaluating overall and cancer-specific survival in neuroendocrine carcinoma of the cervix: a population-based retrospective study. J Endocrinol Invest 2024;47:1545-57. [Crossref] [PubMed]
16. Gao Y, Gan X. A novel nomogram for the prediction of subsyndromal delirium in patients in intensive care units: A prospective, nested case-controlled study. Int J Nurs Stud 2024;155:104767. [Crossref] [PubMed]
17. Zhao L, Leng Y, Hu Y, et al. Understanding decision curve analysis in clinical prediction model research. Postgrad Med J 2024;100:512-5. [Crossref] [PubMed]
18. La Rosa A, Wieczorek DJJ, Tolakanahalli R, et al. Dosimetric Impact of Lesion Number, Size, and Volume on Mean Brain Dose with Stereotactic Radiosurgery for Multiple Brain Metastases. Cancers (Basel) 2023;15:780. [Crossref] [PubMed]
19. Wang Z, Chen H, Chen Q, et al. Multi-institutional study of 'Sandwich treatment' for motor area large brain metastases (LBM) with diameter over 3 cm. Radiol Oncol 2024;58:145-52. [Crossref] [PubMed]
20. Lah TT, Novak M, Breznik B. Brain malignancies: Glioblastoma and brain metastases. Semin Cancer Biol 2020;60:262-73. [Crossref] [PubMed]
21. Scheepens JCC, van der Meer PB, Dirven L, et al. Seizure outcomes in patients with brain metastases and epilepsy: a systematic review on the efficacy of antitumor treatment and antiseizure medication. Neurooncol Pract 2025;12:376-88. [Crossref] [PubMed]
22. Wu D, Chen Q, Chen X, et al. The blood-brain barrier: structure, regulation, and drug delivery. Signal Transduct Target Ther 2023;8:217. [Crossref] [PubMed]
23. Chiang CL, Yang HC, Liao YT, et al. Treatment and survival of patients with small cell lung cancer and brain metastasis. J Neurooncol 2023;165:343-51. [Crossref] [PubMed]
24. Shi H, Zhou K, Cochuyt J, et al. Survival of Black and White Patients With Stage IV Small Cell Lung Cancer. Front Oncol 2021;11:773958. [Crossref] [PubMed]
25. Le TT, Wang MD. Topoisomerase II and etoposide - a tangled tale. Nat Chem Biol 2023;19:546-7. [Crossref] [PubMed]
26. Le TT, Wu M, Lee JH, et al. Etoposide promotes DNA loop trapping and barrier formation by topoisomerase II. Nat Chem Biol 2023;19:641-50. [Crossref] [PubMed]
27. Gao X, Liu T, Fan M, et al. The therapeutic effect of radiotherapy combined with systemic therapy compared to radiotherapy alone in patients with simple brain metastasis after first-line treatment of limited-stage small cell lung cancer: a retrospective study. World J Surg Oncol 2024;22:89. [Crossref] [PubMed]
28. Aizer AA, Tanguturi SK, Shi DD, et al. Stereotactic Radiosurgery in Patients With Small Cell Lung Cancer and 1-10 Brain Metastases: A Multi-Institutional, Phase II, Prospective Clinical Trial. J Clin Oncol 2025;43:2986-97. [Crossref] [PubMed]
29. Liu M, Qiu G, Guan W, et al. Induction chemotherapy followed by camrelizumab plus apatinib and chemotherapy as first-line treatment for extensive-stage small-cell lung cancer: a multicenter, single-arm trial. Signal Transduct Target Ther 2025;10:65. [Crossref] [PubMed]
30. Alcazar-Felix RJ, Cheng L, Srinath A, et al. A mouse model of stereotactic radiosurgery-induced neuroinflammation and blood-brain barrier compromise. Acta Neuropathol Commun 2025;13:195. [Crossref] [PubMed]
31. Wang P, Liu J, Zhang M, et al. Radiation Exposure Induced Blood-Brain Barrier Injury via Mitochondria-Mediated Sterile Inflammation. Adv Sci (Weinh) 2025;12:e02356. [Crossref] [PubMed]

(English Language Editor: L. Huleatt)