Brain metastases in newly diagnosed lung cancer: epidemiology and conditional survival

Introduction

Lung and bronchus primary cancer stands as the foremost cause of cancer-related mortality, boasting a disheartening 5-year survival rate of only 19%. Its prevalence has escalated to an important health dilemma, bearing profound medical and socioeconomic ramifications in both developing and developed nations (1-3). One significant factor contributing to the poor prognosis of these patients is the diagnosis of lung cancer at an advanced stage, where therapeutic efficacy is currently limited. Despite notable advancements in treatment modalities such as targeted molecular therapies and immunotherapies, survival outcomes for patients with tumor distant metastases remain frustratingly low (3,4). Brain metastasis is predominant among both non-small cell lung cancer (NSCLC) and small cell lung cancer (SCLC) patients (5), constituting a debilitating condition associated with significant morbidity and mortality (1,6,7). Some studies have reported that 16% to 22% of lung cancer patients eventually develop brain metastases (3,8). However, despite efforts by some research groups with smaller sample sizes to determine the prognosis of lung cancer with brain metastasis (LCBM), there remains a lack of large-scale studies examining the incidence and survival rates of patients with LCBM.

Conditional survival (CS) serves as a valuable metric predicting the likelihood of a patient surviving for an additional year after having survived for x years since their cancer diagnosis (9-11). The advantageous nature of CS lies in its ability to provide dynamic, real-time prognostic estimates. However, it often overlooks the clinicopathological features of cancer patients. Certain clinical factors, such as age, tumor histology, tumor grade, and treatment information, have been identified as influential in the survival of patients with LCBM (7,12,13). However, the combined effect of these factors in predicting CS for these patients is yet to be thoroughly investigated.

Therefore, this study utilized the Surveillance, Epidemiology, and End Results (SEER) database to perform a thorough analysis of the epidemiological characteristics and CS of patients with LCBM. Additionally, we merged the conventional nomogram model, which can integrate prognostic factors to predict survival, with CS analysis to develop a novel CS-based nomogram model. We present this article in accordance with the TRIPOD reporting checklist (available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-776/rc).

Methods

Data source

In our study, we analyzed cases extracted from the SEER database. The SEER program gathers incidence and survival data from population-based state cancer registries, covering approximately 35% of the population in the U.S. (14). The study was conducted in accordance with the Declaration of Helsinki (as revised in 2013). Since SEER data are publicly accessible, no Institutional Review Board approval is required.

Study population

We conducted a retrospective analysis on lung cancer patients diagnosed with brain metastasis from 2010 to 2015. Cases diagnosed by autopsy or death certificate were excluded, along with patients with unknown detailed information.

Variable declaration

Demographic variables, including age at diagnosis, gender (male or female), race (White, Black, or other), insurance status (insured or uninsured), and marital status (single or married), were obtained from the SEER database. The optimal cutoff values for age were identified as ≤61, 62–77, or ≥78 years using the X-tile program (see Figure S1). Tumor characteristics, such as primary tumor sites, the total number of (in situ/malignant) tumor sites, tumor size, and histological types, were also collected. Tumor size was categorized as <3, ≥3 and <5, ≥5 and <7, and ≥7 cm. Histological types were grouped as adenocarcinoma, squamous cell carcinoma (SCC), SCLC, NSCLC, and others. Additionally, treatment information including surgery, radiotherapy, and chemotherapy was also extracted.

Statistical analysis

Incidence, trends, and relative survival

Using the SEER*Stat statistical software, we initially investigated the incidence, trends, and relative survival among LCBM patients diagnosed from 2010 to 2015. Relative survival rate is a specialized measure that adjusts for all causes of death except cancer. Using the Ederer II method, relative survival was determined by comparing observed survival with expected survival derived from the general U.S. population, matched for age, sex, and race (15).

CS analysis

CS plays a crucial role in assessing the long-term survival of patients diagnosed with cancer. The CS formula, CS(y|x) = CS(y+x)/CS(x), calculates the probability of a patient’s survival for y additional years, given they have already survived x years following their LCBM diagnosis. CS(x) and CS(y+x) were estimated using the Kaplan-Meier method for x- and (x+y)-year survival.

CS-based nomogram development and validation

We randomly divided eligible patients into training and validation groups at a 7:3 ratio. The least absolute shrinkage and selection operator (LASSO) regression analysis with 10-fold cross-validation was then conducted in the training cohort to identify prognostic predictors, avoiding overfitting. These factors were further analyzed via multivariate Cox regression to confirm their significance as prognostic indicators. Utilizing these factors, we developed a predictive nomogram model incorporating the CS formula for personalized and dynamic survival prediction.

Finally, the model underwent evaluation and validation in both the training and validation groups. The model’s accuracy was assessed through calibration plots generated from 1,000 bootstrap samples, and discrimination was evaluated using the area under the time-dependent receiver operating characteristic (ROC) curve (AUC) and concordance index (c-index). Additionally, decision curve analysis (DCA) was employed to assess the clinical utility of the nomogram, measuring the net benefit of nomogram-guided medical interventions.

All statistical tests were two-sided, with P<0.05 considered statistically significant. All statistical analyses were performed using R software (version 4.1.0).

Results

Baseline characteristics

According to the defined inclusion and exclusion criteria, a total of 9,212 patients diagnosed with LCBM were included in our study, with 6,448 in the training group and 2,764 in the validation group. Patient demographics and tumor characteristics are detailed in Table 1. In the entire cohort, 3,343 patients (36.3%) were under 61 years old, 4,655 (50.5%) were aged between 62 and 77 years, and 1,214 (13.2%) were over 78 years old. The majority of patients were male (53.6%), and most were of White ethnicity (79.4%). Adenocarcinoma (53.2%) was the most common histological type of tumors, followed by SCC (13.6%), NSCLC (10.0%), and SCLC (9.1%). Tumors were predominantly located in the upper lobe (57.2%). Surgery was performed on only 582 LCBM patients (6.3%), while the majority underwent radiotherapy (78.6%). Additionally, 5,259 cases (57.1%) opted for chemotherapy.

Incidence of LCBM

The age-adjusted incidence rate of LCBM was 5.82 per 100,000 during the study period, as depicted in Figure 1A. Subsequently, we analyzed the annual incidence of patients across various age groups. Our findings revealed a significant increase in incidence with age, peaking among individuals aged 70 to 79 years, followed by a decline (Figure 1B).

Figure 1 Epidemiology and survival of lung cancer with brain metastasis. (A) Incidence of lung cancer with brain metastasis; (B) incidence of lung cancer with brain metastasis stratified by age groups; (C) relative survival of lung cancer with or without brain metastasis; (D) relative survival of lung cancer with brain metastasis stratified by age groups.

Survival analysis of lung cancer patients with or without brain metastasis

Next, we compared the relative survival of lung cancer patients with and without brain metastasis. As depicted in Figure 1C, LCBM patients exhibited a dismal prognosis. Furthermore, we conducted additional analysis to assess the impact of age on patient outcomes. As anticipated, the relative survival rate of patients continued to decline with increasing age (Figure 1D).

CS of LCBM patients

The overall survival rates for LCBM patients were observed to be 24%, 6%, and 3% for 1, 3, and 5 years, respectively (Figure 2). Furthermore, the CS curve illustrated an increasing survival rate for LCBM patients over time, as shown in Figure 2. Notably, the 5-year survival rate for patients initially diagnosed with LCBM was adjusted from 3% to 13%, 28%, 52%, and 73% over successive years of survival (1–4 years).

Figure 2 Kaplan-Meier method for estimating conditional survival at 5 years after surviving 0–4 years in LCBM patients. Conditional survival curves and their updated survival data adjusted for survived time. LCBM, lung cancer with brain metastasis.

CS-based nomogram development

Based on the training cohort, the LASSO regression method identified eight key prognostic predictors incorporated into the final survival prediction model, including: age at diagnosis, sex, race, tumor size, tumor grade, surgery, radiotherapy, and chemotherapy (Figure 3). To ensure accuracy, these variables underwent multivariate Cox analysis to confirm their prognostic value (Figure 4). Additionally, multivariate Cox regression analysis was employed to construct a nomogram, assigning varying regression coefficients to selected prognostic factors based on their respective impact on the outcome (Figure 5). Unlike traditional predictive models, the CS-based nomogram utilized in this study considered CS, allowing patients to gain insight not only into their 1-, 3-, and 5-year overall survival rates based on their personalized clinicopathologic characteristics but also their 5-year CS rates based on the duration of time since diagnosis.

Figure 3 Predictor screening. The LASSO regression model (A) and 10-fold cross-validation technique for predictor selection (B). LASSO, least absolute shrinkage and selection operator.

Figure 4 Multivariate Cox regression forest plot confirmed the prognostic value of selected variables. HR, hazard ratio; CI, confidence interval; SCC, squamous cell carcinoma; SCLC, small cell lung cancer; NSCLC, non-small cell lung cancer; RT, radiotherapy; CT, chemotherapy.

Figure 5 Conditional survival nomogram (CS-based nomogram) for predicting 1-, 3- and 5-year OS and 5-year CS for LCBMs. OS, overall survival; CS, conditional survival; LCBM, lung cancer with brain metastasis.

CS-based nomogram evaluation and validation

Various validation methods were employed to assess the model’s performance. Calibration plots for both the training and validation cohorts indicated a strong correlation between the model and actual outcomes (Figure 6A,6B). These plots closely aligned with the 45-degree line, indicating the model’s accuracy in reflecting reality. The c-index of the dynamic CS-nomogram was 0.739 and 0.723 in the training and validation cohorts, respectively. Time-dependent ROCs at 1, 3, and 5 years demonstrated satisfactory discriminatory ability, with AUCs of 0.78, 0.79, and 0.80 in the training group (Figure 6C) and 0.77, 0.75, and 0.83 in the validation group (Figure 6D). Additionally, DCA curves consistently showed a favorable net benefit whenever medical interventions were initiated based on the CS-nomogram in both the training and validation groups (Figure 7A,7B).

Figure 6 Conditional survival nomogram model evaluation and validation. Calibration plots of the CS-based nomogram for predicting the probability of survival at 1, 3, and 5 years in both training (A) and validation (B) cohorts. Time-dependent ROC curves for assessing the discrimination of the model in both training (C) and validation (D) cohorts. AUC, area under the curve; CS, conditional survival; ROC, receiver operating characteristic.

Figure 7 Decision curve analysis of the CS-based nomogram in both training (A) and validation (B) cohorts. The x-axis represents the percentage of threshold probability, whereas the y-axis represents the net benefit, calculated by adding the true positives and subtracting the false positives. CS, conditional survival.

Discussion

Brain metastasis is an extremely serious sequela remaining the single largest cause of mortality in patients with lung cancer (16). And mounting evidence suggests that improved detection techniques and clinical awareness, as well as changes in the treatment have led to the increased incidence of the LCBM (17). Several recent studies have attempted to deepen understanding of LCBM, but data on its epidemiology and prognosis remain unclear. A better understanding of LCBM can optimize clinical decision-making for healthcare providers during the clinical evaluation stage. Therefore, we conducted a comprehensive analysis of the epidemiological characteristics and CS of LCBM using data extracted from the SEER database.

Epidemiologically, the overall age-adjusted incidence rate of the LCBM was 5.82 cases per 100,000 and trend was slightly declined during the period of our study. We also compared the incidence in different age groups and found that tumor peak incidence occurred in patients aged 70 to 79 years. Villano et al. analyzed cases diagnosed from 2010 to 2011 using the Kentucky Cancer Registry (KCR) and Alberta Cancer Registry (ACR), and reported that the median incidence of stage I/II and III LCBM was 25.6% and 19.3% respectively (18). In addition, the relative survival rates of patients with brain metastasis were much worse than those without brain metastasis. And we further illustrated the effect of age on the relative survival of LCBM patients. These results have potential extensive value of application and may influence screening paradigms, treatment strategies, and clinical trial design for specific subsets of patients with lung cancer.

CS is a method of survival assessment that estimates changes in real-time based on the time survived. We then described the CS pattern of LCBM patients and found that despite the advanced stage of LCBM, long-term survival improved significantly for each additional year patients lived. The analysis of the CS revealed a significant increase in the 2^nd year post-diagnosis [CS(1|1) =46%], with a gradual improvement continuing in the following years. Therefore, the findings of the CS analysis indicating a significant enhancement in the possibility of survival could alleviate the anxiety of cancer patients, increase their confidence in overcoming the disease, and improve their overall quality of life. Moreover, understanding this dynamic survival pattern may help establish cost-effective methods for LCBM surveillance, including determining the appropriate duration and intensity of follow-up.

The prognostication of patients in real-time is not only time-dependent, but also subject to variations based on individual clinicopathological characteristics (19,20). To address this, a dynamic CS-nomogram was also developed in our study. We employed the LASSO regression to avoid overfitting or underfitting the mode and finally identify eight prognostic factors for our CS-based nomogram model establishment. Finally, we constructed and internally validated a CS prediction model for these patients. This is the first study, to the best of our knowledge, to develop a CS-based nomogram integrating easily-obtained variables together in determining risk of death for LCBM patients. Our prognostic model demonstrated exceptional predictive ability and could be practically implemented in clinical settings to accurately forecast the dynamic and real-time life expectancy of patients with LCBM. Additionally, mutational profiling indicated that an elevated frequency of genetic mutation facilitated the development of diverse patterns of brain metastasis (21,22). Epidermal growth factor receptor (EGFR) mutation and anaplastic lymphoma kinase (ALK) positive have been found to be associated with more frequent brain metastases in patients with lung cancer (23-25). Efforts are actively underway to investigate the impact of various cancer-related factors, including distinct categories of genes that drive metastasis, on the progression of tumors (16,26-28). While our analysis was limited by the lack of genomic data available through the SEER database, we were able to achieve effective dynamic predictions by incorporating basic clinicopathological features.

There are several limitations that should be considered in the present study. The first may be the retrospective nature of the SEER database-based study and selection bias may be virtually brought in. Second, detailed information on treatment variables such as surgical procedures, immunotherapy, chemotherapy regimens, and radiation dose/technology which related to patient outcomes are not collected by SEER database. Third, we only analyzed patients based on the initial diagnosis of LCBM, data on tumor recurrence was unavailable in SEER database. Lastly, the inclusion of additional prognostic factors such as biological markers would enhance the predictive capacity of our model, and external validation is warranted.

Conclusions

Despite its limitations, this study significantly contributes to our understanding of LCBM patients. Our findings provide a comprehensive overview of their epidemiological characteristics and CS patterns. Through LASSO analysis, we identified eight prognostic factors, facilitating the development of a novel CS-based nomogram model for dynamic survival estimation. This model holds the potential to enhance clinical decision-making and optimize disease management strategies for LCBM patients.

Acknowledgments

The authors would like to thank the SEER database for the availability of the data.

Funding: None.

Footnote

Reporting Checklist: The authors have completed the TRIPOD reporting checklist. Available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-776/rc

Peer Review File: Available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-776/prf

Conflicts of Interest: Both authors have completed the ICMJE uniform disclosure form (available at https://tcr.amegroups.com/article/view/10.21037/tcr-24-776/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 (as revised in 2013). Since the study relied on the SEER database, which was freely available to the public (http://seer.cancer.gov/seerstat/), the requirement for ethical approval was waived.

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. Lee H, Jeong SH, Jeong BH, et al. Incidence of Brain Metastasis at the Initial Diagnosis of Lung Squamous Cell Carcinoma on the Basis of Stage, Excluding Brain Metastasis. J Thorac Oncol 2016;11:426-31. [Crossref] [PubMed]
2. Siegel R, Naishadham D, Jemal A. Cancer statistics, 2013. CA Cancer J Clin 2013;63:11-30. [Crossref] [PubMed]
3. Waqar SN, Waqar SH, Trinkaus K, et al. Brain Metastases at Presentation in Patients With Non-Small Cell Lung Cancer. Am J Clin Oncol 2018;41:36-40. [Crossref] [PubMed]
4. Sereno M, Hernandez de Córdoba I, Gutiérrez-Gutiérrez G, et al. Brain metastases and lung cancer: molecular biology, natural history, prediction of response and efficacy of immunotherapy. Front Immunol 2024;14:1297988. [Crossref] [PubMed]
5. Wang SY, Ye X, Ou W, et al. Risk of cerebral metastases for postoperative locally advanced non-small-cell lung cancer. Lung Cancer 2009;64:238-43. [Crossref] [PubMed]
6. Cagney DN, Martin AM, Catalano PJ, et al. Incidence and prognosis of patients with brain metastases at diagnosis of systemic malignancy: a population-based study. Neuro Oncol 2017;19:1511-21. [Crossref] [PubMed]
7. Sánchez de Cos J, Sojo González MA, Montero MV, et al. Non-small cell lung cancer and silent brain metastasis. Survival and prognostic factors. Lung Cancer 2009;63:140-5. [Crossref] [PubMed]
8. Davis FG, Dolecek TA, McCarthy BJ, et al. Toward determining the lifetime occurrence of metastatic brain tumors estimated from 2007 United States cancer incidence data. Neuro Oncol 2012;14:1171-7. [Crossref] [PubMed]
9. Jung SH, Lee HY, Chow SC. Statistical Methods for Conditional Survival Analysis. J Biopharm Stat 2018;28:927-38. [Crossref] [PubMed]
10. Zabor EC, Gonen M, Chapman PB, et al. Dynamic prognostication using conditional survival estimates. Cancer 2013;119:3589-92. [Crossref] [PubMed]
11. Hieke S, Kleber M, König C, et al. Conditional Survival: A Useful Concept to Provide Information on How Prognosis Evolves over Time. Clin Cancer Res 2015;21:1530-6. [Crossref] [PubMed]
12. Bernhardt D, Adeberg S, Bozorgmehr F, et al. Outcome and prognostic factors in single brain metastases from small-cell lung cancer. Strahlenther Onkol 2018;194:98-106. [Crossref] [PubMed]
13. Ampil F, Caldito G, Milligan S, et al. The elderly with synchronous non-small cell lung cancer and solitary brain metastasis: does palliative thoracic radiotherapy have a useful role? Lung Cancer 2007;57:60-5. [Crossref] [PubMed]
14. Katz SJ. Cancer Care Delivery Research and the National Cancer Institute SEER Program: Challenges and Opportunities. JAMA Oncol 2015;1:677-8. [Crossref] [PubMed]
15. Ederer F, Axtell LM, Cutler SJ. The relative survival rate: a statistical methodology. Natl Cancer Inst Monogr 1961;6:101-21. [PubMed]
16. Andratschke N, Kraft J, Nieder C, et al. Optimal management of brain metastases in oncogenic-driven non-small cell lung cancer (NSCLC). Lung Cancer 2019;129:63-71. [Crossref] [PubMed]
17. Ebben JD, You M. Brain metastasis in lung cancer: Building a molecular and systems-level understanding to improve outcomes. Int J Biochem Cell Biol 2016;78:288-96. [Crossref] [PubMed]
18. Villano JL, Durbin EB, Normandeau C, et al. Incidence of brain metastasis at initial presentation of lung cancer. Neuro Oncol 2015;17:122-8. [Crossref] [PubMed]
19. Iasonos A, Schrag D, Raj GV, et al. How to build and interpret a nomogram for cancer prognosis. J Clin Oncol 2008;26:1364-70. [Crossref] [PubMed]
20. Balachandran VP, Gonen M, Smith JJ, et al. Nomograms in oncology: more than meets the eye. Lancet Oncol 2015;16:e173-80. [Crossref] [PubMed]
21. Gu Y, Zhang J, Zhou Z, et al. Metastasis Patterns and Prognosis of Octogenarians with NSCLC: A Population-based Study. Aging Dis 2020;11:82-92. [Crossref] [PubMed]
22. Guan C, Zhang X, Yu L. A Review of Recent Advances in the Molecular Mechanisms Underlying Brain Metastasis in Lung Cancer. Mol Cancer Ther 2024;23:627-37. [Crossref] [PubMed]
23. Li L, Luo S, Lin H, et al. Correlation between EGFR mutation status and the incidence of brain metastases in patients with non-small cell lung cancer. J Thorac Dis 2017;9:2510-20. [Crossref] [PubMed]
24. Hsu F, De Caluwe A, Anderson D, et al. EGFR mutation status on brain metastases from non-small cell lung cancer. Lung Cancer 2016;96:101-7. [Crossref] [PubMed]
25. Khasraw M, Yalamanchili P, Santhanagopal A, et al. Clinical Management of Patients with Non-Small Cell Lung Cancer, Brain Metastases, and Actionable Genomic Alterations: A Systematic Literature Review. Adv Ther 2024;41:1815-42. [Crossref] [PubMed]
26. Burudpakdee C, Wong W, Seetasith A, et al. Economic impact of preventing brain metastases with alectinib in ALK-positive non-small cell lung cancer. Lung Cancer 2018;119:103-11. [Crossref] [PubMed]
27. Li JD, Feng QC, Qi Y, et al. PPARGC1A is upregulated and facilitates lung cancer metastasis. Exp Cell Res 2017;359:356-60. [Crossref] [PubMed]
28. Chen YM, Lai CH, Rau KM, et al. Advanced non-Small cell lung cancer patients at the extremes of age in the era of epidermal growth factor receptor tyrosine kinase inhibitors. Lung Cancer 2016;98:99-105. [Crossref] [PubMed]