Introduction

According to the Central Brain Tumor Registry of the United States Database, the annual average age-specific incidence rate of all primary malignant and nonmalignant tumors in adolescents and young adults was 12.00 per 100,000 population between 2016 and 20201. In China in 2022, new cancer cases and cancer incidence rates (ASIR, age-standardized incidence rate) of brain and other central nervous system (CNS) tumors were 8.75 × 104 and 4.17/105, respectively2. Among CNS tumors, meningioma is the most common benign tumor, and diffuse glioma is the most common malignant tumor1,3,4,5. These CNS tumors impose a significant disease burden, especially diffuse gliomas, which are associated with poor survival.

With successful treatment, patients are less likely to die from primary tumors and are more likely to live with tumors that are not life-threatening, such as the common meningiomas among nonmalignant CNS tumors6. However, patients with better survival are more likely to die from nontumour-related deaths7,8,9,10,11. These nontumour death events, as competing risk events for primary tumor death, play important roles as causes of death (COD) among individuals with specific tumors/cancers. Previous studies have investigated all-cause and cause-specific mortality among patients with different types of cancer8,9,10,12,13,14,15,16,17. Several studies investigated the all-cause and cancer-specific death of cancer patients of different ages, including older adults and children10,13,17. Moreover, all-cause and cause-specific mortalities were also investigated in different types of cancer, including Gynaecological Cancer12, prostate cancer15, gastric precancerous lesions9, metastatic cancer8. To our knowledge, however, these studies focused less on the mortality of competing nontumor events, and studies on CNS tumors are lacking. In particular, all-cause and cause-specific mortality disparities in tumor behavior among patients with CNS tumors have not been investigated. In this study, we aimed to comprehensively investigate the tumor behavior disparities in all-cause and cause-specific mortality (nonmalignant vs. malignant) among patients with CNS tumors, which might be helpful for proposing more refined clinical management strategies for different types of CNS tumors.

Methods

Data source and study population

The data of patients with CNS tumors in this study were obtained from the SEER database (SEER-17; 2000–2021)18. We retrieved and collected data from U.S. population with CNS tumors, including 328,523 patients with malignant and nonmalignant tumors (age: 0 ~ 100 years). The pathological types of all CNS tumors are based on the Hist/Behave classifications of the International Classification of Diseases for Oncology, third edition (ICD-O-3) and SEER Brain and CNS Recode. The inclusion criteria were: (1) confirmed diagnosis of CNS tumors, including all types of malignant and nonmalignant tumors; and (2) availability of critical clinical and survival data (including demographic characteristics, tumor characteristics, treatment, survival outcomes, and survival status). In addition, 3706 patients lacking survival outcomes and survival status data were excluded. A total of 324,817 patients with malignant and nonmalignant tumors were included in this study. A flowchart of case inclusion and study progress illustrated in Fig. 1.

Fig. 1
figure 1

Flow diagram of the case selection and study process.

Full size image

Variables of interest and outcomes

The variables of interest included patient demographics (age, sex, race, and diagnosis year), tumor features (tumor laterality, tumor behavior, and tumor type), treatments (surgery type, radiation, and chemotherapy), survival time (months), final status, and COD. These variables were categorized according to the database codes, including age group (child and adult) and/or (< 18 [child], 18 ~ 64 [young and middle adult], and > = 65 years [older adult]), sex (male and female), race (White, Black, Asian, American-Indian/Native, and unknown), diagnosis year (2000 ~ 2007, 2008 ~ 2015, and 2016 ~ 2021), and tumor laterality (left, right, bi/midline [bilateral/midline], and unknown). According to the tumor behavior and types of World Health Organization (WHO) classification, tumor types were classified into two subgroups (seven types), including four types of nonmalignant tumors (nonmalignant meningiomas, ependymal neuronal tumors, cranial paraspinal tumors, and other nonmalignant tumors) and three types of malignant tumors (lower-grade glioma [LGG], glioblastoma [GBM] & diffuse midline glioma, H3K27M altered [DMG], and other malignant tumors) (Table 1 and Supplemental Table S1). The surgery type was documented as gross total resection (GTR), subtotal resection (STR), partial resection (PR), or biopsy according to the SEER codes for surgery for CNS tumors. In addition, several variables, including coexisting malignant tumors (Co_Malig), coexisting nonmalignant tumors (Co_NonMalig), coexisting tumors, radiation, and chemotherapy were recorded as No or Yes (Table 1).

Table 1 Summary of CNS tumor characteristics according to tumor behavior.
Full size table

The final status of all patients with CNS tumors was categorized into alive, primary tumor death, and competing event death. Competing event deaths refer to deaths from any cause (except for primary tumor death) and are categorized into deaths from cardiovascular disease (CVD), respiratory disease, subsequent malignant cancer, nervous, digestive, endocrine, genitourinary, infection diseases, and unknown (Supplemental Table S2). The proportion of COD was used to describe the most common causes of death. The cumulative mortality and subgroup analyses were estimated and plotted by cumulative mortality curves. Cause-specific hazard ratios for all-cause death, primary tumor death, and competing event death were calculated.

Statistical analysis

In this study, all the variables are categorical, and the Pearson Chi-square test is used to examine the differences in malignant and nonmalignant tumors among the levels of the independent variables. The proportion and cumulative mortality were calculated by comparing different causes of death (COD)6,7. All-cause mortality refers to mortality due to all causes of death. Cause-specific mortality refers to mortality due to different causes of death (including primary tumor mortality and competing event mortality). The proportion of COD was used to describe the most common cause of death, which was defined as the ratio of cause-specific death to general death (see Fig. 2 and Fig. S1)7,19. The cumulative mortality and subgroup analyses were estimated via a Fine‒Gray competing-risk regression model to investigate the relationship between competing event death and primary tumor death7,19. The proportional hazards assumption was assessed graphically by plotting the cumulative mortality curves (see Figs. 3 and 4 and Figs. S2, S3)20,21,22. Cause-specific hazard ratios (HRs) and 95% confidence intervals (CIs) for all-cause death, primary tumor death, and competing event death were calculated using the multivariable Cox proportional hazard model (see Fig. 5; Table 2)12,23,24,25. For the analysis of the cause-specific hazard ratio by tumor type, nonmalignant meningioma was used as a reference. All the statistical analyses were performed via R (version 4.4.0) software. A two-sided P value < 0.05 was considered statistically significant.

Fig. 2
figure 2

Proportion of causes of death among patients with CNS tumors. A, The final status of all patients with CNS tumors. B, Causes of competing event deaths. Abbreviations: CVD, cardiovascular disease; OtherCNS, other central nervous system sites; Malig, malignant; Non-Malig, nonmalignant; Sub_Malig_Cancer, subsequent malignant cancers; NA, not available.

Full size image
Fig. 3
figure 3

Comparisons of cumulative mortality in patients with CNS tumors according to different characteristics. A, COD by CNS tumor site; B, COD by age; C, COD by sex; D, COD by tumor behavior. Abbreviations: COD, causes of death.

Full size image
Fig. 4
figure 4

Stratified analysis of cumulative mortality with different features according to tumor behavior. A-D, Cumulative mortality among nonmalignant CNS tumor patients. A, COD by CNS tumor site; B, COD by age; C, COD by sex; D, COD by coexisting tumor. E-H, Cumulative mortality among patients with malignant CNS tumors. E, COD by CNS tumor site; F, COD by age; G, COD by sex; H, COD by coexisting tumor. Abbreviations: COD, causes of death.

Full size image
Fig. 5
figure 5

Cause-specific hazard ratios according to tumor behavior and tumor type. A, All-cause death; B, Primary CNS tumor death; C, Competing event death. Note: Meningiomas_NonMalig (nonmalignant meningiomas) was used as a reference group (HR = 1). Abbreviations: COD, causes of death; No., number; HR, hazard ratio; CI, confidence interval.

Full size image
Table 2 Cause-specific hazard ratios of primary tumor death and competing event death grouped by tumor behavior.
Full size table

Ethics approval and consent to participate

The patient data in the SEER database are anonymized and available freely. Therefore, ethical consent was not required, and patient informed consent was waived for this study according to the Medical Ethics Committee of West China Hospital of Sichuan University. Research involving human participants and/or human data was performed in accordance with the ethical principles established in the “Declaration of Helsinki”. The research plan was approved by the Medical Ethics Committee of West China Hospital of Sichuan University (No. 2024 − 1355).

Results

Patient characteristics

A total of 324,817 patients with CNS tumors (including 207,273 nonmalignant and 117,544 malignant tumors) were enrolled. A summary of the clinical features and treatments of CNS tumors is shown in Table 1. Among these patients, 303,345 (93.4%) were adults, whereas only 21,472 (6.6%) were children. Among all patients, 41.8% (135,908) were over 65 years (elderly), and more than half (59.0%) were female. According to the tumor behavior and types of WHO classification, the tumor types of nonmalignant tumors included 149,238 (45.9%) nonmalignant meningiomas, 10,708 (3.3%) ependymal neuronal tumors, 33,570 (10.4%) cranial paraspinal tumors, and 13,757 (4.2%) other nonmalignant tumors. The tumor types of the malignant tumors included 44,389 (13.7%) LGG, 60,567 (18.6%) GBM&DMG, and 12,588 (3.9%) other malignant tumors. For the surgery types, 125,183 (38.5%) patients received STR/GTR, 34,932 (10.8%) received biopsy/PR, and 161,724 (49.8%) did not receive surgery. Among the patients receiving adjuvant therapies, 73,068 (22.5%) received radiation therapy, and 55,977 (17.2%) received chemotherapy. In addition, a comparison of the clinical features revealed significant differences between nonmalignant and malignant CNS tumors. Children’s tumors are less common than adult tumors are, but the proportion of malignant tumors (13.4%) is greater than that of nonmalignant tumors (2.8%) in children. Nonmalignant tumors have a female orientation (67.4%), whereas malignant tumors have a male orientation (55.9%).

Proportion of COD

Among all patients with CNS tumors, 23.70% died from primary tumors, and 18.52% died from competing risk events (Fig. 2A). Subgroup analyses revealed that a lower percentage of patients with brain tumors than those with other CNS tumors were alive (51.97% vs. 80.95%). In contrast, patients with brain tumors accounted for a greater proportion of primary tumor deaths (28.93% vs. 2.86%) and competing event deaths (19.10% vs. 16.19%) than did those with other CNS tumors. In terms of age group, adult patients accounted for a greater proportion of all-cause deaths (including primary tumor deaths and competing event deaths) than did children (primary tumor deaths: 23.86% vs. 21.53%; competing event deaths: 19.48% vs. 4.80%). In particular, patients with malignant tumors accounted for a significantly greater proportion of all-cause deaths (70.73% vs. 26.05%) and primary tumor deaths (61.48% vs. 2.28%) than patients with nonmalignant tumors did, whereas patients with nonmalignant tumors accounted for a greater proportion of competing event deaths than patients with malignant tumors did (23.77% vs. 9.25%). For the competing event death of all CNS tumors, CVD death (27.32%) was the most common COD, followed by respiratory death (14.04%), subsequent malignant cancers (11.45%), and others (Fig. 2B). Compared with patients with malignant tumors, those with nonmalignant tumors had greater proportions of CVD and respiratory deaths. Patients in the adult, male, and nonmalignant subgroups accounted for a greater proportion of CVD deaths than did their counterparts.

Further subgroup analyses based on tumor behavior revealed significant disparities in variables, including tumor site, age group, and sex (Supplemental Fig. S1). Patients with malignant tumors accounted for a significantly greater proportion of all-cause deaths and primary tumor deaths than patients with nonmalignant tumors did in the total population and subgroups of variables, including tumor site, age group, and sex. In contrast, patients with nonmalignant tumors accounted for a greater proportion of competing event deaths than patients with malignant tumors in the total cohort and subgroups of variables. Regarding the competing event death, the proportions of the causes of death also differed between patients with nonmalignant and malignant tumors.

Cumulative mortality

We analyzed the cumulative mortality (all-cause and cause-specific mortality) against the months after primary CNS tumors in the total population and different subgroups (Fig. 3). In the total population of patients with CNS tumors, the mortality of those with primary tumors was greater than that of those with competing events throughout the early and middle periods of follow-up. In contrast, the competing risk mortality rate exceeded the death rate due to the primary tumor during the late follow-up period (Supplemental Fig. S2). Subgroup analyses revealed that the mortality of primary tumors was greater than that of competing events for brain tumors throughout the follow-up period, whereas the mortality of competing events was greater than that of primary tumors for other CNS tumors (Fig. 3A). In children or male patients, the mortality of primary tumors was greater than that of competing events for brain tumors throughout the follow-up period. In contrast, in adult or female patients, the mortality associated with competing events was greater than that associated with primary tumors during late follow-up (Fig. 3B-C). The analysis results for the competitive risk model revealed that all-cause deaths were attributed mainly to competitive risk deaths among patients with nonmalignant CNS tumors. In contrast, all-cause deaths were attributed mainly to primary tumor deaths among patients with malignant CNS tumors (Fig. 3D). Among the above variables, the disparities in tumor behavior for all-cause and cause-specific mortality were the most significant.

Furthermore, stratified analysis of tumor behavior revealed differences in cumulative mortality between subgroups in terms of clinical features, including tumor site, age group, sex, and coexisting tumors (Fig. 4). For patients with nonmalignant tumors, the cumulative mortality due to competing events was significantly greater than that due to primary tumors, regardless of the different subgroups of clinical features (Fig. 4A-D). For patients with malignant tumors, however, the cumulative mortality due to primary tumors was significantly greater than that due to competing events in different variables (Fig. 4E-H). Notably, comparative analysis of subgroups revealed significant differences in cumulative mortality rates due to primary tumor death between the tumor site (brain vs. other CNS) and age groups (child vs. adult), whereas there were no significant differences in sex (male vs. female) or coexisting tumors (No vs. Yes).

In addition, all-cause and cause-specific mortality varied by tumor type (Supplemental Fig. S3). Among patients with nonmalignant tumor types (including nonmalignant meningioma, ependymal neuronal tumors, cranial paraspinal tumors, and other nonmalignant tumors), all-cause mortality was attributed mainly to competitive risk death. Among patients with malignant tumor types (including LGG, GBM & DMG, and other malignant tumors), all-cause mortality was attributed mainly to primary tumor death.

Cause-specific hazard ratios

Cause-specific hazard ratios for all-cause death and cause-specific death were determined via multivariable Cox proportional hazard model (Supplemental Table S3). For specific populations with CNS tumors, such as the adult population, male sex, brain tumor site, malignant tumor behavior, presence of coexisting tumors, and no surgery, the cause-specific hazard ratios for all-cause death, primary tumor death, and competing event death were greater (HR > 1, P < 0.001) than those of their counterparts. Notably, regarding all-cause death, patients with malignant tumors were at higher cause-specific mortality risk compared to those with nonmalignant tumors (HR = 5.4, 95% CI: 5.32–5.48, P < 0.001). Moreover, patients with malignant tumors were at higher cause-specific mortality risk for primary tumor death (HR = 47.3, 95% CI: 45.8–48.9, P < 0.001) and competing event death (HR = 1.05, 95% CI: 1.02–1.08, P < 0.001). Further stratified analyses based on tumor behavior revealed significant differences in variables between nonmalignant and malignant tumors (Table 2). The disparities in all-cause and cause-specific mortality were significantly different between patients with nonmalignant and malignant tumors.

The cause-specific hazard ratios of all-cause death and cause-specific death varied by the type of CNS tumor. Compared with patients in the reference group (nonmalignant meningiomas), patients in the nonmalignant tumor groups (excluding Other_Nonmalignant) had lower cause-specific hazard ratios for all-cause death, primary tumor death, and competing event death. For patients with malignant tumors, the cause-specific hazard ratios for all-cause death and primary tumor death were greater than those in the reference group. However, the cause-specific hazard ratio for competing event death was lower than that in the reference group. Among these CNS tumors, the GBM & DMG type had the highest cause-specific hazard ratios for all-cause death and primary tumor death, and nonmalignant meningioma had the highest cause-specific hazard ratios for competing event death (Fig. 5).

Discussion

In this study, we comprehensively investigated the tumor behavior disparities in all-cause and cause-specific mortality via a larger population-based cohort of CNS tumors. This study revealed that 23.70% of all patients with CNS tumors died from primary tumors, and 18.52% of all patients with CNS tumors died from competing risk events. The proportion of COD in subgroups varied according to clinical characteristics, especially tumor behavior. The difference in prognosis among different clinical features is consistent with previous reports in the literature on CNS tumors26,27,28,29. A recent study7 on CVD risk among 21 non-metastatic cancers revealed that 40.2% of deaths were from primary neoplasms and that the other deaths were from competing risk events. Compared with regional cancers, localized cancers accounted for a lower proportion of primary neoplasms (25.5% vs. 59.2%) and a higher proportion of deaths from competing risk events (74.5% vs. 40.8%)7. In this study, CVD death (27.32%) was the most common cause of death among all competing event deaths related to CNS tumors. An increasing number of studies have reported that CVD mortality is the main competing cause of death in various types of cancer, suggesting the prognostic role of competing risk events6,7,11,30,31,32,33.

Furthermore, we compared the cumulative mortality of all-cause death and cause-specific death in the total population and different subgroups. For example, according to the tumor behavior of CNS tumors, all-cause deaths were attributed mainly to competitive risk deaths for patients with nonmalignant tumors, whereas all-cause deaths were attributed mainly to primary tumor deaths for patients with malignant tumors. Similarly, competing risk analysis for different tumor types confirmed significant differences in cumulative mortality according to tumor behavior. The study7 on CVD death in 21 non-metastatic cancers classified all cancer patients into a high competing risk group (14 individual cancers) and a low competing risk group (7 individual cancers). The results revealed that primary neoplasms conferred the highest cumulative mortality of competing risk events among patients with localized cancers, which had a relatively better survival outcome than did regional cancers7. Additionally, cardiotoxicity or other side effects of tumor/cancer treatment may increase the risk of competing events, including CVD death7,30,34,35,36. In our study, patients with nonmalignant tumors, who had better survival than those with malignant tumors, also had the highest cumulative mortality of competing risk events. Therefore, patients with nonmalignant tumors or localized cancers with longer survival are generally associated with a higher cumulative incidence of competing events than patients with malignant tumors or regional cancers7,30. In addition, we noticed that the proportion of non-malignant/malignant changes dramatically from 2000 to 2007 period to later periods (~ 1:1 to ~ 2:1). This phenomenon may be related to the increase in the detection rate of benign intracranial tumors due to advancements in magnetic resonance technology and the growing willingness of patients to undergo treatment1,37.

The increased risk of competing death in patients with nonmalignant or malignant tumors may be multifactorial. Interestingly, stratified analyses on the basis of tumor behavior also revealed different cause-specific hazard ratios among patients with nonmalignant and malignant CNS tumors in this study. A previous study analyzed the cause-specific hazard ratios of 21 non-metastatic cancers and revealed that the risk of death from CVD varied by cancer site7. In our study, we compared the cause-specific hazard ratios among different types of CNS tumors, and the cause-specific hazard ratios for all-cause death, primary tumor death, and competing event death varied by tumor behavior and tumor type. The heterogeneity in the risk of competing event death by tumor type may be multifactorial, including heterogeneity in pathogenesis, age distribution, and prognosis; heterogeneity in comorbidities; and heterogeneity in treatment strategies for different tumors7,8,31,35,38,39.

However, this study had several limitations. First, the data’s wide time period (2000–2021) may have led to a confounding effect from calendar year periods, as the tumor classification guidelines and treatment changed over time. Nonetheless, our results described the risk of death from a broad category (meningioma, glioma, etc.) rather than a specific subcategory. Second, the SEER database did not include variables related to competing events, such as systemic diseases, preventing further exploration. Finally, the lack of information on the recurrence and metastasis of CNS tumors during follow-up in the SEER database limits further exploration. As a result, it is necessary to investigate the cumulative mortality of all-cause death and cause-specific death in large-sample, multiregional, ethnically diverse population cohorts.

Conclusion

This study investigated the tumor behavior disparities in all-cause and cause-specific mortality via a larger population-based cohort. All-cause death was attributed mainly to primary tumor death among patients with nonmalignant tumors, whereas it was attributed mainly to competing event death among patients with malignant tumors. The cause-specific mortality from primary tumor risk and competing event risk varied by behavior and type of CNS tumors. This study provides robust evidence for an in-depth understanding of the tumor behavior disparities in cumulative mortality among patients with CNS tumors.