Exploration of the prognostic value of the resection of adult brainstem high-grade glioma based on competing risk model, propensity score matching, and conditional survival rate

Background Adult brainstem high-grade glioma (HGG) is a refractory disease, and the treatment strategy of resection is still controversial. Objective To investigate the prognostic value of brainstem HGG resection in adults. Methods We collected 126,386 samples from the Surveillance, Epidemiology, and End Results (SEER) database between 1998 and 2016, and screened 973 patients diagnosed with adult brainstem HGG, who were in turn, grouped into 899 cases of non-resections and 74 cases of resections. Competing risk models were used to screen independent prognostic factors. Propensity score matching (PSM) was performed to reduce the influence of confounding factors. Conditional survival (CS) rate was considered to evaluate the changes in overall survival (OS) and disease-specific survival (DSS) of patients with HGG over time. Results Based on the competing risk model and PSM, univariate analysis showed that age ≥ 45 years and male gender were poor prognostic factors for adult brainstem HGG. No previous history of glioma was a beneficial factor. Multivariate analysis revealed only the absence of a history of glioma to be a favorable prognostic factor. Considering the CS rate of the resection group, after the patient had survived for 3 years, the OS and DSS remained unchanged at 100% during the fourth and fifth years, whereas in the non-resection group, the OS and DSS of the patients were 82% and 74%, respectively. Conclusion Adult brainstem HGG resection has a poor prognosis in the early stage; however, patients have a potentially significant survival benefit after 3 years of survival. Supplementary Information The online version contains supplementary material available at 10.1007/s10072-022-06557-z.


Introduction
Brainstem high-grade glioma (HGG), a rare and aggressive tumor, occurs mostly in children [1,2] and rarely in adults, accounting for 1-2% of all adult cases of central nervous system tumors [3][4][5]. There are crucial nerve centers in the brainstem, such as the cardiovascular movement, respiratory, and swallowing centers. Due to its special anatomical location, surgical resection of brainstem HGG is difficult, making the treatment of brainstem HGG challenging [6,7]. Studies have reported that the median survival rate of patients that underwent surgical intervention was only 11 months, and the overall survival (OS) rate was very low [8]. Moreover, the patient's age, gender, radiotherapy, chemotherapy, tumor size, and scope of surgical resection have been reported to be related to HGG prognosis [9][10][11][12]. However, OS in adult patients with HGG and related survival factors are still controversial.
Competing events of non-cancer death, such as heart disease, cerebrovascular disease, and influenza, do exist in patients with cancer [9]. Therefore, compared with the Cox survival analysis, the use of a competing risk model would facilitate accurate assessment of the association of predictor variables with outcome events. Moreover, confounding factors are often ubiquitous in observational research. The current study aimed to achieve "randomization" through propensity score matching (PSM) to control the influence of confounding factors on research conclusions. Additionally, we found the conditional survival rate (CS) of patients to change over time. CS can quantify the degree of improvement in the prognosis of patients over time, facilitating the adjustment of long-term follow-up strategies.
Exploration of the prognostic value of expanded resection of adult brainstem HGG based on a single medical center often leads to a small sample size and low statistical power [13][14][15]. The patient selection bias based on hospital-centered data is often greater than that based on the Surveillance, Epidemiology, and End Results (SEER) data. With the above considerations, the current study aimed to include adult patients with brainstem HGG through the SEER database, based on the competitive risk model, PSM, and CS, to study the prognostic value of resection of adult brainstem HGG.

Baseline data
We collected 126,386 samples from the SEER database. The criteria for inclusion and exclusion were as follows: the histological code of the International Classification of Diseases for Oncology Third Edition (ICD-O-3) was 9380/9401/9440/9441/9442/9451, the age was ≥18 years old, the tumor occurred in the brainstem, and 973 people finally met the criteria. The main variables in this study are non-resection (including no operation/biopsy) and resection (including subtotal resection/total resection). The included covariates were age, gender, race, marital status, diagnosis time, radiotherapy/chemotherapy, tumor size, etc. (Table 1). The SEER database is the authoritative and public source of information about cancer incidence, mortality, and survival rates in the USA. We used SEER-Stat software (version 8.3.5) to download patients' clinical data. The data are publicly available and do not involve privacy of patients; therefore, so no ethical review was required.

Competing risk model
In this study, we chose death from brainstem HGG as the end event while deaths caused by other factors were regarded as competing risk events. Cumulative incidence function (CIF) was used to estimate the cumulative occurrence probability of an outcome event, which was then used to process the survival data from multiple endpoints and competing risk events. The CIF for death due to brainstem HGG and other competing risk events was calculated and was grouped by age, race, chemotherapy, etc. Using cmprsk's R package, we drew a CIF curve for each variable, and performed Gray's test to identify the difference between brainstem HGG and non-brain stem HGG deaths in CIF. Subsequently, for multivariate competing risk survival analysis, we constructed the Fine-Gray proportional sub-distribution hazards model, and used cmprsk and risk regression to predict the potential risk factors of death from brainstem HGG and death from non-brain stem HGG events.

Propensity score matching
This study used PSM to balance the clinical data between the resection group and the non-resection group in the SEER cohort, including the following baseline characteristics: age, gender, race, marital status, time of diagnosis, chemotherapy, tumor size, and past history, to achieve the effect of retrospective randomization. First, we used the multiple logistic regression model to calculate the propensity score (PS) of each patient according to surgery type (resection and non-resection). Second, we used the MatchIt package in R software to analyze the data, set the caliper value to 0.02, and evaluated the effect according to the standardized mean difference (SMD) and P value.

Conditional survival rate
Conditional survival rate was estimated from clinical data according to the Berkson-Gage method, and its variance and confidence interval were derived according to the binomial distribution theory. This study mainly analyzed the binary variables of resection and non-resection, the overall survival rate (OS), and disease-specific survival rate (DSS) after 0, 1, 2, 3, 4, and 5 years following surgery. The probability that a patient who has survived x years after the initial treatment will survive for another n years was expressed as CS(

KM curve of patient OS and DSS
From the KM survival curve of OS (Fig. 1), according to the median survival time, the prognosis of age < 45 years (48 months) is better than that of age ≥ 45 years (12 months) (P = 0.00); no chemotherapy (35 months) was better than chemotherapy patients (16 months) (P = 0.00). Surgically, non-resection (23 months) was significantly better than resection (13 months) (P = 0.00). Similarly, the KM curve of DSS is shown in Fig. A.1.

Screening factors affecting prognosis based on competitive risk model
Due to the impact of non-glioma deaths events, the Cox survival analysis (

CS of resection after PSM
Based on conditional survival, after resection of adult brainstem HGG, patient survival rate remained stable during 3-5 years (Fig. 3A, C). In terms of OS, after 1 year of resection, the survival rate for 3-5 years was 39%; after 2 years of survival, the survival rate for 3-5 years was 72%. After 3 years of survival, the survival rate was 100% at 4-5 years (Fig. 3A). Also in terms of DSS, after 1 year of resection, the survival rate during the 3-5 years was 46%; after 2 years of survival, the survival rate during the 3-5 years was 78%. After 3 years of survival, the survival rate at 4-5 years was 100% (Fig. 3C). However, after non-resection, the survival rate decreased year by year (Fig. 3B, D). In particular, after 3 years of survival, the DSS in the fourth and fifth years were 82% and 74%, respectively (Fig. 3D); both were lower than the DSS in the same period after resection (100%, 100%) (Fig. 3C).

Background on the treatment of adult brainstem HGG
Adult brainstem HGG are highly malignant and associated with a low survival rate; they have always been regarded as the most difficult to treat [16]. They mostly manifest as dizziness, nausea, and vomiting, along with symptoms such as limb numbness, facial paralysis, and dysphagia [17][18][19].
With the development of neuronavigation [20,21], electrophysiological monitoring and diffusion-weighted magnetic resonance imaging technology [22], convection-enhanced separation and application of gamma knife and cyberknife [23][24][25], histological diagnosis [26], continuous upgradation of radiotherapy, and that of anti-angiogenesis, monoclonal antibodies, and other chemotherapeutic drugs [27,28], the treatment of adult brainstem HGG has become more feasible. However, the overall survival of patients with adult brainstem HGG and related survival factors are still unclear. Our current study revealed the following: first, age, tumor size, and previous history of glioma are closely related to adult brainstem HGG. Second, in terms of treatment measures, there is no significant statistical difference in improving the overall survival rate of patients with radiotherapy and chemotherapy. Third, and most importantly, for adults with brainstem HGG resection, patients have a significant survival benefit after 3 years of survival.

The impact of competing events on traditional survival analysis
If there are more than 10% competing events, such as heart disease and car accidents in cancer patients, the use of Cox survival analysis may lead to an incorrect estimation. In this data, we found 20.51% (129/629) of the deaths to be due to non-HGG causes. Therefore, use of a competing risk model would facilitate the accurate assessment of the association between the surgical approach and the prognosis of HGG; multivariate analysis found age ≥ 45 years, chemotherapy, and tumor diameter ≥ 20 mm to be poor prognostic factors. This was basically consistent with the results of Dey et al. [29]. At the same time, compared with low-grade glioma, the more aggressive the tumor, the worse the prognosis of patients with brainstem glioma [30,31]. Furthermore, factors associated with improved 5-year overall survival included female gender, higher income, and fewer comorbidities [32].

Effects of radiotherapy on adult brainstem HGG
Radiotherapy is one of the main treatment methods for tumors. However, studies have reported that radiotherapy is significantly related to an increased risk of secondary malignant tumors [33]. For sub-ependymal tumors, radiotherapy had no statistically significant impact on overall survival [34]. Nevertheless, a study by Reithmeier et al. [35] suggested that postoperative radiotherapy is the cornerstone of treatment and can reduce the risk of death by 0.4 times. Compared with radiotherapy alone, postoperative radiotherapy combined with chemotherapy has a significant survival benefit for HGG patients with WHO grade IV (HR: 0.46, 95% CI [0.28-0.76], P = 0.00) [36]. For HGG of the pons, radiotherapy remains the standard treatment, although it only provides a survival advantage of 3 to 4 months [37]. But, our study showed that radiotherapy (HR = 1.05, 95% CI [0.59-1.86], P = 0.87) had no statistically significant effect on the prognosis of adult brainstem HGG. Research by Hu et al. [38] also confirmed the same.

Effect of resection on adult brainstem HGG
Whether an adult brainstem HGG would need surgery and the scope of surgical resection have remained controversial.
Our study found that resection of adult brainstem HGG has a higher risk of early death, although a stable period is reached  [40] believed that tumor grade is the only factor that has a statistically significant impact on survival time (P = 0.00), whereas younger age, better physical status, and resection surgery showed prolonged survival. However, owing to the special brainstem site, surgery itself is highly risky, and the incidence of complications, such as postoperative bleeding, infection, edema, cerebral hernia, and respiratory insufficiency, is high (36.4%) [35,41]. If there is a clear boundary between the tumor, normal cerebellum, and tentorium in imaging results, the tumor can be biopsied, and its exogenous part, the brainstem surface, and the accessible Fig. 3 The CS of resection and non-resection after PSM. A Overall conditional survival rate after resection; B overall conditional survival rate after nonresection; C disease-specific conditional survival rate after resection; D disease-specific conditional survival rate after non-resection. PSM, propensity score matching; CS, conditional survival part can be roughly/completely removed. Total resection is not possible, since high-grade tumors have strong invasiveness [42].

Limitations and strengths
First, it is a retrospective study with inherent limitations. With modern medical technology development, the survival rate of patients with adult brainstem HGG will also change. Secondly, the SEER database lacked detailed information about the patient's physical condition, such as family history and specific parts of tumor (midbrain, pons, or medulla oblongata), especially performance status (PS), is often positively correlated with the prognosis of patients. Thirdly, the database still lacks data of early postoperative mortality, which is of great significance in evaluating the safety of resection. Nevertheless, considering the huge sample size provided by the SEER database, the statistical results would still be very meaningful. In addition, the CS could be incorporated as a factor of survival time, which could assess the prognosis in survivors with HGG more accurately over time.

Conclusion
The purpose of this study was to investigate the prognostic value of brainstem HGG resection in adults using competing risk models, PSM, and CS based on the SEER database. The data show that the DSS of adult brainstem HGG is 82% and 74% in 1-2 years after non-resection survival for 3 years, but there is a stable survival rate (100%) in 1-2 years after 3 years of resection survival, which will be more effective in helping young patients cope with future uncertainty. Therefore, these data support neurosurgeons to perform maximal safe resection of adult brainstem HGG when available.
Funding National Natural Science Fund, Grant/Award Number: U1804199.

Declarations
Ethical approval This study was approved by the Ethics Committee of the First Affiliated Hospital of Xinxiang Medical College. The clinical study registration number is: EC-021-159.

Conflict of interest
The authors declare no competing interests.
Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http:// creat iveco mmons. org/ licen ses/ by/4. 0/.