High-grade glioma (HGG), such as World Health Organization (WHO) grade III and IV, is the most common type of intracranial malignant tumor. The inherent high heterogeneity of HGG contributes to poor therapeutic efficacy as the dominant factor, resulting to the high mortality rates and rapid progression. The median survival time of grade IV patients is only 15 months (1–3). In addition, in clinical practice, the overall survival (OS) for HGG patients who receive the same treatment may differ significantly at the individual level (4). Therefore, summarizing the different characteristics and identifying effective prognostic factors based on retrospective reviews could stratify the patients for personalized follow-up regimen development and further individualized management improvement of prognosis. Clinically, adequate and reliable prognosis prediction for HGG patients is urgently need but remains challenging.

Inflammation plays a crucial role in tumor microenvironment and tumor progression, namely, glioma. Therefore, inflammatory biomarkers, such as neutrophil-to-lymphocyte ratio (NLR), platelet-to-lymphocyte ratio (PLR), lymphocyte–monocyte ratio (LMR), lactate dehydrogenase (LDH), may not only reflect inflammation status but also indicate glioma progression. Currently, many factors including serum inflammatory factors, have been identified as effective prognostic factors in HGG (5–8). However, the predictive value of serum inflammatory biomarkers in glioma prognosis remains controversial (9). Therefore, the role of serum inflammatory biomarkers needs to be further investigated due to the controversial results. In addition, considering the high heterogeneity of HGG, a single risk factor may be limited in precisely and effectively predicting prognosis. A multivariable model by comprehensively integrating the clinical, laboratory, and pathological risk factors may be more effective and reliable for prognosis prediction.

Hence, our study aims to investigate the predictive value of serum inflammatory biomarkers in HGG, and further develop and validate a risk stratification model for HGG based on risk factors extracted from clinical, laboratory, radiological, and pathological information.

Currently, the maximal safe resection combined with concurrent chemoradiotherapy is considered as the most effective treatment regimen for HGG. Even so, patients who underwent the same treatment regimen have exhibited significant difference in their prognosis. With the rapid development of precision medicine, individualized treatment and follow-up strategy were urgently needed. Precise and reliable survival models could contribute to guide clinicians in formulation of treatment plan and management of individual patients. Among different tumor types, nomograms integrating multiple prognostic factors into a reliable tool has been widely applied in predicting survival of patients (10, 11).

Considering the interactions among the related factors may better reflect the complexity of malignant tumor (12). In this study, we attempted to incorporate factors that may have an effect on survival. In the pursuit of simplicity, serum inflammatory biomarkers in our study were handled by a categorical manner as they are actually continuous variables. This may result in a considerable loss of statistical power with introduction of bias in the multiple regression analysis (13). However, considering those inflammatory biomarkers as categorical variables may be more appropriate in guiding process of clinical decision-making as continuous variables could not necessarily increase the prediction performance (14). Therefore, the serum inflammatory biomarkers in our study were divided as binary variables and valued as 0 or 1 based on their optimal cut-off value instead of absolute value.

Our study utilized the Lasso regression analysis and SVM to analyze the independent prognostic risk factors left after multivariable Cox regression analysis. Comparing to the Cox regression analysis, the Lasso regression analysis is a new approach for variables selection by analyzing all variables at the same time. It could minimize the coefficients and produce some coefficients which are exactly zero. Hence, it decreases the estimation variance and provides appropriate prognostic factors with non-zero coefficients (15). In addition, as a widely-used machine learning-based analysis (16), SVM was also utilized in our study to screen variables. Based on the multivariable Cox regression analysis, Lasso regression analysis, and SVM, PLR value, tumor grade, IDH mutation, and Ki-67 index were independent prognostic factors for OS time. In addition, the results showed that GTR was not significant associated with OS in HGG patients, while advanced age lost its significance in the multivariable Cox analysis. This may be caused by the relatively small sample size and high heterogeneity of the disease.

All the above independent factors were considered as potential prognostic indicators of glioma or other malignancies in the previous studies. In malignant tumors, the evoked inflammatory responses and inflammatory cytokines and mediators regulated by tumor microenvironment played an essential role in tumor progression (17–19). PLR served as a biomarker of systemic inflammation and has been considered as an independent prognostic factor in many solid tumors (20). The IDH genotype, an important genetic hallmark of glioma, has been definitely confirmed its prognostic value (21). There is no doubt that the prognosis of glioma is negatively associated with the tumor grade due to its important indication for malignancy of glioma (22). As a proliferation index, Ki-67 has been widely considered as an effective predictor of prognosis and adjuvant therapy responsiveness (23, 24).

Inflammation response could weaken the immune response to malignant tumor (25), permitting serum inflammatory biomarkers serve as prognostic factors. However, whether they are valuable and how to comprehensively utilize these predictors to predict the prognosis of patients with HGG remains unclear. Thus, based on the 4 independent prognostic factors, we established two models. DCA, IDI, and NRI were used to evaluate the clinical utility of the predictive models. Time-dependent ROC compared the predictive performance among the two different models at various time points instead of a fixed time to determine the final model, which was another advantage of our study. Based on time-ROC analysis, DCA, IDI, and NRI, the model including PLR level showed better performance than another model without the two biomarkers, which was selected as the final model to establish a nomogram.

In our study, the discrimination and calibration abilities of the nomogram were evaluated to ensure its accurate application. Herein, we proposed that the training cohort and validation cohort were homogeneous and comparable based on the results of statistical analysis. C-index could reflect the probability of patients with shorter OS time ranked with higher risk of death according to the model in a random selection process. Therefore, the C-index comprehensively considered both the occurrence of the event and the follow-up duration, which is particularly suitable for time-to event analysis. In the entire cohort, the nomogram showed a strong discriminating capability concluded by high C-index value, which were 0.747 (95%CI, 0.706–0.788). The DCA curves and ROC curves showed that the nomogram had excellent clinical utility and predictive performance. In addition, calibration curve is an evaluation method of the agreement between the observed and predicted prognosis in a predictive model. The calibration curves showed good consistency between the predicted and observed survival probability, indicating the ideal repeatability and reliability of the nomogram.

The most important finding of our study is confirmation of the prognostic value of serum inflammatory biomarkers in HGG patients and further construction of a nomogram that could accurately and reliably predict prognosis of HGG patients in a relatively small sample. Functional status, inflammatory condition, adjuvant treatment, pathological grade and immunochemical features were all taken into account when the nomogram for OS prediction was established. By using a simple and intuitive graphical representation, the nomogram enabled its application in clinical care of individual patient. Specifically, total points of nomogram based on the 4 associated factors could be calculated for individual patients. Thus, the corresponding estimated survival probability could be obtained and used to guide clinicians for making individualized follow-up management. Identifying high-risk patients based on precious prediction and taking timely measures may improve their outcome.

The present study also several inherent limitations that should be discussed. First, as a retrospective and single-institution study with a relatively small sample, it may have been subject to interference and selection bias. Second, some clinical, laboratory, and immunohistochemical factors were not included in our study due to the lack of examination or incomplete medical records. Particularly, some molecular profiling may have impacts on glioma prognosis but could not be available due to the retrospective design. Third, the validity and generalizability of the nomogram required to be validated in other independent patient groups. Fourth, the training and validation samples were not in the same time frame which may cause bias due to advancements in medical practice. Final, though the predictive model consisted of adequate prognostic factors is beneficial to HGG management, it should be well aware that the inherent high heterogeneity in HGG is still hard to settle. Further prospective study is needed to minimize the limitations in our study.

Taken together, our study confirmed the prognostic value of serum inflammatory biomarkers and established a comprehensive scoring system for the OS prediction in HGG patients. The clinical nomogram could assist clinicians when making survival predictions and follow-up management in individual HGG patient. Further studies with larger sample size are required to verify our findings.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by the local ethics committee of the First Affiliated Hospital of Fujian Medical University. The ethics committee waived the requirement of written informed consent for participation.

D-ZK, C-YD, and H-HY performed and designed the study, and wrote the manuscript. X-YC, D-LP, J-HX, YC, W-FX, J-YC, Z-YW, and Y-XL collected, interpreted, and analyzed the data. All authors contributed to the article and approved the submitted version.

This study was supported by grants from the Excellent talent project (No. YJCRC-B-KDZ2021), the Startup Fund for Scientific Research of Fujian Medical University (No. 2020QH1035), the project of improving the diagnosis and treatment ability of complicated diseases (No. PT-YNBZW2018), and China Postdoctoral Science Foundation (No.2020M682070).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.