Development and validation of nomogram model predicting overall survival and cancer specific survival in glioblastoma patients

Abstract

Background: Identifying the incidence and risk factors of Glioblastoma (GBM) and establishing effective predictive models will benefit the management of these patients.

Methods: Using GBM data from the Surveillance, Epidemiology, and End Results (SEER) database, we used Joinpoint software to assess trends in GBM incidence across populations of different age groups. Subsequently, we identified important prognostic factors by stepwise regression and multivariate Cox regression analysis, and established a Nomogram mathematical model. COX regression model combined with restricted cubic splines (RCS) model was used to analyze the relationship between tumor size and prognosis of GBM patients.

Results: The incidence of GBM has been on the rise since 1978, especially in the age group of 65-84 years. 11498 patients with GBM were included in our study. The multivariate Cox analysis revealed that age, tumor size, sex, primary tumor site, laterality, number of primary tumors, surgery, chemotherapy, radiotherapy, systematic therapy, marital status, median household income, first malignant primary indicator were independent prognostic factors of overall survival (OS) for GBMs. For cancer-specific survival (CSS), race is also independent prognostic factors. Additionally, risk of poor prognosis increased significantly with tumor size in patients with tumors smaller than 49 mm. Moreover, our nomogram model showed favorable discriminative ability.

Conclusion: At the population level, the incidence of GBM is on the rise. The relationship between tumor size and patient prognosis is still worthy of further study. Moreover, the proposed nomogram with good performance was constructed and verified to predict the OS and CSS of patients with GBM.

Keywords: Cancer-specific survival; Glioblastoma; Overall survival; Prognostic nomogram; SEER.

Fig. 1

Flowchart of the inclusion and exclusion criteria for this study

Fig. 2

Joinpoint analysis of the incidence rates of glioblastoma in the U.S. between 1975 and 2019. (*) Indicates the annual percent change (APC) that is significantly different from zero (P < 0.05)

Fig. 3

Establishment of overall survival (OS) and cancer-specific survival (CSS) nomograms. A Construction of OS nomogram; B construction of CSS nomogram

Fig. 4

Calibration plot of the nomogram for predicting 0.5-,1-, 2-, 3-, 5- and 8‐year overall survival (OS) and cancer-specific survival (CSS) in training cohort and validation cohort, respectively. A 0.5-,1-, 2-, 3-, 5- and 8‐year OS in training cohort; B 0.5-,1-, 2-, 3-, 5- and 8‐year CSS in training cohort; C 0.5-,1-, 2-, 3-, 5- and 8‐year OS in validation cohort; D 0.5-,1-, 2-, 3-, 5- and 8‐year CSS in validation cohort

Fig. 5

Receiver operating characteristics curve (ROC) comparison of overall survival (OS) and cancer-specific survival (CSS) nomogram in training cohort and validation cohort, respectively. A 0.5-,1-, 2-, 3-, 5- and 8‐year ROC of OS nomogram using training cohort; B 0.5-,1-, 2-, 3-, 5- and 8‐year ROC of CSS nomogram using training cohort; C 0.5-,1-, 2-, 3-, 5- and 8‐year ROC of OS nomogram using validation cohort; D 0.5-,1-, 2-, 3-, 5- and 8‐year ROC of CSS nomogram using validation cohort

Fig. 6

Kaplan–Meier survival curves for glioblastoma patients. A–F 0.5-A,1-b, 2-C, 3-D, 5-E and 8‐F year overall survival (OS) in training cohort; G–L 0.5-G,1-H, 2-I, 3-J, 5-K and 8‐L year cancer-specific survival (CSS) in training cohort

Fig. 7

Cumulative risk curves for glioblastoma patients. A–F 0.5-A,1-B, 2-C, 3-D, 5-E and 8‐F year overall survival (OS) in training cohort; G–L 0.5-G,1-H, 2-I, 3-J, 5-K and 8‐L year cancer-specific survival (CSS) in training cohort

Fig. 8

Adjusted cubic spline models showing association between tumor size and hazard ratio for overall survival (A) and cancer-specific survival (B). The solid line and red zone represent the estimated odds ratio and its 95% confidence interval