Introduction: Gliomas represent a challenging clinical entity due to their heterogeneous nature. We examine a large national database in order to identify the prognostic factors that contribute the most to patient survival, and consequently develop a predictive model for survival.
Methods: NCI SEER program was used to retrieve the records of 27,325 subjects diagnosed with glioma between 1985 and 2008. The primary outcome measure was survival following diagnosis. Parameters examined for effects on this outcome included: age at diagnosis, gender, race, histology, tumor location and size, surgery, radiation, and geographic region. Cox proportional hazard regression and Kaplan-Meier survival and risk functions were used to examine and compare these parameters. A predictive algorithm was developed and tested.
Results: Age at diagnosis was the most significant prognostic factor (X2=5186.4,p<0.0001), with an increase in hazard risk of 3.4% per year. Second was histology (X2=4134.7,p<0.0001), with pilocytic astrocytoma and oligodendroglioma having 84.9% and 80.5% decreased in mortality risk respectively, compared to glioblastoma. Surgery was next (X2=815.8,p<0.0001). Relative to gross total resection, no surgery, biopsy, and subtotal resection were associated with 58.7%, 82.6%, and 14.5% increases in hazard respectively. Location (X2=362.2,p<0.0001) was also significant, with brainstem having a 30.2% increased risk compared to cerebrum. Lack of radiation (X2=359.9,p<0.0001) increased hazard by 36.8%. Females had a 5.3% reduced hazard (X2=16.5,p<0.0001). Unexpectedly, geographic region was also significant (X2=20.2,p=0.0002), with Northern Plains and Southwest showing 4.9% and 7.7% hazard increases respectively, compared to East Coast (X2=6.6 ,p=0.0104; X2=9.9,p=0.0017). Race was not significant (X2=2.5,p=0.28). There was a hazard increase of 0.8% per mm of tumor size (X2=261.7,p<0.0001)(n=15,771). A mixed model algorithm was 98% accurate at predicting survival.
Conclusions: The analyses identified and quantified the relative importance of clinical and demographic parameters on survival for glioma patients. The novel predictive algorithm may be used to assist patient counseling and management decision making.
Patient Care: By being able to provide more information on survival for those with glioma
Learning Objectives: By the conclusion of this session, participants should be able to: 1) Identify the significant prognostic factors for survival in patients with glioma, and 2) Recognize the relative contributions of the prognostic factors to patient survival