Prognostic factors for glioblastoma multiforme--development of a prognostic index

The prognosis for glioblastoma multiforme is generally poor. However, it can be useful to identify groups of patients with varying prognosis. Based on data from 495 consecutive patients admitted to the Norwegian Radium Hospital during 1980-1994, multivariable methods were used to identify prognostic factors independently associated with length of survival. A prognostic index was derived in 384 randomly selected patients, and tested in 111 other patients. Performance status, dexamethasone dependency, age group, surgery and behavioural changes were prognostic discriminators. On this basis a prognostic index was made and three separate subsets with different survival prognoses could be identified. The prognostic index was then tested among 20% of the patients and shown to predict reliable survival curves. In the subset with the highest index (high risk subset) almost all patients in the main group as well as in the the test group were dead within 18 months.