Evaluating glioma therapies: modeling treatments and predicting outcomes.
BACKGROUND: Intra-arterial chemotherapy with carmustine (BCNU) and interstitial radiation therapy with the use of stereotactically placed 125I sources are aggressive local therapies for malignant glioma. These therapies emerged in the 1980s and both appeared promising in phase II studies but yielded disappointing results in subsequent randomized controlled trials by the Brain Tumor Cooperative Group (BTCG). Florell and colleagues had prepared us for the possibility that brachytherapy would have less impact on survival than anticipated from the phase II experience by demonstrating that patients who were judged eligible for interstitial radiation, but treated conventionally, lived significantly longer than those who were ineligible and had better than average outcomes. PURPOSE: To further examine the impact of patient selection on outcome, we used the database of Florell et al. to assess the survival of patients with malignant glioma who were eligible or ineligible for chemotherapy by three intra-arterial methods, one of which was similar to that employed by the BTCG in its randomized, controlled trial evaluating intra-arterial BCNU. METHODS: The medical records and computed tomography (CT) scans of 102 consecutive patients with malignant glioma receiving standard treatment (i.e., maximum feasible surgical resection, external-beam radiotherapy, and often adjuvant systemic chemotherapy) at a single cancer center in Canada during the calendar years 1988 and 1989 were used for this analysis. Based on CT imaging and blind to outcome, an interventional neuroradiologist decided which patients were eligible or ineligible for intra-arterial chemotherapy via injection of two major arteries, via injection of one major artery, or via selective middle-cerebral artery injection. A Karnofsky performance score of greater than or equal to 60 was required. The percent of eligible patients, the median survival time, and the distribution of prognostic factors were analyzed for each group of eligible and ineligible patients. Median survival times were compared with the use of the generalized Wilcoxon (Breslow) test. All P values were based on two-tailed tests. RESULTS: For two-vessel treatment, 72.5% of the patients (74 of 102) were eligible; the eligible patients on average lived longer than the ineligible patients (14.8 versus 3.5 months; P < .00001). For one-vessel treatment, 48% of the patients (49 of 102) were eligible; again, the eligible patients lived longer than the ineligible patients (18.4 versus 5.1 months; P < .00001). For middle-cerebral artery treatment, 30% of the patients (31 of 102) were eligible, and these eligible patients did live somewhat longer than the ineligible patients, but this result did not reach statistical significance (13.6 versus 9.9 months; P = .1304). Trends were similar for patients with glioblastoma multiforme and anaplastic glioma. The median duration of survival was 11.4 months for all patients. CONCLUSIONS: Patients who were eligible for intra-arterial chemotherapy lived significantly longer or somewhat longer (depending on the selection criteria used) than patients who were ineligible and had better than expected outcomes. Patients who were judged eligible for intra-arterial chemotherapy by the two-vessel method and the control group in the BTCG phase III trial of intra-arterial chemotherapy had similar median survival times (14.8 versus 14.0 months). IMPLICATIONS: Modeling treatments with the use of a comprehensive clinical and imaging database of unselected, conventionally treated patients may help investigators decide if new therapies warrant definitive evaluation in randomized trials by measuring the degree to which patient selection may have enhanced phase II study outcomes.