The most reliable prognostic factors for patients with primary malignant brain tumors remain histology, age, and functional status. Management of these individuals might be improved by quantifying pertinent molecular markers. We have measured the gene dosage of the epidermal growth factor receptor (EGFR), mouse double minute 2 (MDM2), and cyclin-dependent kinase 4 (CDK4) genes in a series of brain tumor specimens and correlated their amplification status with standard prognostic factors and survival. Individual tumor DNA was successively hybridized with probes for EGFR, MDM2, and CDK4. The signal was quantified by densitometry, and amplification was defined as gene signal > or = 2 times normal. Survival, age, Karnofsky performance status, and histology were correlated with gene amplification. Nineteen astrocytomas, 20 anaplastic astrocytomas, and 70 glioblastomas had complete data available. Median survival with and without any form of gene amplification was 70.7 and 88.6 weeks, respectively (P = 0.0369). For the EGFR gene alone, those with and without amplification had a median survival of 58.9 and 88.6 weeks, respectively (P = 0.0104). By Cox analysis, only tumor histology (P = 0.04) and Karnofsky performance status (P = 0.0157) were significant independent predictors of survival. Gene amplification by itself was not predictive of survival, even for glioblastomas (P = 0.8249). The lack of correlation between gene amplification and survival for patients with primary malignant brain tumors may be because EGFR, MDM2, and CDK4 are only portions of larger signaling systems. Therefore, the lack of a direct correlation between a single gene and outcome is not entirely unexpected.