The goal of this study was to investigate the use of proton magnetic resonance spectroscopic imaging as a prognostic indicator in gamma knife radiosurgery of recurrent gliomas. Thirty-six patients with recurrent gliomas were studied with proton magnetic resonance spectroscopic imaging at the time of radiosurgery, and with conventional magnetic resonance imaging examinations at regular time intervals until the initiation of a new treatment strategy. Patients were categorized on the basis of their initial spectroscopic results, and their performance was assessed in terms of change in contrast-enhancing volume, time to further treatment, and survival. The trends in the overall population were toward more extensive increase in the percent contrast-enhancing volume, a decreased time to further treatment, and a reduced survival time for patients with more extensive initial metabolic abnormalities. Statistical analysis of the subpopulation of patients with glioblastoma multiforme found a significant increase in relative contrast-enhancing volume (P < 0.01, Wilcoxon signed-rank test), a decrease in time to further treatment (P < 0.01, log-rank test), and a reduction in survival time (P < 0.01, log-rank test) for patients with regions containing tumor-suggestive spectra outside the gamma knife target, compared with patients exhibiting spectral abnormalities restricted to the gamma knife target. Further studies are needed to establish statistical significance for patients with lower-grade lesions and to confirm the results observed in this study. The pretreatment spectroscopic results provided information that was predictive of outcome for this patient pool, both in local control (change in contrast-enhancing volume) and global outcome (time to further treatment and survival). This modality may have an important role in improving the selection, planning, and treatment process for glioma patients.