Purpose: In the absence of high-level evidence or clinical guidelines supporting any given active treatment approach over another for localized prostate cancer, clinician and patient preferences may lead to substantial variation in treatment use. Methods: Data were analyzed from 36 clinical sites that contributed data to the Cancer of the Prostate Strategic Urologic Research Endeavor (CaPSURE) registry. Distribution of primary treatment use was measured over time. Prostate cancer risk was assessed using the D'Amico risk groups and the Cancer of the Prostate Risk Assessment (CAPRA) score. Descriptive analyses were performed, and a hierarchical model was constructed that controlled for year of diagnosis, cancer risk variables, and other patient factors to estimate the proportion of variation in primary treatment selection explicable by practice site. Results: Among 11,892 men analyzed, 6.8% elected surveillance, 49.9% prostatectomy, 11.6% external-beam radiation, 13.3% brachytherapy, 4.0% cryoablation, and 14.4% androgen deprivation monotherapy. Prostate cancer risk drives treatment selection, but the data suggest both overtreatment of low-risk disease and undertreatment of high-risk disease. The former trend appears to be improving over time, while the latter is worsening. Treatment varies with age, comorbidity, and socioeconomic status. However, treatment patterns vary markedly across clinical sites, and this variation is not explained by case-mix variability or known patient factors. Practice site explains a proportion of this variation ranging from 13% for androgen deprivation monotherapy to 74% for cryoablation. Conclusion: Substantial variation exists in management of localized prostate cancer that is not explained by measurable factors. A critical need exists for high-quality comparative effectiveness research in localized prostate cancer to help guide treatment decision making. © 2010 by American Society of Clinical Oncology.