Subthalamic nucleus (STN) deep brain stimulation (DBS) is currently the most common therapeutic surgical procedure for patients with Parkinson's disease (PD) who have failed medical management. However, a recent summary of clinical evidence on the effectiveness of STN DBS is lacking. We report the results of such a systematic review and meta-analysis. A comprehensive review of the literature using Medline and Ovid databases from 1993 until 2004 was conducted. Estimates of change in absolute Unified Parkinson's Disease Rating Scale (UPDRS) scores after surgery were generated using random-effects models. Sources of heterogeneity were explored with meta-regression models, and the possibility of publication bias was evaluated. Patient demographics, reduction in medication requirements, change in dyskinesia, daily offs, quality of life, and a ratio of postoperative improvement from stimulation compared to preoperative improvement by medication from each study were tabulated and average scores were calculated. Adverse effects from each study were summarized. Thirty-seven cohorts were included in the review. Twenty-two studies with estimates of standard errors were included in the meta-analysis. The estimated decreases in absolute UPDRS II (activities of daily living) and III (motor) scores after surgery in the stimulation ON/medication off state compared to preoperative medication off state were 13.35 (95% CI: 10.85-15.85; 50%) and 27.55 (95% CI: 24.23-30.87; 52%), respectively. Average reduction in L-dopa equivalents following surgery was 55.9% (95% CI: 50%-61.8%). Average reduction in dyskinesia following surgery was 69.1% (95% CI: 62.0%-76.2%). Average reduction in daily off periods was 68.2% (95% CI: 57.6%-78.9%). Average improvement in quality of life using PDQ-39 was 34.5% +/- 15.3%. Univariable regression showed improvements in UPDRS III scores were significantly greater in studies with higher baseline UPDRS III off scores, increasing disease duration prior to surgery, earlier year of publication, and higher baseline L-dopa responsiveness. Average baseline UPDRS III off scores were significantly lower (i.e., suggesting milder disease) in later than in earlier studies. In multivariable regression, L-dopa responsiveness, higher baseline motor scores, and disease duration were independent predictors of greater change in motor score. No evidence of publication bias in the available literature was found. The most common serious adverse event related to surgery was intracranial hemorrhage in 3.9% of patients. Psychiatric sequelae were common. Synthesis of the available literature indicates that STN DBS improves motor activity and activities of daily living in advanced PD. Differences between available studies likely reflect differences in patient populations and follow-up periods. These data provide an estimate of the magnitude of the treatment effects and emphasize the need for controlled and randomized studies.