The aim of this study was to analyse and compare published data during the last decade on the different approaches to Parkinson's disease surgery. Eighty-eight papers published between 1990 and 2001 fulfilled the inclusion criteria. Full-text and prospective papers on lesion and stimulation of GPi or STN were assessed. Descriptive analysis of surgery procedure and population under study was performed, as well as a meta-analysis of the most consistently reported variables. A total of 1702 patients underwent surgery with a mean age of 58.75 years (range 46.5 - 72.5), mean duration of illness 13.6 years (8.1 - 18.1) and a male:female ratio 1.5:1. Mean postoperative follow-up was 9 months (1 - 52). Single blind assessment was performed in two papers, while double blind evaluation was used in 6. In the GPi group, no difference was found between the pre- and postoperative levodopa equivalent daily dose (960.39 v. 943.13; p > 0.05), while the STN group showed a marked reduction (1104.8 v. 483.04; p < 0.05) of this dosage. Meta-analysis of the most consistently reported variables (UPDRS total score, UPDRS motor score, UPDRS ADL score and Schwab & England score showed that Nucleus, Bilaterality of Approach and Surgical Procedure were the best moderators for defining outcome. Bilateral DBS STN procedures proved to be associated with better outcome. Microelectrode recording was not found to be a moderator that influenced outcome. Although there was a significant improvement of dyskinesias among the different approaches described in the papers, the heterogeneity of data makes it impossible to perform a structured analysis on this item.