The aim of this overview was to identify interventions that change doctor prescribing behaviour and to derive conclusions for practice and further research. Relevant studies (indicating prescribing as a behaviour change) were located from a database of studies maintained by the Cochrane Collaboration on Effective Professional Practice. This register is kept up to date by searching the following databases for reports of relevant research: DHSS-DATA; EMBASE; MEDLINE; SIGLE; Resource Database in Continuing Medical Education (1975-1994), along with bibliographies of related topics, hand searching of key journals and personal contact with content area experts. Randomised controlled trials and non-equivalent group designs with pre- and post-intervention measures were included. Outcome measures were those used by the study authors. For each study we determined whether these were positive, negative or inconclusive. Positive studies (+) were those that demonstrated a statistically significant change in the majority of outcomes measured at level of p < or = 0.05 between the intervention and control groups. Negative studies (-) showed a significant change in the opposite direction and inconclusive studies (approximately) showed no significant change compared to control or no overall positive findings. We identified 79 eligible studies which described 96 separate interventions to change prescribing behaviour. Of these interventions, 49 (51%, 41%-61%) showed a positive significant change compared to the control group but interpretation of specific interventions is limited due to wide and overlapping confidence intervals.