Opioids are used increasingly for chronic non-cancer pain. Controversy exists about their effectiveness and safety with long-term use. We analysed available randomised, placebo-controlled trials of WHO step 3 opioids for efficacy and safety in chronic non-cancer pain. The Oxford Pain Relief Database (1950-1994) and Medline, EMBASE and the Cochrane Library were searched until September 2003. Inclusion criteria were randomised comparisons of WHO step 3 opioids with placebo in chronic non-cancer pain. Double-blind studies reporting on pain intensity outcomes using validated pain scales were included. Fifteen randomised placebo-controlled trials were included. Four investigations with 120 patients studied intravenous opioid testing. Eleven studies (1025 patients) compared oral opioids with placebo for four days to eight weeks. Six of the 15 included trials had an open label follow-up of 6-24 months. The mean decrease in pain intensity in most studies was at least 30% with opioids and was comparable in neuropathic and musculoskeletal pain. About 80% of patients experienced at least one adverse event, with constipation (41%), nausea (32%) and somnolence (29%) being most common. Only 44% of 388 patients on open label treatments were still on opioids after therapy for between 7 and 24 months. The short-term efficacy of opioids was good in both neuropathic and musculoskeletal pain conditions. However, only a minority of patients in these studies went on to long-term management with opioids. The small number of selected patients and the short follow-ups do not allow conclusions concerning problems such as tolerance and addiction.