BACKGROUND: In view of widespread claims of efficacy, we examined the evidence regarding the effects of hypnosis for pain relief during childbirth. METHODS: Medline, Embase, Pubmed, and the Cochrane library 2004.1 were searched for clinical trials where hypnosis during pregnancy and childbirth was compared with a non-hypnosis intervention, no treatment or placebo. Reference lists from retrieved papers and hypnotherapy texts were also examined. There were no language restrictions. Our primary outcome measures were labour analgesia requirements (no analgesia, opiate, or epidural use), and pain scores in labour. Suitable comparative studies were included for further assessment according to predefined criteria. Meta-analyses were performed of the included randomized controlled trials (RCTs), assessed as being of "good" or "adequate" quality by a predefined score. RESULTS: Five RCTs and 14 non-randomized comparisons (NRCs) studying 8395 women were identified where hypnosis was used for labour analgesia. Four RCTs including 224 patients examined the primary outcomes of interest. One RCT rated poor on quality assessment. Meta-analyses of the three remaining RCTs showed that, compared with controls, fewer parturients having hypnosis required analgesia, relative risk=0.51 (95% confidence interval 0.28, 0.95). Of the two included NRCs, one showed that women using hypnosis rated their labour pain less severe than controls (P<0.01). The other showed that hypnosis reduced opioid (meperidine) requirements (P<0.001), and increased the incidence of not requiring pharmacological analgesia in labour (P<0.001). CONCLUSION: The risk/benefit profile of hypnosis demonstrates a need for well-designed trials to confirm the effects of hypnosis in childbirth.