In 1992 Chalmers called for randomised trials to assess the potential for serious maternal side effects caused by epidural analgesia1. The response to his letter by Reynolds et al.2 suggested that ‘in most large obstetric centres, we have passed the point where randomisation of epidural analgesia is ethical or practical’. Fortunately this was and is not so. The results of several randomised trials of epidural versus systemic opioid analgesia completed since 1992 have been reviewed in a meta-analysis3 and in the Cochrane database4. These randomised trials have certainly not been easy. C. J. Howell et al.5 reporting in this issue of the Journal (see pages 27-33), enrolled the women in their trial over four years: 2840 women were approached in the antenatal clinic and 875 of these provisionally agreed to participate. In the end 369 women were randomised, with a drop-out rate after this of approximately one-third in each group. In one other trial, half the women due to receive intravenous opioid requested different pain relief4.