Many people are surprised to learn how little high quality evidence supports the use of pethidine for pain relief in labour. Of course, pethidine has not been shown to be ineffective, it is just that well-designed placebo-controlled studies have not yet been performed. The reason has been that it is unethical to submit women to a placebo-controlled trial of a drug which everyone thinks they already know is effective. Surely, people argue, even a weak opioid must reduce pain. However, anaesthetists and obstetricians have increasingly questioned this belief, and many have come to genuinely doubt that pethidine is effective at all. Fortunately, a well-designed placebo-controlled randomised trial has now been done by doctors from the Chinese University in Hong Kong (pages 648–655). It was stopped at the first interim analysis after 50 women had been recruited, because they had the answer. Pethidine works, although its analgesic effect is modest. The next step must be to do some decent quality comparative studies of alternative analgesics; the quality of the studies in the current Cochrane review of the topic is, to put it bluntly, rubbish. Such trials will need to be fairly large because we do not just need to know which analgesic is most effective. If there is a trade-off between analgesic effectiveness and fetal safety, doctor and parents need precise estimates of both to make informed decisions. Of all the ‘talking therapies’, cognitive behaviour therapy has the strongest evidence base. It works for treating both depression in general, and postnatal depression in particular. A research group from Perth in Western Australia quite reasonably suggested that it might prevent depression too, and decided to test their hypothesis in a group of high risk mothers, those who had recently delivered a very preterm infant. They were disappointed (pages 641–647). There was no reduction in the rate of major or minor depression over one year of follow up, and such small trends as occurred were in the direction of harm. This is unfortunate as no other preventive strategies have been shown to be effective either, so the best we can do at present appears to be to keep a look out for depression, and treat it early. Uterine artery embolisation has been used as a treatment for fibroids since 1995. We already knew that it reduces uterine size, and symptoms related to pressure and menorrhagia, but the article on pages 700–705 is the first to demonstrate an objective reduction in menstrual blood loss. The VALUE study (Vaginal Abdominal or Laparoscopic Uterine Excision) included data on all hysterectomies performed for benign disease in England, Wales and Northern Ireland between October 1994 and September 1995. On pages 688–694, the authors report the rates of severe complications and relate these to a number of pre-operative risk factors. The results are sobering. Severe complications, defined as death, deep vein thrombosis, pulmonary embolism, myocardial infarction, renal failure, cerebrovascular accident, septicaemia, necrotising fasciitis, secondary haemorrhage, ureteric obstruction or visceral damage occurred in 4.4%, and the overall mortality was 0.38 per 1000. There is much else of interest in this paper. For example, the severe complication rate appears to fall, not rise, with increasing patient age but not to be related to grade of operator. In this country, prophylactic antibiotics are widely used for hysterectomy; they were administered to 72% of patients in the VALUE study. However, practice in other countries differs. On pages 695–699, we report a randomised trial from Canada of povidone iodine gel or nothing, used as a supplement to the usual vaginal vault cleaning prior to abdominal hysterectomy. Less than 10% of participants in either group received prophylactic antibiotics. The result of adding in povidone iodine gel was negative but interesting. There was no difference in the primary outcome, postoperative infectious morbidity or in any of the secondary outcomes, apart from one. The exception was pelvic abscess, which occurred in seven control participants but none of those who received the treatment. If genuine, such a difference would clearly be important. Perhaps the trial should be repeated to test this hypothesis.