As Hazem El-Refaey and his colleagues point out (pages 1104–1110) postpartum haemorrhage is the main cause of maternal death in the developing world. Standard oxytocic regimens are less suitable than in Western countries, for the drugs are unstable in hot climates and have to be administered by syringe and needle involving expense and the risk of transmission of infection. These considerations have stimulated research into misoprostol to prevent postpartum haemorrhage, misoprostol being stable and administered by mouth.

Two randomised trials in this issue compared oral misoprostol with standard oxytocic drugs in the prevention of postpartum haemorrhage. Both trials were conducted rigorously, and one even used placebos, but they both illustrate the problems of conducting randomised trials in this area. Robert Walley and his colleagues (pages 1111–1115) used a continuous variable, decrease in haemoglobin concentration, as the primary outcome of their trial, but this is not so important as the risk of postpartum haemorrhage, a dichotomous variable, requiring a much larger randomised trial. The definition of postpartum haemorrhage is uncertain, for blood loss of 500 ml or more is unlikely to be clinically important in Western countries but may be critical in those women in the developing world who are severely anaemic in late pregnancy. In all settings postpartum haemorrhage of greater than one litre is life-threatening. In these randomised trials of routine clinical practice accurate measurement of blood loss was impossible and was estimated subjectively by the midwife; in an open randomised trial this subjectivity may have introduced bias. The perceptions of postpartum haemorrhage differed between the trials, for the proportion of women with blood loss greater than 500 ml was 12% in London and 0.5% in Accra. El-Refaey and colleagues tested the hypothesis that misoprostol would not double the risk of postpartum haemorrhage compared with standard oxytocic drugs, requiring 1000 women to be randomised, but we need to show that misoprostol is equivalent to standard oxytocic drugs, requiring 27,000 women to be randomised, as the authors acknowledge.

Nevertheless these randomised trials are important, for their results may be combined with those of other small trials to provide conclusive evidence of the efficacy and safety of misoprostol in the prevention of postpartum haemorrhage. Alternatively a trial should be conducted which is sufficiently large to demonstrate the equivalence of misoprostol with standard oxytocic drugs. We eagerly await the results of the WHO multicentre trial.