rics, of 12 993 nulliparous women, showing that even use in the latent phase, compared with later, led to no difference. It was really very early compared with later, rather than late. There was a very high rate of intervention in both arms, but the rate of caesarean section was modest (23-24%), although not at our current common rates of up to or exceeding 30%. Wong et al. 6 reported on a second study by the same group, in which nulliparous women were again exposed to combined spinal and epidural analgesia, and not epidural, but it was of induced women only. Hence its external validity to women in spontaneous labour is questionable.This meta-analysis illustrates the old issue of what constitutes the type of study that should be included in a meta-analysis. In my opinion, the only study that speaks to the issue at hand is that of Ohel et al. 4 One can conclude that in this setting, an early epidural will not lead to an increase in the rate of caesarean section. However, most of the environments in North America and most Western countries do not have a caesarean section rate of 11-13%. Thus, the Ohel study illustrates that if we could approach such a low caesarean section rate, epidurals are not likely to be a problem. In fact we could all learn from Ohel and company about how they achieve such a low rate of caesarean section. j However, we do not agree with his comments. Dr Klein suggests that the meta-analysis is primarily about caesarean section, and that the title is therefore misleading. The primary focus of our meta-analysis was the effect of early epidural analgesia (EA) on the rate of instrumental vaginal delivery or caesarean delivery. Table 3 illustrated the inclusion of six studies (15 399 women) for the rate of instrumental delivery and five studies (14 836 women) for the rate of caesarean delivery. Thus, we disagree with Klein's conclusion, and consider the title appropriate. MC KleinThe fact that Luxman's 3 study was underpowered is irrelevant because the results were pooled with results from other studies. In addition, Klein refers repeatedly to the fact that there is a low caesarean section rate in several studies. He does not argue, however, about what would be the effect of a higher, versus lower, rate of caesarean section on the relative risk. Looking at relative risks for different studies included in the meta-analysis, we see no link between the rate of caesarean section and the relative risk. Three studies with a caesarean section rate of less than 20% described relative risks of 0.67, 3 0.86, 4 and 1.16, 5 respectively, and two studies with caesarean rates over 20% reported relative risks of 1.02 6 and 1.05, 7 respectively. After the first request for analgesia, Wong et al.7 randomised women to receive intrathecal fentanyl or systemic hydromorphone. Subsequently, patient-controlled epidural analgesia was initiated in the intrathecal group at the second request, and in the systemic group at a cervical dilatation of at least 4 cm or at the third request for analgesia.In Wang's study, a min...