Compared to vacuum extraction, the use of forceps resulted in greater success in achieving an instrument-aided vaginal delivery. However, the caesarean section rate was lower with vacuum extraction. This was so because after vacuum extraction failed, an attempt to deliver by forceps was more likely than the use of vacuum extraction after failure to deliver with forceps.
Significant maternal injuries were less likely when vacuum extraction was used. The use of pudendal or general anaesthesia and pain during delivery were also less common in the vacuum group.
The use of vacuum was associated with a higher incidence of cephalhaematoma and retinal haemorrhage, although the latter was evaluated in a small proportion of infants and the result was influenced by the inclusion of a study that was methodologically least sound (see Ehlers 1974 in the review). There was also a tendency towards a higher incidence of low Apgar scores at five minutes with the use of vacuum extraction. It is possible that the higher use of an alternative instrument after extraction failed in the vacuum group, contributed to the worse neonatal outcomes associated with the use of vacuum.
No conclusion can be drawn about the effect on perinatal mortality because of the small numbers. Long-term outcomes were evaluated in only one study (see Portsmouth 1983 in the review) and no differences were observed.
All identifiable, randomized controlled trials have been included.
As the overall reduction in maternal injuries is the main benefit associated with vacuum extraction, a better description of these maternal outcomes would be desirable. The outcome "significant maternal injury" is not defined in the review.
The outcome "Regional and general anaesthesia" is presented in a subgroup analysis. Pudendal, epidural and general anaesthesia, are first analysed independently and a subtotal typical relative risk is calculated for each one. At the end, a total typical relative risk is calculated for the three outcomes together and it is the figure shown in the summary of analysis. This last procedure is not appropriate for these outcomes since it considers the same studies and the same women three times, resulting in a biased analysis and an artificial sample size. Thus, readers do not have to consider the total estimate in this outcome and should only consider the subtotals of the different subgroups.