To the Editor: The 2005 Cochrane review Home-like versus conventional institutional settings for birth1 has been cited in the public media to claim that birth centres are less safe than labour wards as there was an increased risk of a baby dying during or immediately after childbirth.2

This “headline-grabbing” statement is false. Firstly, this finding from the systematic review did not reach statistical significance.1 Secondly, the outcomes reviewed were related to the allocated place of birth, not the care provided. This fact is critically important, as 48% of women who were booked to have their baby in a birth centre did not give birth there.1 This is a predictable effect of the intention-to-treat principle. However, such high rates of “treatment contamination” negatively affect confidence in the study results.3 Additionally, the vast majority of baby deaths examined in the Cochrane review happened before labour and thus had nothing to do with care during childbirth.

One might wonder whether there was a real increased perinatal mortality rate resulting from delayed transfers from birth centres.1 The analysis found 41 deaths in total, but only six that occurred in normally formed babies who reached term (these are the only babies who are eligible to be born in a birth centre). Three of these deaths were associated with birth centre care, and three with standard labour care.

The interpretation of this Cochrane review raises questions about the validity of the underlying randomised controlled trials. In this experimental design, researcher control should ensure that people receive the specific treatment that was planned for them (treatment fidelity).4 The Cochrane handbook gives no guidance as to how to evaluate either the quality of the researchers’ definition of the planned treatments, or the fidelity between the treatments provided and the researchers’ plan.3 Most of the trials that formed the basis of the Cochrane review did not adequately define their treatments, nor adequately control the treatments provided to either group. It is not clear how the birth centre trials could sensibly be considered to have been scientifically controlled. The reviewers attempted to deal with this critical point by claiming that they were looking only at the effect of the “setting”, but their question clearly states that they were examining the effect of “care within a setting”.1

We conclude that the Cochrane review of the setting for birth is unreliable because of the weaknesses of the underlying trials. Rather than using questionable research to attack birth centres, it would be more constructive to engage in rigorously designed research that could provide robust evidence on the safety of all forms of maternity care, including standard medical care.

In reply: Fahy and Tracy highlight the lack of high-level evidence about the relative safety of different models of maternity care. But in criticising the Cochrane review, it is important not to “shoot the messenger”.

There is no doubt that the Cochrane review is not ideal but, like it or not, it remains the best evidence we have. The review of 8677 women in six randomised trials found a relative risk (RR) of perinatal death of 1.83 (95% CI, 0.99–3.38) in birth centres versus conventional institutional settings. In the 3332 pregnancies assigned to continuity of care by midwives who did not also work in conventional delivery suites, the RR was 2.38 (95% CI, 1.05–5.41).1

It would be fair to say that such findings should lead to real concerns about lack of safety rather than reassure the unbiased observer.

Other published evidence has raised similar concerns. A retrospective review of over 183 000 low-risk births in Stockholm, Sweden, found a statistically significant fourfold increase in intrapartum fetal mortality in women planning birth centre care compared with those planning standard care (three intrapartum deaths in 3256 babies of women planning birth centre care versus 36 deaths in 180 380 babies of those planning standard care).2 The increase in intrapartum mortality was almost sevenfold for primigravidae. These findings led to evidence-based changes in the organisation of the birth centre involved to minimise the identified risks.

To paraphrase Fahy and Tracy, rather than criticising the best available evidence reviewing birth centre outcomes, it would be more constructive to engage in rigorously designed research to assess how risk might be minimised in all forms of maternity care.