The Cochrane Pregnancy and Childbirth Database
BMJ 1994; 308 doi: https://doi.org/10.1136/bmj.308.6941.1448 (Published 28 May 1994) Cite this as: BMJ 1994;308:1448
- R J Lilford
The Cochrane Collaboration Annual subscription (pounds sterling )99 (institutional) (pounds sterling )57 (individual) Available on 3.5″ or 5.25″ disk from Update software, Oxford OX44 7QB
It is a sign of the times to be asked to review not a book but a computer database. This is welcome; books are often out of date even by the time they are published. The Cochrane Childbirth and Pregnancy Database (Previously the oxford Database) is a product of secondary research - research that is designed to find and analyse original research publications. that some secondary research in necessary should come as no surprise, given that most individual articles get lost in the mass of medical information. The authors of the database have made an immense contribution by pulling all of this together, and it is extremely useful both to clinicians and to researchers, as well as being very easy to use.
The information contained in structured reviews can provide a clear guide to clinical action. while some treatments - such as administering anti-D immunoglobulin to an Rh negative woman - have obvious advantages, the effects of other interventions - such as the use of antibiotics at the time of caesarean section - cannot be inferred from simple clinical observation or deduced from first principles. Individual studies may give conflicting results because of chance or inadequate power, and the only way to make sensible use of the world's research effort is to ascertain all studies, evaluate them for quality, and combine them in a structured way.
The audit committee of the Royal College of Obstetricians and Gynaecologists has therefore committed itself to conduct, on a three yearly basis, a careful review of all the evidence in the Cochrane database of trials. This was last done a year ago, and 22 clear and auditable clinical recommendations emerged. this might seem a small harvest from such an extensive source - the database has 200 reviews and includes data from nearly 10 000 trials. On the other hand, without it we simply would not have the answers to many classic questions in maternity care. For example, we now know that routine induction of labour a week after term is associated with a clear and large reduction in the relative risk of stillbirth. We also know that antenatal measures to try to every the nipples are of no value whatever in promoting breast feeding. These are clear cut answers to longstanding controversies, and they come form structured reviews updated in the database over the past tow years - hence the value of computer rather than paper publication. The database is updated every three months.
How the data base helps researchers
Clinical researchers find the database essential. Knowledge of previous publication is a prerequisite for any research proposal. in labour ward management will find that the effects of routine rupture of the membranes are well documented but that the jury is still out as far as the effectiveness of oxytocin to “augment” slow labour is concerned. These are particularly good examples, because the relevant systematic reviews include data from unpublished trials. Experienced researchers include evidence from such previous studies in their calculations of sample size - for example, if meta -analysis of previous high quality studies shows a non significant worsening of outcome then it possible to calculate the number of subject required to show an improved outcome across all studies combined. The value of the database for clinical scientists cannot be overemphasised.
Relative risks
The current way of presenting results clinical trials has two large drawbacks. The first is that these data are typically given a odds ratios rather than relative risk. While these are almost identical when risks are low (a relative risk of 1 in 100 is equivalent to odds of 1 to 99), they are very different who the risks are large ( a risk of 1 in 6 is equivalent to odds of 1 to 5). Since relative risk is far more intuitive to clinicians I prefer the method of presenting results, and I am pleased that the database includes this option.
Even relative risk leaves out a very important factor, the absolute risk. For example giving prophylactic antibiotics halves the relative risk of maternal infection after caesarean section irrespective of whether the operation was done before or during labour However, the absolute size of this reduction is much greater in intrapartum operation because the baseline risk of infection is much higher. The database software deals with the by giving differences in event rates.
In addition to giving the characteristic 95 confidence intervals (of odds ratios, relative risks, or differences in event rates) the data base will also give 99% confidence intervals would like to have 80% confidence interval as well, because some unbiased evidence better than none at all, and it is not necessary to be 97.5% confident that one treatment better than another in order to select it - is merely necessary to perceive a greater like hood of benefit.
The statistical technique to combine results of trials by meta-analysis in the (a all other current databases) is the Man Haenszel statistic. This seems an inappropriate statistical method because it assume that all trials come from a hypothetical and infinitely large body of trials - that there one underlying true effect. Clearly this is so (clinical outcomes differ according populations and the characteristics clinicians, in different places and over to and I would therefore prefer statistical teniques that do not rely on this underlying assumption. Nevertheless, this is a critical of current worldwide practice rather than this specific database.