Introduction: Work on identifying research gaps being done by the WOMBAT (Women and Babies Health and Wellbeing: Action through Trials) Collaboration has revealed that the majority of the 500 Cochrane reviews on perinatal topics have inconclusive findings. Users of such reviews often express frustration about the lack of clear findings. Further knowledge about the causes of this uncertainty (either at the review or trial level) may help to reduce it.
Objectives: To ascertain the factors likely to be responsible for inconclusive findings in Cochrane perinatal reviews and to propose actions needed to increase the proportion of reviews with clear findings.
Methods: We developed a classification scheme of 18 factors contributing to uncertainty including items such as sample size, characteristics of the intervention or comparator, participant characteristics, missing outcomes, sources of potential bias, and out of date reviews. We then selected all Cochrane reviews with uncertain findings from 10 perinatal topics (breastfeeding, breathing support, breech birth, feeding preterms, maternal infection, maternal mental health, models of care, pain relief in infants, preterm care and retinopathy of prematurity) and coded each review with one or more contributors to this uncertainty. The results of the coding have been translated into an agenda for addressing particular sources of uncertainty; and detailed proposals for potential solutions will form later phases of this project.
Results: From a total pool of 154 reviews, 100 reviews (65%) had uncertain findings, and the remainder were classified as: findings likely to be effective (26%), likely to be ineffective or harmful (6%) and both benefits and risks (3%). Median number of factors contributing to uncertainty per review was 3 to 4. In 81 of the 100 reviews classified as uncertain, lack of trials or lack of power of trials contributed to the uncertainty (17 had no included trials, 13 had one included trial, and 51 reviews with more one trial were judged to be inconclusive due to insufficient sample sizes in the included trials). Other large contributors to uncertainty were:
o characteristics of the intervention or comparisons - 32.5 out of 100 reviews (interventions and/or comparators in trials did not completely match or align with the all of the questions of the review)
o trials not measuring (or reporting) the right outcomes - 29 out of 100 reviews
o trials not reporting long term outcomes - 27.5 out of 100 reviews
o trials not addressing the right groups of participants - 23 out of 100 reviews (e.g. types of participants too broadly or too narrowly defined).
Standard outcomes were identified as an issue in 12.5 reviews, but these were clustered in particular topics (feeding preterms and pain relief for infants). The lack of currency of reviews may have contributed to uncertainty in 11 of the 100 reviews. In five reviews, ongoing trials were mentioned as likely to be able to answer the review question when completed.
Conclusions: Small sample sizes and lack of trials are severely limiting the ability of perinatal systematic reviews to reveal potentially important differences between treatments. Greater communication between triallists and review authors may help to address this and other sources of uncertainty. Prospective meta-analysis and other forms of cooperation between triallists as well as production and dissemination of detailed research gaps from systematic reviews is likely to positively influence trial design.