This meta-analysis by Berghella and colleagues adds to the debate about the role of cervical-length measurement in determining candidacy for cerclage in an effort to reduce the rate of preterm birth. The authors are clearly passionate about the prevention of preterm birth—as we all are—but the conclusions they reach must be questioned.
First, it is misleading to report these results as Level-1 evidence. A meta-analysis can never, strictly speaking, be Level-1 evidence, although it may be based on an analysis of Level-1 evidence.
Sound confusing? Let’s take, as an example, the study of the role of calcium supplementation in the prevention of preeclampsia. In JAMA, in 1996, a meta-analysis fairly conclusively demonstrated that calcium supplementation was effective in preventing preeclampsia (odds ratio, 0.38; 95% CI, 0.22–0.65).1 Yet, a subsequent large randomized trial failed to confirm the findings of this meta-analysis.2
The lesson here? Level-1 evidence consists of appropriately powered, large-scale, randomized clinical trials. To date, we lack such trials with respect to cervical-length measurement and indications for cerclage. In fact, two of the authors of this paper are “on the record” as saying this very thing.
A 2009 paper by Owen and coworkers demonstrated only that cerclage for a cervical length below 25 mm reduced previable birth and perinatal death, but did not prevent births before 35 weeks unless the cervical length was less than 15 mm—and that bit of information came from a secondary analysis of the data.3 In a follow-up study, Owen and coworkers concluded that cervical length did not predict preterm birth before 37, 35, or 28 weeks, whether or not cervical length was viewed as a continuum or was stratified.4 And in an earlier meta-analysis reported by Berghella and associates in 2005, the authors conclude that “…cerclage may reduce preterm birth, and a well-powered trial should be carried out on this group of patients.”5
We should continue to rely on clinical assessment and history to make cerclage decisions, a conclusion reached in a recent randomized, controlled trial.6
In the meantime, those of us who practice maternal-fetal medicine would be wise to stop spending time massaging the data (i.e., meta-analysis and secondary analyses) from trials that have already been performed and start spending time, effort, and money to conduct the well-powered trials that I (and Dr. Berghella and colleagues) believe that we need. This is not to say that cervical-length measurement is without value. We simply don’t yet have the strength of association to accurately determine what that value is—most certainly not in the form of a screening tool for low-risk populations.
—John T. Repke, MD
We want to hear from you! Tell us what you think.