Details in a meta-analysis comparing mitral valve repair to replacement for ischemic regurgitation

Although meta-analyses are considered to be of great value to establish an overall effect on an investigated outcome, there are some basic ‘rules’ to such an analysis. Several statements have guided authors in reporting results from systematic reviews and have been increasingly cited to inform readers of the quality of the review [1]. In the March 2011 issue, however, a meta-analysis of short-term and long-term survival after mitral valve repair or replacement for ischemic regurgitation was published in which several details are missing [2].

First of all, the article did not adhere to the preferred reporting items for systematic reviews and meta-analyses (PRISMA) guidelines [3]. These guidelines may prevent authors from missing important steps in the review and analysis process. They, for example, guide to describe the search strategy in full detail. In the current meta-analysis, this is however incomplete as the authors did not report when the search was performed and therefore the inclusion and exclusion of some published articles cannot be verified. In addition, only the MEDLINE database was searched while it is well recognized that an extended literature search of EMBase and/or the Cochran Library can identify further potentially valid studies that can be included.

The authors also did not mention the length of follow-up of the operated patients. The expression ‘long-term survival’ is a subjective phrase, and a hazard ratio only demonstrates an effect for the defined follow-up time. It is, therefore, important to mention these data.

In the meta-analysis section, we furthermore question the accuracy of the acquired forest plots. Fig. 2 shows an overall odds ratio (OR) of short-term mortality in the replacement group versus the repair group. There is an outlying OR of 17.241 (95% confidence intervals 2.330–127.575) in the study from Al-Radi et al. [4]. We were, however, unable to find a matching OR in this article. The reported hospital mortality was 1.5% and 21% in the repair and replacement group, respectively, which yields an OR of 21/1.5 = 14. After propensity adjustment, an OR of 8.3 was reported.

In addition, the difficulty in comparing repair versus replacement is originated by surgeon preferences, the complexity of mitral valve injury, and patient characteristics. Patients who underwent repair therefore represent a different cohort than patients with a replaced valve. Even in propensity-matched studies in which baseline variables are considered equal between groups, valve-related characteristics remain incomparable. Therefore, significant treatment bias cannot be avoided and an overall effect of preferred repair or replacement should not be established based on a meta-analysis of retrospective studies.