Reply to Head et al.

We thank Dr Kappetein’s group for their comments [1] regarding our manuscript [2] and we appreciate the opportunity to provide clarification. The first comment raises the concern that our article does not adhere to the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines [3]. On a thorough review of the 27-point checklist, we have to admit that we have not provided the time frame in which the search for relevant articles was completed. All articles that appear in the MEDLINE database up to 12/1/2009 were searched for inclusion. We also did not provide registration information for our meta-analysis, as the registration process remains optional. Nevertheless, our study addresses all other relevant items from this extensive checklist to various degrees. The point of including other sources to identify relevant articles remains valid. However, the PRISMA guidelines do not require inclusion of other database sources such as Cochrane and Embase, as the authors of the letter seem to imply. In fact, as stated in the PRISMA document, the MEDLINE database remains one of the most comprehensive sources of healthcare information [3]. Dr Kappetein et al. note that in Fig. 2 of the manuscript, there is an outlying odds ratio (OR) of 17.241 from the Al-Radi et al. article [4]. They question how this was calculated when the reported hospital mortality was 1.5% for repair and 21% for replacement, and note that this would yield an OR of 21/1.5 = 14. Actually, what the authors of the comment have calculated is not an OR but a relative risk. The OR calculation in our manuscript was correct: 1/65 patients in the repair group and 29/137 in the replacement group died, yielding an OR of 17.18 (64 × 29/1 × 108). The slight discrepancy from the reported OR of 17.241 stems from the fact that we initially calculated the OR in the opposite direction to reflect survival for replacement compared to repair which was 0.058. Subsequently, we took the reciprocal that was rounded off for ease of presentation, since repair did show superiority compared to replacement. Our method for evaluating log-term survival has been well described and we refer the reader to the well-written article by Parmar et al. for greater insight into understanding the calculation and application of this method for obtaining summary statistics from time-to-event studies [5]. We examined the survival curves from each of the included studies and calculated log hazard ratios (and variances) using non-overlapping 6-month time intervals. This approach allows the investigator to calculate an overall log hazard ratio for each study. All of the included studies provided follow-up of at least 5 years with the exception of the study by Hickey et al. [6]. We are in agreement and have explicitly stated in our article that a meta-analysis of retrospective studies has inherent limitations which need to be considered. Unfortunately, upon searching the literature, we only found studies of retrospective nature. Short of a prospective randomized trial, patient characteristics, surgeon preference and technical ability will continue to play a role in mitral procedure selection. In trying to provide an evidence-based meta-analysis of clinically relevant studies, we selected for inclusion only those articles that would not obviously bias treatment strategy. For example, we excluded patients with hemodynamic instability, who would be more likely to undergo replacement. Therefore, our study provides a useful source of additional information to guide decision making in patients with ischemic mitral regurgitation.