It is difficult to find fault with the analysis by Irwig and colleagues [1]. The authors' criteria and recommendations appear logical, attesting to their thoughtfulness and ability to effectively communicate important and often subtle issues. Their discussion and examples of the distinction between diagnostic accuracy and shifts in classification thresholds are particularly illuminating.

The article provoked my curiosity and made me wish that the authors had provided some additional information. I could not help but wonder about the extent to which judgments by two of the three independent raters were needed to reach "consensus" on the application of each of the guidelines in the 11 meta-analyses. The abstract's reference to these occasions as "consensus" made me uncomfortable, and the authors' use of the phrase "majority view" in the text seemed more appropriate. This raises a question: What actually constitutes "consensus"? A formal test of inter-rater agreement, such as Cohen's statistic [2], might provide insight into the difficulty of this judgment.

A summary Table of the 11 meta-analyses classified by inclusion of the various criteria also might prove useful in answering various questions about the degree of quality within individual meta-analyses, for example, whether the studies that examined the interdependence of sensitivity and specificity in estimations of diagnostic accuracy were the same studies that used an appropriate reference standard or avoided verification bias. Such a summary Table would provide much the same utility offered by tables of primary studies in meta-analyses and would help readers more clearly identify better exemplars of meta-analysis. These could be models for those wishing to learn more about or to conduct meta-analyses. As in the best of life, the authors' work begs us to ask for more.