Reviewer #4: The authors are still failing to take the problem of heterogeneity seriously. The latest response was "well, yes, the results are all over the place, but most results are pretty good". First off, it is not at all clear to this reviewer that a negative predictive value of 85%, the threshold used by the authors, is acceptable. Would most patients be happy to be told that they only have a 15% risk of high-grade cancer, so not to worry about it? That is approximately the risk for a PSA of 10. How many urologists would argue against giving a biopsy to a man with a PSA of 10?Second, the point about heterogeneity is absolutely not to find out where the majority of results lie. For instance, if this was a meta-analysis of complication risk after abortion, and there was gross heterogeneity, with most clinics having very low rates and a few having very high rates, I'm pretty sure the conclusion wouldn't be only that "abortion is a safe procedure". More likely, there would be something about there being some unsafe clinics, and then an analysis to determine the characteristics of those clinics (e.g. solo practice) or a call for further research to determine why some clinics have poor outcomes. The current paper reads as if the authors came up with the conclusion first and then went through the motions of reporting forest plots and I2 statistics and so on. The conclusion that "Multiparametric MRI of the prostate is an accurate test for ruling-out clinically significant prostate cancer" is simply not an appropriate reflection of the data presented by the authors.Formatted: Font: Bold Commented [AL1]: I know we'd already done it but we don't need to point that out to AV. Hopefully he'll think we've now spotted the wisdom of his words and made the change requested. Don't want to give him any reason to dig his heals in.
