Pignone et al.

Comments for EDITORS

This is a meta-analysis of primary prevention trials of CHD using lipid-lowering drugs. Despite a fairly exhaustive sweep the authors have finished up with only four trials satisfying their criteria, and involving three different drugs. Only two trials from the 1980s and two from the 1990s satisfied their criteria. Although this was a meta-analysis, for the three endpoints one trial for each provided 40%, 47% and 37% of the events. Spreading their net a little wider means including some fairly elderly studies such as the Atromid S (clofibrate) trial of the 1970s.

The problem with meta-analyses is that anyone can do them as the data are published. The authors contradict the results of a previous meta-analysis of 1997 by the addition of one trial. Whereas a 1997 meta-analysis had suggested that statins in primary prevention reduced all causes mortality (ie prolonged life) this was strongly influenced by the WOSCOPS study in which 73 of 135 deaths in the control group (a very high proportion) were from coronary disease.

The addition of the TexCAPS study in which 25 of 145 deaths in the control group (a very low proportion - I would expect coronary disease to account for about 25-30% of deaths in a coronary-prone age group) has altered this conclusion.

Is their study sufficiently good and well presented to outclass what other people might have done with the same data? I found it, for the most part, clearly presented and argued and succinct. I do not have any substantial criticisms although not being a clinical triallist I am aware that some of those who are argue the finer points like mediaeval theologians or metaphysicians and might claim that this analysis is not showing anything new to them.

Statement of respectability

I have no personal or vested or financial interest in publication/ non-publication of this paper.

Hugh Tunstall-Pedoe
 

Comments for the AUTHORS

I found the argument and the presentation clear and concise and have no major differences with the authors.

Abstract: I would have preferred the inclusion criteria stated before the search strategy which initially seems to imply that only studies published from 1994 were involved.

Introduction: This is an evolving situation where dates of publications are critical and a few publication dates would help the argument.

Page 7. I would not assume that everybody knows the Peto and DerSimonian and Laird methods so that they do not need referencing.

Overall discussion. No major problems except the point that WOSCOPS had 73 coronary deaths out of 135 in the control group which is more than 50% and therefore very high, whereas TexCAPS had 25 out of 145 which is amazingly low. In the other two trials the ratio is also around 50%.

Figure headings. I am not sure why total comes into every title! I personally prefer All cause mortality to total but anyway what is "Total CHD mortality"? Otherwise I found them a very clear summary of the trials.