Name: SJW Evans
Position:Prof of Pharmacoepidemiology

This paper has shown some considerable improvement, though the tables are rather extensive and, perhaps inevitably, indigestible.

The argument from the authors in favour of publication is resoanbly well put, though there are hints that the authors are less open- minded than they suggest.

The problems with misclassification of exposure are not dealt with as simply as they state. The lack of significant associations in shorter follow-up may be due to the low statistical power when any effect is unlikely to appear until follow-up is extensive.

The first phrase of the conclusions in the abstract might be deleted. It is likely to be misused; the results equally could be said not to rule out a causal relationship- they simply suggest that any realtionship if it exists is not as high as some authors have suggested.

I believe that there are a number of studies that suggest that smoking of 40 or more cigarettes per day has a much greater increase in CHD than the RR of 1.9 in males and 2.4 in females found in this study. It remains possible that this study has under-estimated effects overall.

I am unceratin what the authors mean in the penultimate paragraph of the paper about "author bias". This should be referenced or amended.

I would prefer to see that more reservations are included about the strength of conclusions that can be drawn from epidemiology of this type. The fact that others, based on equally weak studies, have drawn strong conclusions is not the point. We are concerned with this paper and it must make it clear that with observational data of this type, the conclusions must be temepered with acknowledgement of bias and confounding that limit strong statements.