Authors' reply

Since our study had an observational design, with participants not randomised into groups, we adopted a cautious approach to interpreting findings, and there is the possibility that confounding factors might account for the effect. Questions have now been raised concerning the previously well-accepted belief that moderate alcohol consumption confers protection against ischaemic heart disease, with the possibility that either uncontrolled confounding or unmeasured effect modification in observational studies may account for the purported protective association (Jackson et al, 2005). Therefore, we welcome the suggestion of Dr O'Connell that personality differences may partially account for the difference in outcomes for non-drinkers and moderate drinkers, which increases the plausibility of our findings. Nevertheless, we reiterate the need for a conservative approach when interpreting non-experimental data.