of this bias is also towards underestimations, as many research studies have shown that conditions like undernutrition or lack of sanitation which are more prevalent in low and mid income countries, increase the effect of alcohol [10]. Also, the modelling of ischemic heart disease used in the 2000 CRA is based on a meta-analysis for high income regions(see above), overestimating the beneficial effect, as several of the underlying studies did not differentiate between lifetime abstainers and former drinkers(e.g., [42]). For other regions, the modelling of the effects of alcohol on ischemic heart disease is based on multi-level analyses of aggregate data, and there is no overestimation of the protective effect because of misclassification of abstention in this type of analysis. Together with the lack of inclusion of NCD categories in the underlying GBD where alcohol has a detrimental effect (see above), overall the presented figures should be considered as conservative estimates; i.e. the net detrimental impact of alcohol consumption is underestimated.
