Commentary on Cobiac et al. (2009): How to use science to improve alcohol policy?Cobiac and colleagues [1] have convincingly demonstrated that science can help the decision making of alcohol policy in the eve of a global alcohol strategy [2]. They combined sophisticated modelling in the tradition of WHO CHOICE methodology (CHOsing Interventions which are Cost Effective: http://www.who.int/choice/en/; for applications to reduce alcohol-attributable harm see [3]) with local expertise (such as derived from the Advisory Board) to compare realistic interventions in Australia for reducing alcohol-attributable health harm. This is a remarkable approach, which constitutes a marked improvement over the more intuitive expert weightings of different policy options, without explicitly modelling the different options given current knowledge about both interventions and population composition and predicted growth (e.g. [4]). While the overall result is important and hopefully will impact alcohol policy, I still see room for improvement.The most crucial aspect of these calculations is the estimation of exposure in the different scenarios, which the authors derived from a nationally representative survey. Yet, what does this mean? First, such surveys notably underestimate per capita consumption as derived from sales and/or production figures. In Australia [5] the level of underestimation in recent surveys was typically by more than 40% [6]. Such underestimates are common in high-income countries [7]. In addition, in case of the underlying survey, the participation rate was less than 50% with an estimated response rate of less than 40%. Moreover, heavy drinking populations such as the homeless or institutionalized were excluded by sampling design for the underlying survey, as well as for most others. As few drinkers account for substantial amounts of overall alcohol consumed (concentration of consumption; see e.g. [4]), the underestimation of overall exposure in such surveys is not surprising. For the conclusions of Cobiac and colleagues [1] this has two consequences: first, the cost-effectiveness ratios for interventions are all underestimated: alcohol interventions such as those examined are even more cost effective compared to other interventions in the health care field. Consequently, there is an even stronger argument for alcohol policy in the current climate of rationing health care! For future publications reporting on surveys and survey-based research, particularly those claiming national representativeness, reporting the coverage rate (i.e. the proportion of per capita consumption covered by the survey) should be made standard in any good journal, as only knowledge of this coverage rate allows comparison with other results [7]. Per capita consumption figures for these comparisons for all countries are available calculated in a standardized way from the World Health Organization (Global Information System on Alcohol and Health: http://www.who.int/ substance_abuse/activities/gad/en/; see also [8], for background).Other as...