Underestimation of alcohol consumption in cohort studies and implications for alcohol's contribution to the global burden of disease

Background and aims Estimated alcohol consumption from national self-report surveys is often only 30-40% of official estimates based on sales or taxation data. Global burden of disease (GBD) estimates for alcohol adjust survey estimates up to 80% of total per capita consumption. This assumes that cohort studies needed to estimate relative risks for disease suffer less from under-reporting than typical national surveys. However, there is limited evidence on which to base that assumption. This paper aims to assess the extent of underestimation of alcohol consumption in cohort studies concerning alcohol and mortality compared with official total consumption estimates. Design Comparisons of estimated per capita consumption from a comprehensive sample of cohort studies against official estimates by country and year. Participants A total of 1 876 046 participants in 40 cohort studies from 18 countries on alcohol use and all-cause mortality identified by systematic review. Measurements Alcohol consumption data from the cohort studies were converted into usual grams of ethanol per day and then to total age 15+ per capita consumption. Matched estimates were sourced from the World Health Organization (WHO) Global Health Observatory. Findings The cohort studies had mean coverages of age 15+ per capita alcohol consumption of 61.71% (ranging from 29.19% for Russia to 96.53% for Japan), after weighting estimates by sample size for within-country estimates and by number of studies per country for the overall estimate. Regional estimates were higher for the United States (66.22%) and lower for western European countries (55.35%). Conclusions Underestimation of alcohol consumption in cohort studies is less than in typical population surveys. Because some under-coverage is caused by under-sampling heavier drinkers, the current practice of uplifting survey estimates to 80% of total population consumption in global burden of disease studies appears to be appropriate.