Ascertainment bias may explain some of the paradoxical genetic correlations associated with alcohol consumption (61). Population based cohorts, such as UKB and 23andMe, are based on voluntary participation and tend to attract individuals with higher education levels and socioeconomic status than the general population and, crucially, lower levels of problem drinking. In contrast, ascertainment in the PGC and MVP cohorts (12, 18) was based on DSMIV AD diagnosis and ICD (International Classification of Diseases) codes for AUD, respectively. Collider bias (the biased estimation of the causal effect of an exposure on an outcome) has been proposed to underlie some of the genetic correlations between alcohol consumption and BMI (62); however, BMI has been consistently negatively correlated with alcohol use in several subsequent studies (18, 21, 22, 63). Furthermore, it is also possible that the genetic overlap between AD and aspects of alcohol consumption are dependent on the specific patterns of drinking. For example, Polimanti et al (64) identified a positive genetic correlation between AD and alcohol drinking quantity (rg = 0.75), but not frequency.
