A multivariate assessment of alcohol consumption

Subjects attending a large, multiphasic health screening centre in Sydney, Australia estimated their alcohol consumption and specimens of their blood were analysed. The most useful univariate estimates of alcohol consumption were erythrocyte mean corpuscular volume and plasma aspartate-aminotransferase, gamma-glutamyl-transpeptidase, triglycerides and uric acid. The most statistically significant of these tests have been combined to form a multivariate predictor of alcohol intake which is more successful in identifying heavy-drinkers than single tests. To describe this population further, and to aid comparisons between populations, information about non-drinkers has also been provided.