The purpose of the study was to provide updated estimates of alcohol-caused mortality rates in Australia between 1990 and 1997, making adjustments for changes in the prevalence of high-risk alcohol use estimated on the basis of per capita alcohol consumption (PCAC). Deaths wholly and partially attributable to high-risk alcohol consumption were extracted from the Australian Bureau of Statistics Mortality Datafile (1990-1997) and multiplied by specific aetiologicalfractions, which in turn were adjusted by changes in the prevalence of high-risk alcohol use estimated on the basis of annual changes in PCAC. The yearly trends in age-standardized rates of estimated alcohol-caused deaths were compared with those using (i) aetiological fractions unadjusted for changes in PCAC, and (ii) wholly alcohol-caused conditions only (thus requiring no application of aetiological fractions). The age-standardized rates of all alcohol-caused deaths among males aged 15 + years declined from 1990 (4.01110000) to 1993 (3.19/10000) and decreased far more slowly up to 1997 (3.15/10000)-16% overall. For females, these rates declined steadily from 1990 (1.75/10000) to 1997 (1.33/10000)-19% overall. Similar patterns in time trends were noted for estimated alcohol-caused death rates calculated as in (i) and (ii). However, the proportional decreases in rates (21.6%for males; 24. O%for females) would have been underestimated by 16% (males) and 19% (females) if the alcohol aetiological fractions had not been adjusted to take account of the estimated annual changes in the prevalence of high-risk drinking. The declines in estimated alcohol-caused death rates were more pronounced than the 9% decline in PCAC, and were due mainly to decreasing death rates for stroke (men and women), alcoholic liver cirrhosis and road injuries (men only). When aetiological fractions are used to measure temporal trends in estimated alcohol-caused death rates from official mortality statistics, they should account for annual changes in the prevalence of high-risk drinking. Such changes in prevalence can be deduced from yearly fluctuations in PCAC.