One-year drinking history and mortality.

The one-year drinking history of 94 men was recorded by recurrent interviews (mean: 20 per person). The cohort was followed for 18.3 years; during that time, 13 men died. Cox's proportional hazards survival models including age and social class as confounders, indicated that mortality was significantly associated with total annual alcohol consumption, frequency of drinking, and frequency of intoxicating drinking. Estimates of risk of death for various consumption levels are presented: For having 10 drinks (each containing 12 grams of pure ethanol) a week vs one drink a week, the estimated relative risk of death (95% confidence limits in parentheses) was 2.3 (1.6-3.3). For being intoxicated once a week vs no intoxications at all during one year the respective risk was 2.1 (1.3-3.1). The risk estimates for the frequency of intoxication were found to be higher than those in an earlier study using single interview data on drinking. This suggests that more accurate measurement of alcohol consumption may yield higher risk of death estimates than found in studies based on single interview data on alcohol.