Longitudinal trajectories of heavy drinking in adults in the United States of America.

AIMS: To estimate age, period, cohort and other demographic influences on heavy alcohol consumption and trajectories of heavy drinking in American adults.

DESIGN: Prospective cohort of 14 127 participants, aged 25-74 years at baseline. Generalized estimating equations to model longitudinal change in the probability of heavy drinking and its association with demographic factors. Setting National, population-based sample of non-institutionalized civilians.

MEASUREMENTS: Heavy alcohol consumption (usual number of drinks per occasion >or= five for men; >or= four for women) at baseline (1971-74) and three follow-ups until 1992.

FINDINGS: Heavy alcohol consumption declined with increasing age (age effect) and tracked national average consumption (period effect). There was no cohort effect. Higher probability of heavy drinking was associated with male gender (relative risk: RR = 2.4), being not married (RR = 1.4), having less than high school education (RR = 1.7), having annual income below the median (RR = 1.5), not living in the South-east (RR = 1.7), and smoking (RR = 3.4). Getting married and quitting smoking during the study were each associated with reduction in heavy drinking (RR = 0.55 and 0.61, respectively). Slower age-related decline in the probability of heavy drinking was seen in men (P < 0.0001), married individuals (P = 0.03), and smokers (P = 0.05).

CONCLUSIONS: Demographic predictors of trends in heavy drinking are different from those for trends in average alcohol consumption. The likelihood of heavy drinking declined more slowly with increasing age in men and smokers, suggesting that the negative health effects of alcohol in older ages may be greatest in these groups.