BACKGROUND: Using alcohol intake at one point in time, numerous studies have shown a J- or U-shaped relation with all-cause mortality. Mortality is lowest among the light to moderate drinkers, with the risk of dying from coronary heart disease higher among nondrinkers and the risk of dying from cancer higher among heavy drinkers. We studied whether changes in individual alcohol intake result in corresponding changes in mortality. METHODS: In a longitudinal study of 6644 men and 8010 women, age 25 to 98 years, who had attended at least 2 health surveys with a 5-year interval between them, we addressed the risk of death after combinations of changes in alcohol intake. RESULTS: Mortality after changes in alcohol intake was consistent with the mortality observed among those who reported stable drinking. Stable drinkers showed a U-shaped all-cause mortality, with relative risks of 1.29 (95% confidence interval [CI] = 1.13-1.48) for nondrinkers ( 13 drinks per week) compared with light drinkers (1 to 6 drinks per week). For coronary heart disease mortality, stable nondrinkers had a relative risk of 1.32 (0.97-1.79) compared with stable light drinkers and those who had reduced their drinking from light to none increased their risk (1.40; 1.00-1.95), and those who had increased from nondrinking to light drinking reduced their relative risk ratio (0.71; 0.44-1.14). Cancer mortality was increased in all groups of heavy drinkers. CONCLUSION: Persons with stable patterns of light and moderate alcohol intake had the lowest all-cause mortality. Individual changes in alcohol intake were followed by corresponding changes in mortality.