An inverse association between estrogen therapy (ET) and Alzheimer disease (AD) has been reported in some, but not in all studies. We investigated the association between ET and AD in postmenopausal women using a population-based case-control design. Women who developed AD from 1985 through 1989 in Rochester, MN (cases, n=264) were individually matched by age (±1 y) to control women free of dementia from the same population (controls, n=264). ET exposure (≥6 mo after menopause) was ascertained by abstracting the complete medical records archived in the records-linkage system of the Rochester Epidemiology Project. The frequency of ET use was similar in cases (11.4%) and controls [10.6%; odds ratio=1.10; 95% confidence interval (CI)=0.63-1.93]. However, cases who used ET had a suggestive trend for an earlier age at start of ET compared with controls (median, 49.0 vs. 50.5 y; P=0.06). Although smoking (ever vs. never) was not associated with AD overall, we observed an interaction between smoking and ET. The odds ratio of AD in ET users was 4.55 (95% CI=1.33-15.53) among smokers, but was 0.68 (95% CI=0.35-1.32) among never-smokers (P for interaction=0.01). Our findings do not confirm a significant association between ET and AD overall; however, the possible interaction with smoking deserves further study.