Observational studies and randomized controlled trials have produced divergent results concerning the effect of hormone replacement therapy (HRT) on cardiovascular disease and, to a lesser extent, dementia. Residual confounding (confounding that remains even after adjustment for various socioeconomic and lifestyle factors) is one explanation that has been offered for these divergent results. The authors used data collected between 1990 and 1995 from 6,697 French women aged 61-72 years participating in a prospective cohort study to explore the hypothesis that nutritional intake varies according to HRT use and thus may be a source of residual confounding. After the authors adjusted for health and lifestyle factors, HRT users, compared with never users, had significantly higher intakes of alcohol; omega3 fatty acids; vitamins B6, B12, and D; and phosphorus and a lower intake of starch. These differential nutrient intakes were related to differences in eating habits. In particular, HRT users in the studied sample, compared with nonusers, ate significantly more fish. Most of the dietary differences were seen in both early users and delayers of HRT. To limit residual confounding in observational studies, dietary factors may be important parameters to be taken into account in analyses of HRT use and health outcomes.