BACKGROUND: Genetic variation in the estrogen receptor (ESR) may be associated with the incidence of Alzheimer's disease (AD), but this association could be modified by genetic and environmental factors. METHODS: The association between five ESR α (ESR1) and β (ESR2) polymorphisms with 7-year dementia incidence was examined among 6959 older men and women from the Three City Study using multivariate-adjusted Cox regression models with delayed entry. Gender, the apolipoprotein E (APOE) ε4 allele, and hormone treatment were considered as potential effect modifiers of this association. RESULTS: Among women, the CC genotype of ESR1rs2234693 was specifically associated with a small increased risk of AD (adjusted hazard ratio [HR]: 1.54, 95% confidence interval [CI]: 1.03-2.28, P = .03). However, women with this genotype had a substantially increased risk of AD associated with the APOE ε4 allele (adjusted HR: 3.24, 95% CI: 1.81-5.79 for women rs2234693 CC; compared with HR: 1.87, 95% CI: 1.37-2.56 for all women). There was also evidence of a nominally significant interaction between the ESR1 and ESR2 polymorphisms on the risk of all dementias (P = .04). Hormone treatment did not modify these associations, and there were no significant associations in men. CONCLUSIONS: Although there was only weak support for a gender-specific association between the common ESR1rs2234693 polymorphism and AD, this polymorphism may act as an effect modifier, modifying the association between an ESR2 polymorphism and dementia, as well as the risk of AD associated with the APOE ε4 allele.