P3-144: Association of hormone replacement therapy and cognitive impairment: Evidence of confounding by demographics and serum α-tocopherol status

Abstract

Studies of hormone replacement therapy (HRT) and cognitive function in older women have varied from beneficial to possibly detrimental. Some studies have suggested that cognitive effects of HRT may differ by age at HRT initiation, or interact with other cognitive risk factors. Cognitive testing methodology has differed between studies, and observational studies have not always had data on relevant potential confounders. This analysis of data from an observational ancillary study to the Women's Health Initiative (WHI) examines whether HRT use is associated with cognitive impairments of memory or mixed–domain type, determined by a neuropsychological test battery. Study subjects were 523 women age 60 years and over at time of cognitive testing. Estrogen use history (“never” (N=243), “past” (N=140) or “current” (N=140)) was determined from previous WHI data and updated at cognitive test visit. Memory impairment was considered present if women scored at least 2 sd below age and education–specific norms for one or more tests of memory. Mixed domain impairment was defined as scoring below norms on one or more tests in two or more domains (memory, executive function, language, attention, and visual function). In unadjusted analyses, neither current (OR 1.1 95% CI 0.7–1.7) nor past (0.6 95% CI 0.3–1.0) HRT use was associated with memory impairment. Reduced odds of mixed–domain impairment were seen for past HRT users (0.3; 0.1–0.7) but not current users (0.7; 0.4–1.3). Among ever–users, no association was seen between mixed domain impairment and age at first use (p=.58) or use of progestins (p=.23). Adjusting for previously–identified covariates (age, education, race/ethnicity, and lowest quartile serum α–tocopherol), the association of past HRT use with mixed impairment was no longer statistically significant (OR 0.5; 0.2–1.1). Adjusting additionally for Apo e4 or cardiovascular disease did not affect these results. No evidence of interactive effects was seen between HRT use and age, serum α–tocopherol, or e4. No association was seen between current HRT use and memory or mixed–domain impairments. Confounding by age, education, race/ethnicity, and serum tocopherol status appeared to account for the bivariate association between past HRT use and mixed domain cognitive impairment.