Cognitive function and endogenous sex hormones in older women

OBJECTIVE: To determine if endogenous hormone levels predict cognitive function in older women. DESIGN: A longitudinal, population-based study. SETTING: Rancho Bernardo, California PARTICIPANTS: A total of 393 community-dwelling women aged 55 to 89 years who were not using replacement estrogen. METHODS: Between 1984 and 1987, sera were collected for measurement of total and bioavailable testosterone, total and bioavailable estradiol, and estrone. Between 1988 and 1991, 12 standard neuropsychological tests were administered, including two items from the Blessed Information-Memory-Concentration Test, three measures of retrieval from the Buschke-Fuld Selective Reminding Test, a category fluency test, immediate and delayed recall from the Visual Reproduction Test, the Mini-Mental State Examination (MMSE) with individual analysis of the Serial 7's and the "World" Backwards components, and the Trail-Making Test part B (Trails B). The association between the five hormones and the 12 cognitive function tests was tested in age- and education-adjusted analyses using linear regression, partial correlation, quintile and categorical analyses. RESULTS: Women with better MMSE scores (> 23) had significantly higher adjusted mean total and bioavailable testosterone levels (P = .009; P = .02, respectively). Using linear regression, the mean total testosterone was significantly associated with better performance on the World component of the MMSE (b = .12; P = .08). With regard to estrogen, the only statistically significant (P = .02) association was better performance on one test in women with very low levels of estradiol. CONCLUSIONS: In these older women, higher endogenous estrogen levels were not associated with significantly better performance on any cognitive function test. In contrast, higher levels of testosterone predicted better categorical performance on the MMSE and the World component of the MMSE. These novel findings warrant further research.