Associations of Polycystic Ovary Syndrome With Indicators of Brain Health at Midlife in the CARDIA Cohort

Polycystic ovary syndrome (PCOS), a prevalent disorder frequently diagnosed at adolescence or in young adulthood, is known primarily for its reproductive manifestations, including oligomenorrhea, hyperandrogenism, and infertility.^1,2 PCOS is also associated with adverse metabolic features, including insulin resistance (IR), which is independent of, but exacerbated by, obesity.³ Additional cardiovascular risk factors are also common in PCOS, including dyslipidemia, hypertension, and inflammation.⁴ Finally, PCOS associates with mental health outcomes, including depression and anxiety.^5,6

Increasingly, in population studies, adverse brain health outcomes that occur later in life, such as dementia, have been linked to metabolic and cardiovascular health factors at midlife.⁷ Furthermore, the antecedents of these late-life brain changes may be detectable by midlife or earlier,⁸ with midlife cognitive changes also associating with cardiometabolic health.^9,10 Surprisingly, whether women with PCOS are at risk of adverse brain health outcomes has been only minimally investigated. Existing reports on cognitive function, all with relatively small samples, have focused on young adult populations and have suggested possible decrements in specific areas of cognitive functioning, particularly executive functioning and verbal fluency.^11-13 Although it is unclear which aspects of PCOS might contribute to differences in cognition, possible mechanisms include altered signaling by insulin and/or androgens in the brain, inflammation, or early vascular changes. Even fewer reports have examined brain MRI findings in PCOS, and again, existing studies have focused on young populations.^12,13

Significant changes in cognitive function and brain structure at midlife may foretell a greater risk of cognitive decline with aging.^14,15 However, to date, there have been no reports regarding brain health among women with PCOS at midlife and beyond. Accordingly, the goal of this study was to investigate measures of brain health, including cognitive performance and brain MRI, in women with and without a history of PCOS enrolled in the Coronary Artery Risk Development in Young Adults (CARDIA) study, a 30-year longitudinal cohort of Black and White Americans.

Of the 1,163 women in the CARDIA Women's Study (CWS), 907 completed cognitive testing, and of these, 65 (7.1%) met criteria for PCOS determined by hyperandrogenism and oligomenorrhea during young adulthood. The Figure illustrates the timing of measures relevant for our analysis, including PCOS determination. Characteristics of the PCOS and no PCOS groups are presented in Table 1. The groups were similar for age, education, income, and drinking/smoking status, but the PCOS group was more likely to be White (61.6% vs 46.4%; p = 0.03) and to have diabetes (27.3% vs 13.8%; p = 0.009) as well as higher glucose. In analyses that controlled for age, race, years of education, and study center, the PCOS group had lower performance (ß coefficient for difference in z scores (95% CI)) on the Stroop (−0.323 (−0.69 to −7.37); p = 0.008), RAVLT (−0.254 (−0.473 to −0.034); p = 0.002), and category fluency (−0.267 (−0.480 to −0.040); p = 0.02) tests, with no difference on letter fluency test and DSST. Participants with PCOS also demonstrated lower performance on a composite score that comprised the average z scores for all 5 tests (−0.190 (−0.327 to −0.052); p = 0.007). We found no evidence of significant interactions between PCOS status and age, education, and race for cognitive outcomes, although these findings may be limited by the PCOS sample size.

To interrogate the independence of the association between PCOS and cognitive test outcomes, we constructed 2 additional models (Table 2). Model 2 included covariates that differed across the groups at the p < 0.15 level, including diabetes, depression, and fasting glucose. In this model, the effect of PCOS was maintained for the composite score, though with a small reduction in magnitude and significance (−0.171 (−0.309 to −0.035); p = 0.02). A third model incorporated additional variables with potential to affect cognitive outcomes, including current smoking and alcohol use, BMI, systolic blood pressure, and vigorous physical activity (model 3). Here, again, PCOS remained similarly associated with the composite score (−0.176 (−0.315 to −0.040); p = 0.01). In considering individual cognitive tests, it is notable that there was no attenuation of the PCOS effect for both the RAVLT (−0.281 (−0.502 to −0.059) p = 0.01) and category fluency test (−0.338 (−0.571 to −0.104); p = 0.005) in the fully adjusted model (Model 3), although the magnitude of the effect of PCOS for the Stroop test was diminished (−0.229 (−0.474 to 0.017) p = 0.08), suggesting confounding by one or more of the model factors.

To determine potential relationships between cognitive outcomes and the key hormonal and metabolic parameters frequently disrupted in PCOS, we explored associations between serum androgens (free and total testosterone; year 16) and metabolic factors (glucose and insulin; year 15) (Table 3). Given the role of androgens in determining PCOS status, analyses were stratified by PCOS group. We found that for the PCOS group, androgens, particularly free testosterone (log), were associated with cognitive results, including for the DSST, category frequency, and composite scores. For non-PCOS participants, free testosterone was also associated with the composite score and DSST.

We next turned our attention to a comparison of brain MRI outcomes across the PCOS groups. To pursue this question, we used the CARDIA Brain substudy, which offered CARDIA participants the opportunity to complete MRI examinations at years 25 and 30. We found that of 291 women who completed an MRI examination at one of these time points, 25 (8.5%) met PCOS criteria. Here, we did not find a difference in abnormal white matter across PCOS groups; however, we found that the PCOS group had decreased total white matter fractional anisotropy, a measure of white matter integrity, compared with those without PCOS (coeff (95% CI) −0.013 (−0.021 to −0.005); p = 0.002). We found broad impacts across brain regions, although the most robust difference was noted in the corpus callosum, an area important for connectivity and executive functions (Table 4). All MRI analyses were controlled for age, race, education, and study center.

PCOS is a common disorder that affects approximately 8%–15% of women and presents with hyperandrogenism, oligomenorrhea, and polycystic ovaries.²⁶ Although not part of the diagnostic criteria, women with PCOS frequently have IR that is independent of, but exacerbated by, obesity. Using a longitudinal cohort of Black and White individuals, we found that those with PCOS demonstrated lower cognitive performance compared with non-PCOS counterparts at midlife. In addition, in a smaller group of participants of a brain imaging substudy, we found evidence of lower white matter integrity for PCOS. Given that both hormonal and metabolic factors affect the brain, it is surprising that only handful of studies to date have examined cognition in PCOS, and no study has yet reported on cognition in midlife women. Indeed, PCOS research for the past several decades has focused on fertility treatments and management of cardiovascular risk, but our results suggest a need for additional investigations focused on brain health.

Longitudinal studies in general populations indicate that age-related declines in cognitive function may begin at midlife,⁸ and both cognitive batteries and brain MRI comprise some of the modalities that might suggest preclinical changes in brain health.¹⁵ To date, the extant literature describing cognitive function in PCOS has largely focused on links between hyperandrogenism and performance on specific cognitive tests in younger women.^12,13,27-29 In 2007, investigators found lower scores on tests of verbal fluency, verbal memory, manual dexterity, and visual-spatial learning in participants with PCOS (n = 29; mean age 27 years) compared with age-matched controls.²⁹ More recently, Sukhapure showed that higher free testosterone levels in women with and without PCOS (mean age 28.7 years) specifically affected psychomotor speed and visuospatial learning.²⁷ By contrast, we recently reported lower performance in verbal fluency, working memory, and cognitive control in participants with PCOS (n = 48; mean age 31 years) with hyperandrogenism compared with nonandrogenic counterparts.¹¹ In this study, we further found that IR seemed to confer an additional negative effect on cognitive outcomes, expanding upon several other small investigations that have described links between decreased cognitive performance and metabolic health in PCOS.^12,13

This study is the first to compare cognitive performance between individuals with and without PCOS at midlife. We observed differences in measures of memory, cognitive control, and verbal fluency using models that were controlled for age, race, education, and study center. Inclusion of diabetes and depression, both of which are common in PCOS and are known to associate with cognitive performance, only slightly attenuated the magnitude and significance of the PCOS relationship with results from the composite score. In a third model, incorporating a wider array of potential confounders, including smoking/drinking behavior, systolic blood pressure, and BMI, the link between PCOS and cognitive performance persisted. Taken together, these models describe an independent relationship between PCOS and cognitive outcomes.

The mechanistic underpinnings of these findings remain to be elucidated, although our finding of an association between free testosterone, measured in earlier adulthood, and cognition at midlife highlights a potential role for androgens. Alternatively, cardiovascular risk factors, particularly diabetes, are also more common in PCOS, although the PCOS associations with cognition were not meaningfully diminished when controlling for diabetes and other cardiovascular risk factors, including systolic blood pressure and adiposity. Depressive symptoms have also been consistently shown to be increased in people with PCOS^6,30 and could affect cognitive function. Although our cognitive findings were robust to models controlling for depressive symptoms, it remains possible that greater attention to the mental health needs of those with PCOS might represent a window of opportunity to improve cognitive health.

Brain imaging modalities, including structural MRI and diffusion tensor imaging, are increasingly used in research settings to assess preclinical changes that may portend cognitive decline. White matter fractional anisotropy is a well-validated, continuous measure of white matter integrity that associates with subtle cognitive deficits, particularly in executive functions and processing speed.^31,32 Only one other study used diffusion tensor imaging in a PCOS cohort and found lower fractional anisotropy in the corpus callosum compared with controls, although this study was small (n = 19 participants with PCOS) and participants were in the reproductive years.¹³ Recent reports using the CARDIA cohort have reported associations between cardiovascular risk factors and lower fractional anisotropy.^24,33 Whether midlife brain health in PCOS could be improved through early interventions targeting cardiometabolic health should be a focus on future research.

Important strengths should be noted for this study. For one, we used a population-based cohort of individuals with and without PCOS defined by evidence of both oligomenorrhea and hyperandrogenism during the reproductive years. This approach minimizes the impact of referral bias, which can exaggerate the differences between PCOS and control populations. Furthermore, the rigorous quality control in the CARDIA study lends reassurance about the validity of the brain health outcomes and the wide range of covariates. Nonetheless, there are several limitations to consider. As a cross-sectional study, we are unable to determine whether our findings represent a more rapid decline in cognitive function that begins at midlife for PCOS or persistence of a difference that was present before midlife. We are also unable to account for the degree to which unmeasured confounders may have contributed to our results. The diagnosis of PCOS was not made by a physician, rather on the basis of serum androgen levels and self-report of oligomenorrhea and hirsutism. As such, it is likely that some misclassification is present, although we believe this would be likely to bias our findings toward the null. Similarly, because the CARDIA study did not allow for assessment of ovarian markers required for Rotterdam diagnostic criteria, our results pertain to those meeting the narrower NIH PCOS criteria, which require both hyperandrogenism and oligomenorrhea. Our results, therefore, may not be generalizable to those with PCOS who do not meet the NIH criteria. Finally, our results describe a relative difference in brain health measures between those with and without PCOS, but we are unable to assess the degree to which these findings represent a clinically important deficit.

In summary, this report of midlife cognition in PCOS raises a new concern about another potential comorbidity for individuals with this common disorder. At the individual level, cognitive deficits at midlife may affect quality of life, professional attainment, and financial security. Moreover, our findings may indicate the potential for an accelerated cognitive decline trajectory in PCOS. Given that up to 10% of women may be affected by PCOS, these results have important implications for public health at large. Larger studies with longitudinal designs are needed to both validate our findings and to identify potential differences in cognitive trajectories. On a positive note, our results suggest that several modifiable factors, such as cardiovascular and mental health, could contribute to our findings. Current recommendations for PCOS include management of cardiovascular risk and screening for depression, and our work suggests that appropriate management of these aspects may serve to also improve brain aging for this population.

The Coronary Artery Risk Development in Young Adults (CARDIA) study is conducted and supported by the National Heart, Lung, and Blood Institute (NHLBI) in collaboration with the University of Alabama at Birmingham (HHSN268201800005I and HHSN268201800007I), Northwestern University (HHSN268201800003I), University of Minnesota (HHSN268201800006I), and Kaiser Foundation Research Institute (HHSN268201800004I). This manuscript has been reviewed by CARDIA for scientific content. CARDIA was also partially supported by the Intramural Research Program of the National Institute on Aging (NIA) and an intra-agency agreement between NIA and NHLBI (AG0005). We would like to acknowledge R.N. Bryan and the MRI reading center of the Department of Radiology at the University of Pennsylvania.

CME Course: NPub.org/cmelist