Menstrual cycle characteristics across the reproductive lifespan and cognitive function in midlife women.

In the analytical sample of 19,904 participants, the mean age (SD) at cognitive assessment was 62 (4.9) years (range, 50–74 years). At 29 to 46 years, 12.5% of the participants had irregular cycles, and compared to those with very regular cycles, they had a higher BMI, were older at exposure and outcome assessment, were more likely to report gynecologic conditions, and were less likely to use OC ( Table 1 ). Cycles ≤25 days and ≥32 days were present in 16.7% and 14.2% of the participants, respectively. Compared to women with cycles 26 to 31 days, those with cycles ≤25 days were older at exposure and outcome assessment, were more likely to be past or current smokers and to report uterine fibroids, whereas those with cycles ≥32 days had a higher BMI, were more likely to report PCOS phenotype and less likely to use OC. We did not observe differences in cognitive function by cycle regularity at 14 to 17 and 18 to 22 years ( Table 2 ). However, cycle regularity at 29 to 46 years was related to lower learning/working memory compared to regular cycles (regular: β , −0.02; 95% CI, −0.05–0.00; irregular cycles: β , −0.05; 95% CI, −0.08 to −0.01). For reference, each year of age was associated with a −0.03 (95% CI, −0.03 to −0.03) score in learning/working memory. Therefore, the differences in this composite for women with irregular cycles are equivalent to being 1.7 years older. The findings were consistent after adjusting for T2D ( Supplemental Table 1 ). Cycle length at younger but not older ages was associated with cognitive function ( Table 3 ). Compared to women with a cycle length of 26 to 31 days at 18 to 22 years, those with cycles ≤25 days scored lower in learning/working memory ( β , −0.05; −0.09 to −0.02), while women with cycles ≥32 days scored higher in both composite scores (psychomotor speed/attention: β , 0.04; 0.01–0.07; learning/working memory: β , 0.03; 0.01–0.05). The findings were unchanged after adjusting for T2D ( Supplemental Table 2 ). We then evaluated the relation between changes in cycle characteristics over time and cognitive function ( Table 4 ). Compared to women without changes in cycle regularity from 14 to 17 years to 29 to 46 years, women whose cycles became irregular later in life had lower learning/working memory scores ( β , −0.05; −0.10 to −0.01). A similar association was observed for change in cycle regularity from 18 to 22 years to 29 to 46 years. Interestingly, women with cycles ≤25 days at 18 to 22 years whose usual cycle length changed to 26 to 31 days at 29 to 46 years had lower learning/working memory scores and a suggestion of lower psychomotor speed/attention scores than women whose usual cycle length was 26 to 31 days at both time points. When we categorized women who reported no periods separately from those reporting irregular cycles, those with no menstrual periods at 29 to 46 years exhibited lower scores for learning/working memory in midlife ( Supplemental Table 3 ). The findings were consistent with the main analysis when we conducted analyses for cycle length using categories aligned with recommended clinical cutoffs ( Supplemental Table 4 ). The results observed in the main analysis remained consistent when we adjusted for history of depression ( Supplemental Table 5 ), when we excluded women with gynecological conditions ( Supplemental Table 6 ) and participants over 40 years in 1993 ( Supplemental Table 7 ). When we studied OC users in a separate category, the results were mostly consistent with the main findings, although the positive associations of cycles ≥32 days at 18 to 22 years with both cognitive composites were attenuated ( Supplemental Table 8 ). This analysis within a large longitudinal cohort showed that adult cycle characteristics, specifically cycles ≤25 days at 18 to 22 years and irregular cycles at 29 to 46 years, were associated with lower learning/working memory scores. The differences in cognitive performance associated with these characteristics were equivalent to being 1.7 years older and are consistent with the impacts of known factors affecting cognitive function. For example, midlife individuals exhibiting any adverse lifestyle behavior—such as poor diet quality, physical inactivity or smoking—showed a difference of −0.05 z scores in learning/working memory compared to individuals without any adverse behavior. 35 Menstrual cycle characteristics are potential markers of hormonal exposures relevant to chronic disease risk and cognitive health. 7 , 36 Previous studies have shown changes in cognitive performance across the menstrual cycle that coincide with fluctuations in the levels of ovarian hormones. 7 For example, there is evidence that aspects of cognition, such as implicit and working memory, are enhanced in the late follicular and mid-luteal phases, which are characterized by high levels of estradiol, the predominant type of estrogen during the reproductive years. 37 , 38 Fox et al hypothesized that a higher cumulative exposure to estrogens, reflected by the number of lifetime menstrual cycles, would lower the risk of Alzheimer’s disease (AD). 39 In this study of 89 British women, each additional month of having a menstrual cycle decreased AD risk nonsignificantly by 0.3%. 39 It has been suggested that women with short cycle lengths experience more cycles during the reproductive years and greater cumulative exposure to estradiol, 36 which would translate into better cognitive outcomes. However, short cycles have also been linked to lower ovarian reserve and, consequently, a higher risk of early natural menopause (menopause <45 years). 40 , 41 Early natural menopause is associated with a 19% to 36% increased risk of dementia. 42 , 43 One hypothesis for this heightened risk is reduced lifetime exposure to estrogens. 42 , 43 This supports the observed association between cycles ≤25 days at 18 to 22 years and lower learning/working memory, a cognitive domain particularly influenced by estrogens. 8 We observed that women with irregular cycles at 29 to 46 years had lower performance in learning/working memory in midlife. To our knowledge, the only other study on cycle characteristics and cognitive function was conducted in 242 women with schizophrenia or schizoaffective disorder (mean age 38.1 years, SD 10.2 years). 16 In this study, compared to premenopausal women with regular cycles, those with irregular cycles had lower performance in verbal memory, psychomotor speed, and verbal fluency. Discrepancies between the studies, including disease status, age at cognitive assessment, and the cognitive domains assessed, could account for the inconsistent findings. Reduced exposure to estrogens may be an underlying factor linking irregular menstrual cycles and lower learning/working memory. 44 Our sensitivity analysis revealed a strong association between no periods at 29 to 46 years, possibly indicative of greater estrogen deprivation, and lower learning/working memory scores, supporting the hypothesis of estrogen deficiency as a contributing factor to cognitive outcomes in women with irregular cycles. Another possible explanation is the presence of PCOS, a condition characterized by irregular cycles and hyperandrogenism. 45 However, the association persisted when we excluded women with a PCOS phenotype. Nevertheless, we cannot rule out residual confounding as we relied on a proxy variable. An additional explanation may be that the association observed with irregular cycles at ages 29 to 46 years is a different manifestation of the same signal we observed with short cycles at ages 18 to 22 years. As oocyte depletion progresses across the reproductive years, there is a parallel decline in anti-Mullerian hormone levels, which may lead to a shortening in cycle length and, eventually, irregular menstrual cycles. 46 – 48 Therefore, we hypothesize that the combined pattern of short menstrual cycles during the early reproductive years followed by irregular cycles during mid-reproductive years may indicate a faster decline in reproductive function which could shorten lifetime exposure to estrogens and potentially compromise cognitive function. Further research is warranted to elucidate the mechanisms underlying our findings. While outside of the scope of this study, reproductive factors related to cycle characteristics such as parity, gravidity, and breastfeeding history 49 – 51 might influence cognitive health by altering exposure to hormones like estrogens. Future research should explore these relationships to understand how reproductive traits impact cognitive health. Our study has several strengths. To our knowledge this is the first study to evaluate the association between menstrual cycle characteristics at different times of the reproductive lifespan and cognitive function. The large sample size, assessment of menstrual cycle characteristics years prior to cognitive assessment, the assessment of cognition at an early aging stage, and the adjustment for several important confounders are key strengths. Some limitations must be acknowledged. First, cycle characteristics at 14 to 17 and 18 to 22 years were retrospectively reported, potentially introducing recall bias. Second, the wide age range in 1993 when cycle characteristics were reported is a limitation, as having irregular cycles could reflect distinct pathophysiology and stages of ovarian aging. Third, caution is needed in interpreting our findings since we did not adjust for multiple comparisons. Fourth, the study utilized a single cognitive function rather than a longitudinal approach. Fifth, while key confounders were adjusted for, we cannot rule out residual confounding. Furthermore, we did not adjust for precision covariates such as family history of dementia. Sixth, we did not explore subcategories based on combinations of cycle characteristics and OC use. Seventh, predominantly non-Hispanic White and highly educated nurses comprised our population, limiting generalizability. Differences in cycle characteristics across racial and ethnic groups, 52 and varying rates of cognitive decline and dementia, 53 , 54 underscore the need for more diverse studies. Finally, while menstrual cycles might serve as markers of hormonal exposures, the underlying pathophysiological processes causing these imbalances and potentially affecting cognitive function were beyond our study’s scope.

The NHS II is an ongoing cohort of 116,429 female nurses aged 25 to 42 years at enrollment in 1989. Participants complete mailed or online questionnaires biennially. 17 We assessed cognitive function in a substudy with 2 enrollment waves, 2014 to 2019 or 2018 to 2022. Appendix A in the supplement describes details regarding this substudy. Of 20,282 participants with 1 cognitive assessment, we excluded those with all menstrual cycle characteristics missing (n=7) and those diagnosed with cancer by 1993 (n=370) due to chemotherapy-induced changes in cycle characteristics and cognitive function. 18 , 19 The analytical sample included 19,904 participants with data on at least 1 cycle characteristic. The Institutional Review Boards of Brigham and Women’s Hospital and the Harvard T.H. Chan School of Public Health approved the study protocol. At enrollment in 1989, participants retrospectively reported their cycle regularity during high school (14–17 years) and cycle regularity and length at 18 to 22 years, excluding periods of pregnancy, lactation, or when using oral contraceptives (OCs). Participants reported their current cycle regularity and length in 1993 (age range: 29–46 years). We excluded postmenopausal women by 1993 (n=1356) from the analysis of cycle characteristics at this time point. Participants reported cycle characteristics in the categories of very regular (within 3–4 days of the expected period), regular (within 5–7 days), usually irregular, always irregular, or no periods for regularity, and 50 days or too irregular to estimate for length, defined as the interval from the first day of the period to the first day of the next period. We studied a single assessment of cognitive function obtained with the Cogstate brief battery—an online, validated battery with proven clinical utility in identifying cognitive impairment and dementia. 20 , 21 This battery includes the tasks of detection (to measure psychomotor function and information processing speed), identification (to measure visual attention and vigilance), 1 card learning (to measure visual learning and short-term memory), and 1 back (to measure attention and working memory). 22 , 23 As previously described, 24 , 25 we transformed task scores to improve normality and then internally standardized them to z scores with a mean of 0 and a standard deviation (SD) of 1. We then averaged the standardized scores to create the composite score of psychomotor speed/attention (mean of the detection and identification scores) and learning/working memory (mean of the 1 card learning and 1 back scores). Higher z scores reflect better cognitive performance. We recorded age at cognitive assessment. At enrollment in 1989, the participants reported their race and ethnicity, height, age at menarche, body size at ages 5 and 10 years (ie, somatotype), 26 smoking, physical activity, and alcohol intake at various age intervals preceding enrollment to the study (eg, high school and ages 18–22). Participants reported their highest educational level in 2018, as well as their parents’ education and occupation in 2005. We queried several characteristics in 1989 and biennially, including age, weight, smoking, physical activity, OC use, physician-diagnosed T2D, and depression. We calculated body mass index (BMI) (in kg/m 2 ) from self-reported weight and height at age 18 and each assessment time point. 27 We estimated neighborhood socioeconomic status (SES) at enrollment and each assessment time point. 28 We assessed alcohol intake and diet quality with a validated semiquantitative food frequency questionnaire applied every 4 years beginning in 1991. Diet quality was summarized with the 2010 Alternative Healthy Eating Index. 29 In 1989, women reported any history of severe acne during their teenage years; in 1991, they reported history of physician-diagnosed hirsutism. Self-reported physician-diagnosed polycystic ovaries were recorded in 1989 and 1993. We defined a PCOS phenotype considering any self-reported clinical signs of hyperandrogenism (history of severe acne or hirsutism), or polycystic ovaries by 1993. Participants reported information on endometriosis confirmed by laparoscopy, and uterine fibroids in 1993. The primary exposures were menstrual cycle regularity at 14 to 17, 18 to 22, and 29 to 46 years, and cycle length at 18 to 22 and 29 to 46 years. For the analyses, we considered the categories of very regular (reference), regular, and irregular (combining usually irregular, always irregular, and no periods) for cycle regularity, and ≤25 days, 26 to 31 days (reference), and ≥32 days for cycle length. Secondary exposures were changes in cycle length and regularity from age 14 to 17 or 18 to 22 years to age 29 to 46; for regularity, we compared the categories of regular (combining very regular and regular) and irregular. We estimated mean differences (95% confidence intervals [CIs]) in each cognitive composite z score by cycle characteristics using linear regression. We tested linear trends by treating cycle characteristics as ordinal variables. For all the exposures, we adjusted for age at cognitive assessment, non-Hispanic White race and ethnicity, participants’ education, wave of cognitive assessment, parental education, parental occupation, neighborhood SES in 1993, and age at menarche. Furthermore, we accounted for exposure-specific covariates reflecting OC use, adiposity, and lifestyle at different time points. For instance, for cycle regularity at 14 to 17 years, we further adjusted for childhood body size as a metric of adiposity preceding high school, as well as smoking status, alcohol intake, strenuous physical activity, and OC use during high school. For regularity and length at 18 to 22 years, we accounted for characteristics at 18 to 22 years, including BMI, smoking status, alcohol intake, strenuous physical activity, and OC use. The models for cycle regularity and length at 29 to 46 years also accounted for current age, BMI, smoking and OC use, as well as diet quality, alcohol, and physical activity assessed in 1991. We carried forward data 1 assessment cycle for covariates with missing values or imputed the mean for continuous variables or the mode for categorical variables. To contextualize our results, we used linear regression to estimate the cognitive function and age association while adjusting for race and ethnicity. In a sensitivity analysis, we adjusted the primary exposures for T2D diagnosed by the 1993 assessment to control for confounding, considering its known association with both menstrual cycle characteristics and cognitive function. 30 – 32 Additionally, recognizing that the absence of menstrual periods may reflect a distinct hormonal environment compared to irregular cycles, we analyzed women reporting no periods as a separate group. Furthermore, we conducted analyses for cycle length using categories aligned with recommended clinical cutoffs for identifying short and long cycles. 33 We conducted further sensitivity analyses only on the primary exposures where we observed significant associations. We first adjusted for a self-reported history of physician-diagnosed depression (29.4%) prior to the cognitive assessment (ie, by 2013 for wave 1 and 2017 for wave 2) as depression may be an independent predictor of cognitive function. 34 Furthermore, we excluded women with PCOS phenotype and subsequently excluded those with endometriosis and uterine fibroids to rule out confounding by these gynecological conditions. Then, we excluded women >40 years in 1993 from the analysis of cycle characteristics at this time point to avoid misclassification of women experiencing early menopause signs. Finally, since OC are often used to manage menstrual cycle disturbances, we categorized OC users into a separate exposure category. We used SAS statistical software version 9.4 (SAS Institute, Cary, NC) for all the analyses. We relied on 95% CIs to assess the precision of our estimates and to infer statistical significance. We did not adjust for multiple comparisons.

In this cohort of US women, we found that cycles ≤25 days at 18 to 22 years and irregular cycles at 29 to 46 years were associated with lower performance in learning/working memory in midlife. These findings are important because memory impairment is characteristic of AD, the leading cause of dementia. It is crucial for future research to validate and further explore the biological mechanisms behind these associations.

Dementia is a significant and growing health concern affecting around 57 million individuals globally. 1 , 2 Women are affected by dementia 1.7 times more than men, 1 leading to questions about the role of female-specific reproductive exposures in this disparity. 3 , 4 Menstrual cycle characteristics during the reproductive years are potential indicators of hormonal exposures relevant to cognitive health and possibly dementia. Estrogens have neuroprotective effects and influence different aspects of cognition, such as verbal memory, fluency, and fine motor skills. 5 – 7 Altered menstrual cycle regularity or length may reflect disruption in the hypothalamic-pituitary-ovarian axis and signal an imbalance in hormone levels, including reduced estrogen levels, that could contribute to a later risk of cognitive impairment and dementia. 7 – 9 These hormonal imbalances and altered menstrual cycles might result from conditions such as polycystic ovary syndrome (PCOS), which is characterized by irregular cycles and high levels of androgens and has been associated with poor cognitive performance. 10 , 11 Beyond their reflection of reproductive hormones, menstrual cycle characteristics can also be associated with cognitive outcomes through risk factors like type 2 diabetes (T2D) and cardiovascular disease. 12 – 15 Despite the plausible associations between menstrual cycle characteristics and cognitive function, epidemiologic evidence is limited. 16 We examined the association between menstrual cycle characteristics at 3 stages of a woman’s reproductive lifespan and cognitive function in midlife, using data from the Nurses’ Health Study II (NHS II).