Natural vs. surgical postmenopause and psychological symptoms confound the effect of menopause on executive functioning domains of cognitive experience

Abstract

Objective:

The menopause transition is associated with difficulties in executive function. However, it is unclear whether these difficulties persist past perimenopause. This study investigated whether potential confounders, including natural vs. surgical postmenopause and menopause-related psychological symptoms, influence whether executive dysfunction persists into postmenopause.

Study Design:

A cross-sectional sample of women aged 35–65 years (N = 1971) in one of four groups, premenopause, perimenopause, natural postmenopause, and surgical postmenopause, were surveyed. Participants self-reported executive functioning with the Brown Attention Deficit Disorder Scale (BADDS), anxiety symptom severity with the Generalized Anxiety Disorder Questionnaire (GAD-7), and depression symptom severity with the Center for Epidemiologic Studies Depression Scale (CES-D).

Main Outcome Measures:

We analyzed the association between group and BADDS scores using linear regression models – first, by controlling for age, education, and self-reported attention deficit hyperactivity disorder (ADHD) diagnosis (Model #1) and, second, by further controlling for current difficulty sleeping, anxiety, and depression (Model #2).

Results:

In both models, BADDS scores were significantly elevated (indicating more difficulties in executive function) among women in the perimenopausal and surgical postmenopausal groups compared with those in the premenopausal group. Likewise, the perimenopausal and surgical postmenopausal groups had the highest proportions of participants who reported difficulty sleeping and clinical levels of anxiety and depression. BADDS scores were significantly higher in natural postmenopausal vs. premenopausal women without controlling for difficulty sleeping, anxiety, and depression (Model #1), but not when adjusting for these variables (Model #2).

Conclusions:

The type of menopause and psychological symptoms are important confounders of the relationship between the menopause transition and executive dysfunction, and help explain whether executive dysfunction persists or recovers in postmenopause.

1. Introduction

The menopause transition involves dramatic changes in levels of reproductive hormones and functioning of the hypothalamic-pituitary-ovarian axis. During this time, women frequently report struggles with aspects of cognition, including learning, memory, and executive functioning. Executive functioning domains of cognition include initiating and sustaining focus on tasks, temporarily retaining and utilizing information (working memory), processing speed, and motivation for work [1–4]. These executive functioning cognitive domains are especially relevant to the menopause transition because they are highly dependent on the prefrontal cortex (PFC), the neural structure and function of which is modulated by ovarian hormones such as estradiol [5–7]. However, the majority of research into the effects of the menopause transition on cognition have focused on domains of learning and memory, as opposed to exclusively examining executive functioning.

On learning and memory tasks, perimenopausal and postmenopausal women show poorer performance compared to premenopausal women, independent of [1, 8–10]. However, the findings in this literature are inconsistent as to whether these cognitive difficulties persist past perimenopause or if they resolve in postmenopause. One longitudinal study found that cognitive difficulties peak during perimenopause and recover in postmenopause to premenopausal levels [8]. Other longitudinal studies found that the perimenopausal decline in cognitive function does not recover or only partially recovers in postmenopause compared to premenopausal levels [1, 11]. With the negative impact that cognitive difficulties have on quality of life, it is critical to determine what factors contribute to these symptoms during the menopause transition and what domains are most affected. The present study investigates how confounders, such as the type of postmenopause (natural vs. surgical) and current psychological symptoms, impact domains of executive functioning in different stages of menopause.

Natural and surgical postmenopause are distinct experiences with differing effects on cognitive function. Surgical postmenopausal women show more cognitive impairments compared to natural postmenopausal women, independent of age [12–14]. Several differences between natural and surgical postmenopause provide rationale for why cognitive dysfunction is relatively more pronounced following surgical postmenopause. First, the most common indication for surgical menopause (bilateral oophorectomy) is as part of gynecological or breast cancer care or risk reduction, necessitating procedures that induce an abrupt and permanent cessation of ovarian hormone production compared to the natural menopause transition [15]. Second, surgical menopause often occurs before women have entered perimenopause, resulting in an earlier age at postmenopause [16], and earlier postmenopause age is associated with poorer performance on verbal fluency and visual memory tasks [17]. Third, women who undergo surgical menopause as part of cancer care may experience cognitive impairments related to having cancer and/or undergoing chemotherapy [18, 19]. Thus, postmenopause type (natural vs. surgical) is important to consider when studying cognition across menopause stages. However, studies about cognition and the menopause transition often exclude participants with surgical postmenopause, group natural and surgical postmenopause together, or include relatively small sample sizes of surgical postmenopausal women [8, 11, 20].

Psychological symptoms (e.g., disrupted sleep, increased anxiety and depression) also confound the relationship between cognition and menopause stage/type, as these symptoms commonly occur across the menopause transition [21, 22] and negatively impact cognitive functioning [23, 24]. More specifically, disrupted sleep, anxiety, and depression also negatively impact PFC-dependent executive functions [25–27]. However, studies control for these psychological variables inconsistently, which may help explain the observed variability among findings as to whether cognitive problems persist into postmenopause. Furthermore, the prevalence of symptoms that negatively impact cognition, such as depression and anxiety, is higher in women who underwent surgical menopause compared to age-matched women who did not have a bilateral oophorectomy [28]. Despite this evidence, at present, no previous studies have investigated executive functioning and the menopause transition while accounting for possible confounding effects of both postmenopause type and psychological symptoms.

The purpose of this research is to investigate within the same study whether postmenopause type (natural or surgical) and current psychological symptoms (difficulty sleeping, anxiety, and depression) account for the persistence or recovery of cognitive difficulties, specifically executive functioning domains, in postmenopause. We utilized the Brown Attention Deficit Disorder Scale (BADDS) [29] to assess executive functioning in premenopausal, perimenopausal, and natural and surgical postmenopausal women. The BADDS is a validated self-report measure of symptoms of attention hyperactivity deficit disorder (ADHD) that captures multiple domains of executive functioning [30]. The BADDS has previously been used to assess executive dysfunction during the menopause transition and the effectiveness of pharmaceutical treatments aimed at improving executive function [2, 24, 31, 32]. We analyzed the effect of menopause stage and type (premenopause, perimenopause, and natural and surgical postmenopause) on BADDS scores, with and without controlling for possible confounding variables, to better understand the relationship between the menopause transition and self-reported cognitive complaints and why discrepancies in the literature exist about the persistence of these cognitive difficulties, including executive dysfunction.

2. Methods

2.1. Participants

We used convenience sampling to recruit biological females ages 35–65. To sufficiently sample women who had undergone surgical menopause, we recruited participants from ResearchMatch with hereditary cancer risk who were affiliated with the Facing Our Risk of Cancer Empowered (FORCE) group (https://www.facingourrisk.org/). We also recruited participants from listservs at the Penn Center for Women’s Behavioral Wellness. To be eligible, participants were required to live in the United States, be able to speak and read English, and not be currently pregnant or breastfeeding.

At the start of the online survey, participants were provided with information about the study including contact information for study staff and the Office of Clinical Research at the University of Pennsylvania. Participants were then asked to respond to the question “Do you agree to take part in this study?” with the option to provide consent to participate by selecting “Yes, I agree to take part in this study”, or to select “No, I do not agree to take part in this study”, in which case they were not asked any further questions. The study was approved by the University of Pennsylvania and the University of Colorado Anschutz Medical Campus human investigations review boards.

2.2. Measures

2.2.1. Menopause stage and type

Participants self-selected their menopause stage based on terminology from the 2001 Stages of Reproductive Aging Workshop (STRAW) [33] (summarized in Table 1; full wording is available in Supplementary Materials). Women who identified as postmenopausal were also asked whether their transition to postmenopause was natural or surgical. Women who reported having experienced premature ovarian failure, primary ovarian insufficiency, chemical menopause after chemotherapy or radiation, or any type of menopause other than natural or surgical (bilateral oophorectomy) were excluded from the study.

Table 1.

Stages of reproductive aging.

Stage	Menstrual cycle criteria
Premenopause	Relatively regular menstrual cycles Absence of menopausal symptoms such as hot flashes and night sweats
Perimenopause	Either a fewer or greater number of days between menstrual periods Possible presence of menopausal symptoms such as hot flashes and night sweats
Postmenopause	Absence of a menstrual period for 12 months or more

2.2.2. Medical history

Participants completed a questionnaire about their medical history, including yes/no questions about whether they had a history of cancer or had received chemotherapy. Diagnosis of ADHD and difficulty sleeping were determined by categorical (yes/no) responses to the questions: “Have you been diagnosed with Attention Deficit Hyperactive Disorder (ADHD)?” and “Do you have difficulty sleeping?” (Table 2). Women with a history of brain cancer, a brain tumor, or meningioma were excluded from analyses.

Table 2.

Participant demographics according to menopause group.

Variable	N with data	Overall (N = 1971)	Pre (N = 597)	Peri (N = 331)	Natural Post (N = 724)	Surgical Post (N = 319)	p value
Age (years) (mean ± SD)	1914	51.5 (8.7)	42.2 (4.3)	49.0 (4.4)	59.0 (4.3)	54.2 (7.9)	<0.00
BMI (mean ± SD)	1967	29.2 (7.8)	28.8 (8.0)	30.3 (8.2)	28.3 (7.2)	31.0 (7.7)	<0.00
Ethnicity	1964						0.153
Hispanic/Latino descent		91 (4.6%)	37 (6.2%)	17 (5.1%)	23 (3.2%)	14 (4.4%)
Not Hispanic/Latino		1873	558	313	697 (96.3%)	305 (95.6%)
Race	1961						0.041
White		1705	498	285	648 (89.8%)	274 (86.2%)
Black/African American		118 (6.0%)	38 (6.4%)	17 (5.2%)	38 (5.3%)	25 (7.9%)
American Indian/Alaska Native		7 (0.4%)	3 (0.5%)	1 (0.3%)	1 (0.1%)	2 (0.6%)
Others/multiracial		131 (6.7%)	53 (9.0%)	26 (7.9%)	35 (4.8%)	17 (5.3%)
Marital status	1966						<0.00
Single (never married)		331	165	48	92 (12.7%)	26 (8.2%)
Married or domestic partnership		1179	338	218	407 (56.2%)	216 (67.7%)
Widowed		51 (2.6%)	6 (1.0%)	5 (1.5%)	34 (4.7%)	6 (1.9%)
Divorced		370	77	51	179 (24.7%)	63 (19.7%)
Separated		35 (1.8%)	10 (1.7%)	8 (2.4%)	9 (1.2%)	8 (2.5%)
Education	1968						<0.00
Less than college degree		559	142	90	205 (28.4%)	122 (38.2%)
College graduate		626	178	123	224 (31.0%)	101 (31.7%)
Graduate degree		783	276	117	294 (40.7%)	96 (30.1%)
Household income	1971						0.001
< $50,000		472	135	60	180 (24.9%)	97 (30.4%)
$50,000 to $100,000		667	213	111	232 (32.0%)	111 (34.8%)
$100,000 to $200,000		620	203	117	224 (30.9%)	76 (23.8%)
> $200,000		146 (7.4%)	37 (6.2%)	29 (8.8%)	55 (7.6%)	25 (7.8%)
Unknown or did not disclose		66 (3.3%)	9 (1.5%)	14 (4.2%)	33 (4.6%)	10 (3.1%)
Has ever had cancer	1971						<0.00
Yes		275	34 (5.7%)	35	106 (14.6%)	100 (31.3%)
No		1696	563	296	618 (85.4%)	219 (68.7%)
Chemotherapy for any cancer	1971	73 (3.7%)	12 (2.0%)	5 (1.5%)	20 (2.8%)	36 (11.3%)	<0.00
Has difficulty sleeping	1971	1240	338	237	443 (61.2%)	222 (69.6%)	<0.00
Has been diagnosed with ADHD	1971	203	71	53	48 (6.6%)	31 (9.7%)	<0.00
Meets GAD-7 criteria for anxiety	1971	481	159	100	123 (17.0%)	99 (31.0%)	<0.00
Meets CES-D criteria for depression	1971	473	139	103	135 (18.6%)	96 (30.1%)	<0.00
Age at final menstrual period (FMP) (years)	n/a	n/a	n/a	n/a	49.7 (5.3)	40.8 (7.8)	n/a

2.2.3. Executive function

The Brown Attention Deficit Disorder Scale (BADDS) [29] is a 40-item questionnaire that assesses five subscales of executive functioning (summarized in Table 3):

Table 3.

Description and score range of BADDS subscales.

Scale	Description	Number of	Score range
Total	Summary of all subscales	40	0–120
Subscale 1	Organizing and activating to work	9	0–27
Subscale 2	Sustaining attention and concentration	9	0–27
Subscale 3	Sustaining energy and effort	9	0–27
Subscale 4	Managing affective interference	7	0–21
Subscale 5	Utilizing working memory and accessing	6	0–18

Subscale 1: Organizing and Activating for Work. Nine items assessed excessive difficulties in getting organized, getting started on work-related tasks, and self-activating for daily routines.

Subscale 2: Sustaining Attention and Concentration. Nine items assessed chronic problems in sustaining attention when doing work-related tasks, such as excessive daydreaming or distractibility when listening or doing required reading, or repeatedly losing track and needing to re-read assigned material.

Subscale 3: Sustaining Alertness, Effort, and Processing Speed. Nine items assessed problems in keeping up consistent alertness and effort for work-related tasks, daytime drowsiness, slow processing of information, or inadequate task completion.

Subscale 4: Managing Affective Interference. Seven items assessed difficulties with mood, sensitivity to criticism, lack of motivation, excessive frustration, or discouragement.

Subscale 5: Using Working Memory and Accessing Recall. Six items assessed issues with forgetfulness in daily tasks and routines, problems in recall of learned material, and losing track of necessary items.

For each item in the questionnaire, participants reported the extent to which it had been a problem over the last six months (0 = never, 1 = once a week or less, 2 = twice a week, or 3 = almost daily). Total BADDS scores range from 0–120, with higher scores indicating more self-reported difficulties with executive functioning.

2.2.4. Anxiety

Participants completed the Generalized Anxiety Disorder Questionnaire (GAD-7) to assess the presence and severity of anxiety symptoms. A score of ≥ 10 represents moderate to severe anxiety symptoms and the cutoff for clinically significant anxiety [34, 35]. We grouped participants with a GAD-7 score ≥ 10 as those who met the clinical cutoff for anxiety and those with a score < 10 into those who did not (Table 2).

2.2.5. Depression

Participants completed the Center for Epidemiologic Studies Depression Scale (CES-D) to assess the presence and severity of depression symptoms. A score of ≥ 16 represents the cutoff for clinical depression in the general population [36]. We grouped participants with a CES-D score ≥ 16 into those who met the cutoff for clinically significant depression and those with a score < 16 into those who did not (Table 2).

2.3. Statistical Analysis

2.3.1. BADDS validation

We tested the internal validity of the BADDS assessment in our sample of women in premenopause, perimenopause, and natural or surgical postmenopause by running Pearson’s correlations between total BADDS scores and scores for each of the five subscales (Table 4), with the hypothesis that if the measurement were valid, total BADDS scores and subscale scores should be highly positively correlated with one another (i.e., participants who scored highly on one subscale would likely score highly on other subscales as well).

Table 4.

Matrix of Pearson’s correlation coefficients for BADDS total and subscale scores in our sample of women in premenopause, perimenopause, and postmenopause (natural or surgical).

	Total	Subscale 1	Subscale 2	Subscale 3	Subscale 4	Subscale 5
Total	1.00***	0.92***	0.87***	0.89***	0.76***	0.81***
Subscale 1	0.92***	1.00***	0.73***	0.82***	0.66***	0.65***
Subscale 2	0.87***	0.73***	1.00***	0.68***	0.55***	0.69***
Subscale 3	0.89***	0.82***	0.68***	1.00***	0.63***	0.67***
Subscale 4	0.76***	0.66***	0.55***	0.63***	1.00***	0.50***
Subscale 5	0.81***	0.65***	0.69***	0.67***	0.50***	1.00***

Numbers indicate the Pearson’s r correlation between items in the matrix.

^***p < 0.001.

2.3.2. Linear regression models

To analyze the effect of menopause group on BADDS scores, we ran linear regression models with premenopause, perimenopause, natural postmenopause, and surgical postmenopause as predictors for each of the BADDS outcome variables (total BADDS and five subscale scores) separately and conducted pairwise comparisons between the four menopause groups, adjusting for multiple comparisons using the Bonferroni correction (α = 0.0083).

2.3.3. Covariate selection

All models included current age and education as covariates, as these variables have been shown to impact BADDS scores independent of menopausal stage/type [11, 20] and are frequently included as covariates in analyses of executive functioning (Fuh et al., 2003; Ryan et al., 2014). All models also included ADHD as this condition impacts BADDS scores [30].

The first linear regression model (Model #1) only included age, education, and ADHD as covariates. In addition to these covariates, the second linear regression model (Model #2) also included difficulty sleeping (self-report), anxiety symptom severity (GAD-7 score), and depression symptom severity (CES-D score) to test whether these variables confounded the effect of menopause group on BADDS scores observed in Model #1 (Figure 1). We report associations in the form of unstandardized regression coefficients, β, from the multivariable linear regression models. Statistical analyses were performed in R version 4.2.0 [37].

Figure 1. Covariates for the linear regression models analyzing the effects of menopause group on BADDS scores.

Menopause group consisted of premenopause, perimenopause, natural postmenopause, and surgical postmenopause. GAD-7: Generalized Anxiety Disorder Questionnaire; CES-D: Center for Epidemiologic Studies Depression Scale; BADDS: Brown Attention Deficit Disorder Scale.

3. Results

3.1. Participants

Demographic data for study participants is presented overall and by menopause group in Table 2. Based on the self-selection criteria for menopause stage and type, women were divided into premenopause (N = 597), perimenopause (N = 331), natural postmenopause (N = 724), and surgical postmenopause (N = 319) groups. Most participants were White (86.9%), not Hispanic/Latino (95.0%), and married or in a domestic partnership (59.8%). The most represented level of education and household income was a graduate degree (39.8%) and $50,000-$100,000 (33.8%), respectively. A history of cancer was present in 14.0% of participants. Raw means for each group, prior to adjusting for any covariates, are presented in Figure S1.

The number of participants in each menopause group who reported a diagnosis of ADHD, difficulty sleeping, and clinically significant levels of anxiety or depression are presented in Table 2. Perimenopausal women and surgical postmenopausal women had the highest prevalence of difficulty sleeping, anxiety, and depression, while the prevalence was lowest in premenopausal and natural postmenopausal women.

3.2. Validation of BADDS among the study sample

Total BADDS and subscale scores all correlated significantly with one another (p < 0.001) with large effect sizes (r ≥ 0.50) [38], indicating high internal reliability of the BADDS in this study (Table 4).

3.3. Covariate selection

Given the possibility that a history of cancer and/or chemotherapy could impact executive functioning [18, 39, 40], we tested for their associations with BADDS scores. We first analyzed the effect of cancer and chemotherapy on BADDS scores in separate unadjusted linear regression models. Cancer history was not significantly associated with total BADDS score (p = 0.573) or any of the five subscale scores (p ≥ 0.074). Chemotherapy history was significantly associated with total BADDS score (p = 0.036) and Subscale 5 (p = 0.001). However, when chemotherapy history was included as a covariate in the model with menopause group, age, education, and ADHD, it no longer had a significant effect on total BADDS score (p = 0.133). Therefore, cancer and chemotherapy histories were not included as covariates in our analyses. Previous studies have shown that an earlier postmenopausal age and a longer time in postmenopause are both associated with greater cognitive impairments [17, 41]. Surgical postmenopausal women had a younger average age at final menstrual period (FMP) than natural postmenopausal women (40.8 vs. 49.7 years; Table 2). However, we did not include age at FMP or time since FMP as a covariate in our models for several reasons. First, FMP data were only available for the natural and surgical postmenopausal groups and therefore could not be included in our models, all of which compared these groups with premenopausal and perimenopausal women. Second, in analyses comparing only natural and surgical postmenopausal groups, current age had a stronger association with BADDS scores than age at FMP or time since FMP, likely due to the cross-sectional nature of this study. Therefore, we adjusted for current age rather than age at FMP or time since FMP.

3.4. Effects of menopause stage and type on BADDS scores

3.4.1. Model #1: Controlling for age, education, and ADHD

Model #1 examined the effect of menopause group on BADDS scores, adjusting for age, education, and ADHD (F_7,1903 = 43.49; p < 0.001; adjusted R² = 0.14). Menopause group significantly impacted total BADDS score (F_3,1903 = 27.37, p < 0.001), as well as BADDS scores on each of the subscales (Subscale 1 [organizing and activating to work]: F_3,1903 = 23.53, p < 0.001; Subscale 2 [sustaining attention and concentration]: F_3,1903 = 23.25, p < 0.001; Subscale 3 [sustaining energy and effort]: F_3,1903 = 24.16, p < 0.001; Subscale 4 [managing affective interference]: F_3,1903 = 19.35, p < 0.001; Subscale 5 [utilizing working memory and accessing recall]: F_3,1903 = 21.92, p < 0.001).

Pairwise comparisons between groups for Model #1 are presented in Table 5 (Bonferroni corrected at α = 0.0083). Compared to premenopausal women, perimenopausal women had a total BADDS score that was 11.13 points higher (p < 0.001), as well as significantly higher scores on each subscale (p < 0.001). Natural postmenopausal women had a total BADDS score that was 8.18 points higher than premenopausal women (p < 0.001), as well as significantly higher scores on each of the subscales except for Subscale 3. No significant differences emerged between perimenopausal women and natural postmenopausal women.

Table 5.

Model #1: Pairwise comparisons between menopause groups, adjusted for age, education, and ADHD.

Comparison →	Peri – Pre		Natural Post – Pre		Natural Post – Peri
Outcome ↓	Mean difference (95% CI)	p value	Mean difference (95% CI)	p value	Mean difference (95% CI)	p value
BADDS total score	11.13 (7.29, 14.97)	<0.001	8.18 (3.46, 12.90)	<0.001	−2.95 (−7.03, 1.13)	0.156
Subscale 1	2.33 (1.33, 3.33)	<0.001	1.94 (0.71, 3.17)	0.002	−0.38 (−1.44, 0.68)	0.479
Subscale 2	2.62 (1.61, 3.63)	<0.001	1.99 (0.75, 3.23)	0.002	−0.63 (−1.70, 0.44)	0.245
Subscale 3	2.27 (1.37, 3.17)	<0.001	1.25 (0.14, 2.36)	0.028	−1.02 (−1.98, −0.06)	0.036
Subscale 4	2.07 (1.32, 2.82)	<0.001	1.46 (0.54, 2.38)	0.002	−0.61 (−1.40, 0.19)	0.133
Subscale 5	1.84 (1.18, 2.51)	<0.001	1.54 (0.72, 2.36)	<0.001	−0.30 (−1.01, 0.40)	0.403
Comparison →	Surgical Post – Pre		Surgical Post – Peri		Surgical Post – Natural Post
Outcome ↓	Mean difference (95% CI)	p value	Mean difference (95% CI)	p value	Mean difference (95% CI)	p value
BADDS total score	15.75 (11.29, 20.21)	<0.001	4.62 (0.44, 8.80)	0.030	7.57 (3.98, 11.16)	<0.001
Subscale 1	3.86 (2.70, 5.02)	<0.001	1.53 (0.44, 2.62)	0.006	1.92 (0.98, 2.85)	<0.001
Subscale 2	3.95 (2.78, 5.12)	<0.001	1.33 (0.24, 2.43)	0.017	1.97 (1.02, 2.91)	<0.001
Subscale 3	2.77 (1.72, 3.81)	<0.001	0.50 (−0.48, 1.48)	0.32	1.52 (0.68, 2.36)	<0.001
Subscale 4	2.27 (1.40, 3.14)	<0.001	0.20 (−0.61, 1.02)	0.627	0.81 (0.11, 1.51)	0.023
Subscale 5	2.90 (2.13, 3.68)	<0.001	1.06 (0.33, 1.78)	0.004	1.36 (0.74, 1.98)	<0.001

Differences were calculated based on estimated marginal means controlling for age, education, and ADHD. Significance was adjusted for multiple comparisons at α = 0.0083 using the Bonferroni correction. Bolded effects indicate p < 0.0083.

Women who underwent surgical menopause had a total BADDS score that was 15.75 points higher compared to premenopausal women (p < 0.001), as well as higher scores on each of the subscales (p < 0.001). Between surgical postmenopausal women and perimenopausal women, there was no significant difference in total BADDS score (p = 0.030); however, scores on Subscale 1 (p = 0.006) and Subscale 5 (p = 0.004) were significantly elevated in surgical postmenopausal compared to perimenopausal women. Compared to natural postmenopausal women, surgical postmenopausal women had a total BADDS score that was 7.57 points higher (p < 0.001), as well as significantly higher scores on each subscale except for Subscale 4.

Overall, surgical postmenopausal women self-reported the most difficulties with executive functioning followed by perimenopausal women, as can be observed in Figure 2 and Table 6.

Figure 2. Model #1: Total BADDS scores by menopause group, adjusted for age, education, and ADHD.

Data are presented as estimated marginal mean (± standard error) relative to premenopausal participants. Multiple pairwise comparisons are Bonferroni corrected (α = 0.0083) and presented in Table 4. Values are presented as raw data in Table 5. ***p < 0.001.

Table 6.

Model #1: Estimated marginal means (± standard error), adjusted for age, education, and ADHD.

Marginal mean (SE)	Pre	Peri	Natural Post	Surgical Post
BADDS total score	43.7 (1.7)	54.9 (1.6)	51.9 (1.5)	59.5 (1.7)
Subscale 1	11.4 (0.4)	13.8 (0.4)	13.4 (0.4)	15.3 (0.4)
Subscale 2	10.8 (0.4)	13.5 (0.4)	12.8 (0.4)	14.8 (0.4)
Subscale 3	9.0 (0.4)	11.3 (0.4)	10.2 (0.4)	11.8 (0.4)
Subscale 4	6.9 (0.3)	9.0 (0.3)	8.4 (0.3)	9.2 (0.3)
Subscale 5	5.5 (0.3)	7.4 (0.3)	7.1 (0.3)	8.4 (0.3)

3.4.2. Model #2: Controlling for age, education, ADHD, difficulty sleeping, anxiety, and depression

Model #2 examined the effect of menopause group on BADDS scores, adjusting for age, education, ADHD, difficulty sleeping, anxiety, and depression (F_10,1900 = 160.80, p < 0.001, adjusted R² = 0.46). Model #2 was a significantly better fit for the data than Model #1 as indicated by an ANOVA comparison between the two models for total BADDS scores (F_3,1900 = 374.79, p < 0.001). Each of the additional covariates in Model #2 had a significant effect on total BADDS scores (difficulty sleeping: β = 7.40, p < 0.001; anxiety: β = 15.27, p < 0.001; depression: β = 24.04, p < 0.001).

As with Model #1, menopause group significantly impacted total BADDS score (F_3,1900 = 43.50, p < 0.001), as well as scores on each of the subscales (Subscale 1 [organizing and activating to work]: F_3,1900 = 34.72, p < 0.001; Subscale 2 [sustaining attention and concentration]: F_3,1900 = 30.71, p < 0.001; Subscale 3 [sustaining energy and effort]: F_3,1900 = 35.15, p < 0.001; Subscale 4 [managing affective interference]: F_3,1900 = 33.06, p < 0.001; Subscale 5 [utilizing working memory and accessing recall]: F_3,1900 = 27.19, p < 0.001).

Pairwise comparisons between groups for Model #2 are presented in Table 7 (Bonferroni corrected at α = 0.0083). As with Model #1, perimenopausal women had significantly elevated total BADDS scores relative to premenopausal women (5.11 points, p < 0.001). However, Subscales 1 and 3 were no longer significantly different between these groups. In contrast to Model #1, natural postmenopausal women no longer showed any significant differences in total BADDS score or subscale scores relative to premenopausal women. As with Model #1, no significant differences in BADDS scores emerged between perimenopausal and natural postmenopausal women.

Table 7.

Model #2: Pairwise comparisons adjusted for age, education, ADHD, difficulty sleeping, anxiety, and depression.

Comparison →	Peri – Pre		Natural Post – Pre		Natural Post – Peri
Outcome ↓	Mean difference (95% CI)	p value	Mean difference (95% CI)	p value	Mean difference (95% CI)	p value
BADDS total score	5.11 (2.04, 8.18)	0.001	2.95 (−0.81, 6.71)	0.124	−2.16 (−5.40, 1.07)	0.190
Subscale 1	0.87 (0.04, 1.70)	0.040	0.68 (−0.34, 1.70)	0.190	−0.19 (−1.06, 0.69)	0.672
Subscale 2	1.30 (0.41, 2.18)	0.004	0.84 (−0.24, 1.92)	0.129	−0.46 (−1.39, 0.47)	0.334
Subscale 3	1.00 (0.24, 1.75)	0.010	0.15 (−0.78, 1.07)	0.757	−0.85 (−1.65, −0.06)	0.036
Subscale 4	0.89 (0.32, 1.47)	0.002	0.41 (−0.30, 1.12)	0.255	−0.48 (−1.09, 0.13)	0.120
Subscale 5	1.06 (0.45, 1.66)	<0.001	0.87 (0.14, 1.61)	0.020	−0.18 (−0.82, 0.45)	0.572
Comparison →	Surgical Post – Pre		Surgical Post – Peri		Surgical Post – Natural Post
Outcome ↓	Mean difference (95% CI)	p value	Mean difference (95% CI)	p value	Mean difference (95% CI)	p value
BADDS total score	7.65 (4.07, 11.22)	<0.001	2.53 (−0.79, 5.86)	0.135	4.70 (1.85, 7.55)	0.001
Subscale 1	1.90 (0.94, 2.87)	<0.001	1.03 (0.14, 1.93)	0.024	1.22 (0.45, 1.99)	0.002
Subscale 2	2.18 (1.15, 3.21)	<0.001	0.88 (−0.07, 1.84)	0.070	1.34 (0.52, 2.16)	0.001
Subscale 3	1.06 (0.18, 1.94)	0.018	0.06 (−0.75, 0.88)	0.883	0.91 (0.21, 1.61)	0.011
Subscale 4	0.63 (−0.04, 1.30)	0.065	−0.26 (−0.88, 0.36)	0.413	0.22 (−0.31, 0.76)	0.418
Subscale 5	1.87 (1.17, 2.57)	<0.001	0.82 (0.17, 1.47)	0.014	1.00 (0.44, 1.56)	<0.001

Differences were calculated based on estimated marginal means controlling for age, education, ADHD, difficulty sleeping, anxiety, and depression. Significance was adjusted for multiple comparisons at α = 0.0083 using the Bonferroni correction. Effects in italics are those that are no longer significant compared to Model #1, which controlled for only age and education. Bolded effects indicate p < 0.0083.

Like Model #1, women who underwent surgical menopause still had a significantly elevated total BADDS score compared to premenopausal women (7.65 points, p < 0.001). However, in contrast to Model #1, Subscales 3 and 4 were no longer significantly different between these groups. No significant differences emerged between surgical postmenopausal and perimenopausal women in Model #2. Surgical postmenopausal women had total BADDS scores that were 4.70 points higher than natural postmenopausal women (p = 0.001); all subscales except for Subscales 3 and 4 were also significantly higher.

Overall, surgical postmenopausal women again self-report the most difficulties with executive functioning, followed by perimenopausal women as can be observed in Figure 3 and Table 8.

Figure 3. Model #2: Total BADDS scores by menopause group, adjusted for age, education, ADHD, difficulty sleeping, anxiety, and depression.

Data are presented as estimated marginal mean (± standard error) relative to premenopausal participants. Multiple pairwise comparisons are Bonferroni corrected (α = 0.0083) and presented in Table 6. Values are presented as raw data in Table 7. **p = 0.001; ***p < 0.001.

Table 8.

Model #2: Estimated marginal means (± standard error), adjusted for age, education, ADHD, difficulty sleeping, anxiety, and depression.

Marginal mean (SE)	Pre	Peri	Natural Post	Surgical Post
BADDS total score	53.8 (1.4)	58.9 (1.3)	56.7 (1.3)	61.4 (1.4)
Subscale 1	13.9 (0.4)	14.7 (0.4)	14.5 (0.3)	15.8 (0.4)
Subscale 2	12.9 (0.4)	14.2 (0.4)	13.7 (0.4)	15.1 (0.4)
Subscale 3	11.2 (0.3)	12.2 (0.3)	11.3 (0.3)	12.3 (0.3)
Subscale 4	9.1 (0.3)	10.0 (0.3)	9.5 (0.2)	9.7 (0.3)
Subscale 5	6.7 (0.3)	7.8 (0.3)	7.6 (0.2)	8.6 (0.3)

3.5. Adjusting for varying ranges in BADDS subscales

In our analyses of the effects of menopause stage and type on BADDS scores, we also wanted to investigate whether menopause stage and type differentially affected specific domains of executive functioning measured by the BADDS subscales (Table 3). Based on the pairwise comparisons between groups in Model #1 (Table 5) and in Model #2 (Table 7), no clear differences between menopause groups emerged on the subscales. Since the BADDS subscales had different ranges, we also analyzed raw subscale scores as percentages of the maximum subscale score to correct for differences in ranges between the subscales. The percentage transformation revealed that scores for all menopause groups were highest on Subscale 1 (organizing and activating to work), followed closely by Subscale 2 (sustaining attention and concentration) (Table S1), further indicating that menopause stage and type does not appear to differentially impact the domains of executive functioning measured by the subscales.

4. Discussion

To our knowledge, this is the first study to examine executive dysfunction across the menopause transition while considering the potential confounding effects of both postmenopause type and psychological symptoms. Although postmenopause type (natural vs. surgical) and symptoms such as difficulty sleeping, anxiety, and depression are all factors that are known to impact PFC-dependent executive functions, these factors had not been considered conjointly within the same study prior to our investigation. The main conclusions from this research are twofold. First, this study supports the premise that natural and surgical postmenopause are distinct experiences when it comes to difficulties with executive functioning domains of cognition. In statistical models with and without potential confounders, surgical postmenopausal women had BADDS scores that were significantly higher than both premenopausal and natural postmenopausal women. These findings indicate that surgical postmenopause is associated with heightened subjective executive dysfunction. All analyses controlled for age at the time of the study, so these findings are not confounded by surgical postmenopausal participants being younger on average than natural postmenopausal participants (54.2 vs. 59.0 years). Notably, the BADDS scores of surgical postmenopausal women were not significantly different from those of perimenopausal women in either model, suggesting that executive dysfunction is similar between these groups. Second, psychological symptoms – such as sleep disruption, anxiety, and depression – are critical confounders of the relationship between the menopause transition and executive dysfunction that need to be considered. Specifically, when these variables were added as covariates in our model, BADDS scores in natural postmenopausal women no longer significantly differed from premenopausal women, indicating that, in women who undergo menopause naturally, self-reported executive dysfunction does seem to improve after the perimenopause peak in executive function difficulties. However, BADDS scores in natural postmenopausal women were not significantly lower than in perimenopausal women, who still had higher BADDS scores than premenopausal women, suggesting that this recovery in natural postmenopause is only partial. Together, these findings support the conclusion that executive functioning after perimenopause may partially recover to premenopausal levels, but only for natural postmenopausal women and only when accounting for the confounding effects of important psychological symptoms – namely, difficulty sleeping, anxiety, and depression.

In this study, cancer and chemotherapy history did not explain why surgical postmenopausal women report worse executive dysfunction, as these variables were not significantly associated with BADDS scores in models adjusted for age, education, and ADHD. This finding emphasizes the relevance of the psychological covariates we considered (difficulty sleeping, depression, and anxiety). The absence of an effect of chemotherapy on BADDS is consistent with past research that found a history of chemotherapy did not impact BADDS scores among women with BRCA1 or BRCA2 germline mutations who had undergone risk-reducing salpingo-oophorectomy [24]. However, the prevalence of chemotherapy was low among participants in the present study (Table 2), possibly limiting our ability to detect an association with BADDS scores. Although cancer and chemotherapy history are still important to consider when investigating cognitive impairment in surgical postmenopausal women, studies on cancer-related cognitive impairment have largely disagreed on the extent to which cancer and chemotherapy impair cognition and which domains of performance are most affected and for how long [18]. The lack of a significant effect of cancer or chemotherapy on BADDS scores emphasizes the importance of difficulty sleeping and clinically significant levels of anxiety and/or depression on executive dysfunction. Interestingly, the prevalence of these confounders was highest among surgical postmenopausal women and perimenopausal women, further highlighting the similarities between these two groups.

Although the psychological symptoms examined in this study are clearly relevant to executive function across the menopause transition, it is not possible to determine whether the relationship is causal and, if so, the direction of that causality. For example, hormonal changes associated with the menopause transition increase sleep difficulties and symptoms of anxiety and depression, which may then impair executive function [23, 27, 42]. Alternatively, or additionally, the menopause transition may increase executive dysfunction, leading to difficulty sleeping and depression and anxiety symptoms due to stress and reduced quality of life stemming from cognitive difficulties [4]. Difficulty sleeping and anxiety and depression symptoms may also have independent effects on, or be independently affected by, executive dysfunction. Future studies designed to test for mediation effects are necessary to disentangle the causal relationship between psychological symptoms and executive function during the menopause transition, which may be bidirectional. Future research should also consider more detailed measures of the psychological variables examined in this study, as well as others not investigated here. For example, using a validated metric for sleep like the Pittsburgh Sleep Quality Index [43] would allow for an investigation into how the frequency, severity, and nature of sleep difficulties are associated with executive functioning difficulties. Vasomotor symptoms, such as hot flashes, are also common during the menopause transition and are associated with greater self-reported struggles with cognitive function [4].

Despite the important contributions of this study for understanding the role of potential confounders in the relationship between cognition and menopause, four main limitations are critical to consider. First, all measures were assessed via self-reports, which are more susceptible to participant biases than objective measures. While the BADDS has been used to accurately assess severity of ADHD symptoms and efficacy of ADHD medications in clinical trials [44, 45], previous research has shown that perceptions of cognitive function do not reliably map onto concurrent measures of cognitive task performance [46, 47]. However, the BADDS has the advantage of being able to gauge perceptions about multiple domains of executive functioning quickly within the same instrument, including ability to initiate engagement with a task, maintain focus on the task, regulate emotions that may interfere with focus, and utilize working memory to sustain performance [44, 45]. By contrast, objective tests of cognitive performance are time consuming and limited to measuring a specific domain of cognition. Utilizing the BADDS allowed us to observe that menopause stage and type did not clearly relate to perceived deficits on particular subdomains of executive functioning over others but rather was more important for perceptions of global executive functioning (total BADDS scores). Furthermore, subjective reports of cognitive difficulty are relevant to perceived quality of life [45] and have been shown to predict future declines in objective measures of cognitive performance [48]. Self-reports of cognition are therefore important to consider in research and clinical practice, as they are highly relevant to menopausal women’s quality of life in real-world contexts, such as in the workplace.

The second limitation is that we sampled participants cross-sectionally rather than longitudinally, meaning that we compared groups of participants in different stages and types of menopause rather than measuring changes in executive functioning within the same participants across the menopause transition. Longitudinal studies have greater power to detect cognitive changes over time by accounting for individual differences in cognitive baselines and variation in the rate of change [49, 50]. However, using a cross-sectional design and measuring executive function via self-report also offered several advantages. First, it allowed us to achieve a large sample size within each menopause group, including surgical postmenopausal participants, who are often underrepresented in research on the menopause transition [51, 52]. Additionally, we were able to avoid repeated cognitive testing over time, which can lead to performance increases due to practice, making it more difficult to isolate the effect of age as well as menopause stage and type on cognitive function [53]. Nevertheless, a longitudinal study would be necessary to confirm our findings from this cross-sectional study of executive function difficulties.

The third limitation is that this study was not equipped to examine the impact of hormone therapy (HT) use on executive function. A thorough analysis of the effects of HT need to include details such as type of HT, when HT was initiated, and duration of use, both according to chronological age and relative to menopausal stage and type. These details about HT were not possible to ascertain in this cross-sectional survey study. However, HT during the menopause transition is important to consider given previous findings that hormonal birth control [54, 55] and HT [8, 56] may impact cognitive function. However, the directionality of HT effects on cognition are complex as HT during menopause can be either beneficial or detrimental depending on the cognitive domain examined [17] and/or the timing and duration of HT use [20, 56–58]. For instance, estrogen therapy may benefit cognition if started during perimenopause but not postmenopause, though results are inconclusive [14, 15, 56]. For surgical postmenopausal women, estrogen therapy benefits cognition if initiated prior to approximately age 50, but the efficacy declines with age [56]. However, estrogen is contraindicated for some postmenopausal women, such as those at high risk for breast cancer [59]. More research into the use of HT for cognition in postmenopausal women, including the timing of treatment (according to chronological and reproductive age) and the associated risks vs. benefits, is needed. Given the complexity of the relationship between the menopause transition and executive function, it will be important in future studies, especially longitudinal ones, to evaluate the impact of HT.

Lastly, the fourth limitation is that the findings from this study primarily come from White, educated, English-speaking women who had access to the internet to complete the online questionnaires. Experiences and symptoms during the menopause transition can vary in different countries [60] and depend on social and cultural factors [61–63]. Future investigations should consider postmenopause type (natural vs. surgical) and psychological symptoms in a broader, cross-cultural sample of participants.

5. Conclusions

Collectively, this research supports the conclusion that natural and surgical postmenopause need to be considered as distinct experiences and that difficulties with executive functioning are heightened in surgical postmenopause. Psychological symptoms, including difficulty sleeping, depression, and anxiety, also need to be considered as potential confounders in the relationship between executive dysfunction and the menopause transition, specifically whether executive functioning difficulties persist in postmenopause. Without accounting for psychological symptoms, natural postmenopausal women have increased difficulties with executive functioning similar to perimenopausal women. However, when accounting for difficulty sleeping, depression, and anxiety, natural postmenopausal women show executive function difficulties that are partially improved to premenopausal levels. The relevance of psychological symptoms and postmenopause type to perceived executive dysfunction will be important for researchers and clinicians to consider in the context of health-related quality of life for menopausal women.

Highlights.

• Compared with premenopausal women, perimenopausal and surgical postmenopausal women experience more perceived executive dysfunction.

• Problems with perceived executive dysfunction were significantly higher in natural postmenopausal than in premenopausal women without controlling for difficulty sleeping, anxiety, and depression, but not when adjusting for these variables.

• Menopause type (natural vs. surgical) and psychological symptoms (sleep, anxiety, and depression) are important confounders of the relationship between menopause and perceived executive function.