Statement of Research Significance
Research Question(s) or Topic(s): This study examined whether cognitive difficulties during menopause, often described as “brain fog,” are associated with measurable changes in cognitive performance. It also reviewed the tools currently used to assess subjective cognition in this population. Main Findings: A small but significant correlation was found between subjective cognition and learning efficiency, or the ability to learn and retain new information. No significant correlations were observed between subjective complaints and other cognitive domains. Across studies, 12 different tools with varying psychometric properties were used to assess subjective cognition. Study Contributions: This study is the first to synthesize correlations between subjective and objective cognition during menopause using meta-analysis. Findings highlight a need for more reliable and standardized tools to assess menopause-related cognitive concerns and suggest that reduced learning efficiency may underlie self-reported menopausal brain fog. Findings inform future research and may improve clinical recognition and support for cognitive concerns during menopause.
Introduction
Menopause, which is defined as the day of the final menstrual period and is diagnosed retrospectively after 12 consecutive months of amenorrhea, has a median age of 51 years and results from an age-related diminished supply of ovarian follicles. Perimenopause, the transitional phase around the menopause, spans approximately four to 10 years leading up to menopause and ends a year after the final menstrual period (Harlow et al., Reference Harlow, Gass, Hall, Lobo, Maki, Rebar, Sherman, Sluss and de Villiers2012). Perimenopause is associated with fluctuating and declining levels of sex hormones, including estradiol and progesterone. For many women, perimenopause is also accompanied by a range of physical, behavioral, and cognitive changes, including hot flushes, night sweats, sleep disturbance, and depression, all of which can have adverse impacts on a woman’s quality of life (Harlow et al., Reference Harlow, Gass, Hall, Lobo, Maki, Rebar, Sherman, Sluss and de Villiers2012).
It is estimated that between 62 and 67% of women experience cognitive symptoms during perimenopause, with common symptoms including difficulty concentrating, losing their train of thought, forgetfulness, and challenges with multi-tasking (Reuben et al., Reference Reuben, Karkaby, McNamee, Phillips and Einstein2021). Many report that these cognitive symptoms impact their daily lives, affecting work performance, social interactions, and overall quality of life (Woods et al., Reference Woods, Coslov and Richardson2023). For some, the cognitive symptoms raise concerns about early-onset dementia or reflect underlying ADHD symptoms (Epperson et al., Reference Epperson, Shanmugan, Kim, Mathews, Czarkowski, Bradley, Appleby, Iannelli, Sammel and Brown2015; Maki & Jaff, Reference Maki and Jaff2024).
Colloquially, the term “brain fog” has become synonymous with cognitive symptoms during menopause. Although brain fog is not a recognized clinical syndrome, it provides a meaningful way for women to describe their experience of subjective cognitive symptoms during the menopause transition years. Additionally, the term has been increasingly utilized in research and clinical practice (Maki & Jaff, Reference Maki and Jaff2022). Hence, we define “menopausal brain fog” as subjective cognitive symptoms related to menopause and will use these terms interchangeably. As the neuropsychological profile of brain fog during menopause remains unclear, the aim of this systematic review and meta-analysis is to provide an overview of current research that has examined the relationship between subjective cognitive symptoms and objective cognitive performance during the menopause transition years.
There is currently no formal measure of menopausal brain fog. Research studies that aim to identify or measure subjective cognitive decline during menopause typically use self-report questionnaires assessing symptoms of forgetting or other memory-related complaints, such as the Memory and Cognitive Confidence Scale (MACCS; Ballantyne et al., Reference Ballantyne, King and Green2021), Memory Functioning Questionnaire (MFQ; Maki et al., Reference Maki, Gast, Vieweg, Burriss and Yaffe2007), Attentional Functional Index (AFI; Grummisch et al., Reference Grummisch, Sykes Tottenham and Gordon2023), or the Multifactorial Memory Questionnaire (MMQ; Unkenstein et al., Reference Unkenstein, Bryant, Judd, Ong and Kinsella2016). It is also unclear whether menopausal brain fog represents an objective change in cognition that can be reliably detected by neuropsychological tests (Weber & Mapstone, Reference Weber and Mapstone2009). Studies that have examined this question often find that other menopausal symptoms, such as fatigue, mood disturbances, and vasomotor symptoms, may better predict subjective cognitive decline than measures of objective cognition, implying that reducing or treating these symptoms could improve cognitive functioning (Triantafyllou et al., Reference Triantafyllou, Armeni, Christidi, Rizos, Kaparos, Palaiologou, Augoulea, Alexandrou, Zalonis, Tzivgoulis and Lambrinoudaki2016; Unkenstein et al., Reference Unkenstein, Bryant, Judd, Ong and Kinsella2016; Weber et al., Reference Weber, Mapstone, Staskiewicz and Maki2012). However, there is also contrasting evidence pointing to a subtle decline in objective measures of verbal memory and attention (Armeni et al., Reference Armeni, Apostolakis, Christidi, Rizos, Kaparos, Panoulis, Augoulea, Alexandrou, Karopoulou, Zalonis, Triantafyllou and Lambrinoudaki2018; Drogos et al., Reference Drogos, Rubin, Geller, Banuvar, Shulman and Maki2013; Greendale et al., Reference Greendale, Wight, Huang, Avis, Gold, Joffe, Seeman, Vuge and Karlamangla2010; Grummisch et al., Reference Grummisch, Sykes Tottenham and Gordon2023; Schaafsma et al., Reference Schaafsma, Homewood and Taylor2010; Weber & Mapstone, Reference Weber and Mapstone2009; Weber et al., Reference Weber, Rubin and Maki2013), as well as a potential association between subjective cognitive complaints and decreased hippocampal volume (Conley et al., Reference Conley, Albert, Boyd, Kim, Shokouhi, McDonald, Saykin, Dumas and Newhouse2020). Furthermore, some research has suggested that cognitive changes can occur in the absence of other menopausal symptoms (Maki et al., Reference Maki, Springer, Anastos, Gustafson, Weber, Vance, Dykxhoorn, Milam, Adimora, Kassaye, Waldrop and Rubin2021).
It is crucial to characterize the typical neuropsychological profile of menopausal brain fog in relation to established cognitive constructs to facilitate an evidence-based assessment of its symptoms and severity. As a multifactorial syndrome, clinical presentations of menopausal brain fog are likely heterogeneous, meaning that two women reporting brain fog may be experiencing different types of symptoms, with the degree of diminishment in each construct varying from person to person (Grewal et al., Reference Grewal, Weinman, Hebron, Brown and Shackelford2023; Maki & Jaff, Reference Maki and Jaff2022).
Even though brain fog or cognitive symptoms during menopause may not reach established thresholds for statistical or clinical significance using neuropsychological tests, it is important to characterize the neuropsychological profile of brain fog for several reasons. First, up to 67% of women report a substantial impact of brain fog on work and quality of life (Harper et al., Reference Harper, Phillips, Biswakarma, Yasmin, Saridogan, Radhakrishnan, C Davies and Talaulikar2022). Hence, it is crucial that we better understand how it relates to objective cognitive performance, measurable by neuropsychological tests. Second, there is a need to untangle the individual factors that contribute to subjective cognitive decline during menopause to identify effective interventions to alleviate symptoms tailored to individual needs. Third, given that subjective cognitive concerns in adulthood are a risk factor for later-life dementia (Pike et al., Reference Pike, Cavuoto, Li, Wright and Kinsella2022), it is important to better understand the nature of the subjective cognitive symptoms or brain fog in menopause and characterize the associated neuropsychological profile. Finally, understanding the neuropsychological profile of brain fog is essential for research focused on understanding the direct effects of menopause-related changes on cognition.
In this study, we used objective cognitive performance, assessed using neuropsychological tests, as a proxy for brain health and aimed to determine the extent to which existing measures of subjective cognition correlate with neuropsychological outcomes. By synthesizing the current literature, we sought to identify the subjective measures currently used in menopausal populations and assess whether any patterns emerge in their correlation with objective cognitive performance. Accordingly, this study aimed to review the available measures for assessing brain fog, evaluate their quality, and examine how they relate to neuropsychological test performance. Specifically, this study aimed to systematically review and meta-analyze correlations between subjective and objective cognitive measures in perimenopausal and postmenopausal women.
Method
Preregistration and protocol
This study was preregistered on PROSPERO (Study ID: CRD42024541330). All human data included in this manuscript was obtained in compliance with the Helsinki Declaration.
Eligibility criteria
Eligible studies met the following criteria: (a) reported outcomes for at least one subjective measure of cognition and at least one neuropsychological test score in a perimenopausal or postmenopausal cohort of women, and (b) were in English, or an English version of the study could be accessed via Google Translate. Exclusion criteria included (a) studies involving participants with current comorbid neurological or psychiatric conditions known to impact cognition, (b) participants with a history of chemotherapy, and (c) studies not published in English. Specifically, we excluded studies that focused exclusively on clinical populations (e.g., major psychiatric, neurodevelopmental, or neurodegenerative disorders), but retained community-based samples that screened for these conditions. Studies that included individuals who had undergone surgical menopause as part of a larger sample were retained but noted for later sensitivity analyses if they were included in the meta-analysis. However, studies that exclusively examined cohorts of women with surgical menopause were excluded, as their findings may not generalize to the natural menopause transition, which was the primary focus of the review.
Search strategy
Search databases included MEDLINE, EMBASE, and PsycINFO, accessed via the Ovid interface. Three terms were used in the search strategy, including (a) neuropsychological assessment or cognition, (b) menopause, and (c) subjective cognitive decline. The full search strategy is available in Supplementary Table 1. The reference lists of eligible studies identified were also searched to identify any additional studies meeting the eligibility criteria. The final search date was November 22nd 2024.
Study selection
Search records were imported into Covidence screening (Veritas Health Innovation, 2024). Title and abstract screening was completed independently by the first author (R.F.). A total of 155 studies were retained for full-text screening and were independently reviewed by the first author (R.F.). The last author (C.G.) reviewed a subset of 55 studies. There was substantial agreement for inclusion between reviewers (agreement = 94.5%, Cohen’s kappa = .77), and disagreements were resolved by consensus.
Data extraction
Data extraction was completed independently by the first author (R.F.), with the last author (C.G.) reviewing accuracy in a subset of studies. No automation tools were used in the data extraction process. For longitudinal studies or randomized controlled trials (RCTs), the baseline estimates were recorded.
Study characteristics
Data items extracted for study characteristics included author information, publication year, the primary objective of the study, and the country where the study was conducted. The dates and setting of data collection were also recorded.
Participant characteristics
For participant characteristics, extracted data items included the sample size, mean age and education, and menopause stage for each group (classified as perimenopause, postmenopause, or combined). The criteria used to define the stage of menopause were also recorded. Any reported estimates of hormone levels, including estradiol, progesterone, testosterone, follicle-stimulating hormone (FSH), luteinizing hormone (LH), and use of menopausal hormone therapy, were also recorded for each group.
Measures of subjective cognition
The measure of subjective cognition used in each study was recorded. The number of items, scoring scale, test–retest reliability, and validation method were also recorded for each measure. If this information was not reported in the included study, it was sought from the original publication of the measure.
Measures of objective cognition
The names and outcomes of neuropsychological tests administered in each study were recorded. Following data extraction, tests were categorized into broad cognitive domains following the Cattell-Horn-Carroll (CHC) model of cognitive abilities (Schneider & McGrew, Reference Schneider, McGrew, Flanagan and McDonough2018). This model was selected because it is the most comprehensive and empirically supported model of cognition and is the basis for the most updated neuropsychological test batteries such as the Fifth Edition of the Wechsler Adult Intelligence Scale (WAIS-5; Wechsler et al., Reference Wechsler, Raiford and Presnell2024). These cognitive domains included learning efficiency (Gl; the ability learn and store information over time), working memory (Gwm; the ability to maintain and manipulate information in the mind), retrieval fluency (Gr; the efficiency with which information can be retrieved from long-term memory), processing speed (Gs; the ability to control attention to perform simple cognitive tasks quickly and fluently), fluid reasoning (Gf; the ability to solve novel problems without the use of previously learned information), and acquired knowledge (Gc; the ability to understand and communicate culturally relevant knowledge) (Schneider & McGrew, Reference Schneider, McGrew, Flanagan and McDonough2018). Additionally, general intelligence (G), or general cognitive ability, is a higher-order domain that underpins performance across all broad cognitive ability domains and reflect shared variance across distinct cognitive abilities. For each test administered, the mean score for each group reported in each study was recorded in the format used, including mean raw scores, z-scores, or t-scores.
Measures of other menopause symptoms
Any administered measures of mood, sleep or fatigue, and vasomotor symptoms were documented, with the names and outcomes of each recorded. Mood measures included those assessing overall mood, depressive symptoms, anxiety symptoms, or stress symptoms. These were grouped under a single category of mood.
Correlations between measures
Any reported Pearson’s correlation coefficients (Pearson’s r) or beta regression coefficients (beta estimates) between measures of objective and subjective cognition, subjective cognition and menopausal symptoms, and objective cognition and menopausal symptoms were recorded. Beta estimates are considered appropriate substitutions for Pearsons’s r, with evidence supporting the relative accuracy and stability of results in meta-analyses using both types of estimates (Peterson & Brown, Reference Peterson and Brown2005).
Quality evaluation
The risk of bias and applicability concern for each study was evaluated using a modified QUADAS-2 form (Whiting, Reference Whiting2011). The risk of bias was evaluated for four domains, including patient selection, the measure of subjective cognition, the measure of objective cognition, and timing and flow. Applicability concerns were evaluated for three domains, including patient selection, the measure of subjective cognition, and the measure of objective cognition. Signalling questions for each domain were answered as yes, unclear, or no, and the risk of bias for each domain was then determined as low, high, or unclear. A list of signalling questions and completed ratings for each included study are shown in Supplementary Table 2. The quality evaluation was completed independently by the first (R.F.) and last (C.G.) authors, with agreement between reviewers assessed using Cohen’s kappa (κ). There was a substantial agreement across domains, including patient selection (agreement = 95.5%, κ = .65), objective measure (agreement = 1, κ = 1), subjective measure (agreement = 1, κ = 1), and timing and flow (agreement = 91.1%, κ = 82.1). Disagreements were resolved by consensus.
Data synthesis
Studies that reported correlations between any measures were included in the meta-analyses. Meta-analyses were completed when three or more relevant outcomes were reported (Higgins et al., Reference Higgins, Thomas, Chandler, Cumpston, Li and Page2023). Three meta-analyses were planned a priori, including synthesizing correlations between objective and subjective measures of cognition (Model 1), subjective cognition and menopausal symptoms (Model 2), and objective cognition and menopausal symptoms (Model 2). All analyses were multilevel random-effect meta-analyses conducted using the meta package in R Studio (RStudio: Integrated Development for R, 2020; Schwarzer, Reference Schwarzer2024). The multilevel approach was used to account for dependent effect sizes, where level one represents the study level and level two represents the effect level, with results to be displayed visually using forest plots.
Effect measure
For all analyses, the effect measures were Pearson’s correlation coefficients (r). Heterogeneity was evaluated using Cochran’s Q with its associated degrees of freedom, tau-squared, and I-squared values (Higgins & Thompson, Reference Higgins and Thompson2002). Pearson’s correlation coefficients were interpreted according to Cohen (Reference Cohen1988) as small (r = .10), moderate (r = .30), or large (r = .50).
Subgroup and meta-regression analyses
Subgroup analyses planned a priori included examining the effect of the menopause stage (all models), the cognitive domain of the objective tests (Models 1 and 3), and menopausal symptoms type (Models 2 and 3). Meta-regressions were used to examine sources of within- and between-study heterogeneity on an exploratory basis.
Sensitivity analyses
Two sensitivity analyses were planned. The first was to evaluate the effect measure (Pearson’s r or beta estimate) on the results of any meta-analysis. The second was to examine the effect of studies rated as having a high risk of bias in any QUADAS-2 domain.
Results
Study characteristics
A flowchart of the study selection process is shown in Figure 1. Twenty-two eligible studies were included in the systematic review, including 16 cross-sectional studies, six RCTs, and two longitudinal studies. A summary of the study characteristics is displayed in Table 1.
Figure 1. PRISMA flowchart of study selection.
Table 1. Characteristics of included studies
Sample characteristics
Overall, there were 5007 participants, including 512 perimenopausal women, 4352 postmenopausal women, and 143 women across combined samples. The mean age for the perimenopausal group was 49.50 years (SD = 0.98), while the postmenopausal group had a mean age of 55.87 years (SD = 4.25). The combined group had a mean age of 53.46 years (SD = 0.49). The perimenopausal group had an average of 16 years of education (SD = 0.30), compared to 13.19 years (SD = 2.58) for the postmenopausal group and 15.75 years (SD = 0.92) for the combined group. Only eight studies reported estradiol, testosterone, follicle-stimulating hormone (FSH), or luteinizing hormone (LH) levels. No studies reported progesterone. A summary of these details, including studies with missing data, is available in Supplementary Table 3.
Quality evaluation
The assessment of risk of bias across studies was conducted across four domains, with ratings of low, high, or unclear risk for each. Individual study scores for each signalling question are available in Supplementary Table 2. In the patient selection domain, 9% of studies were rated as having a low risk of bias, 54.5% as high risk, and 36.5% as unclear. Those rated as having high risk typically used convenience sampling to recruit participants rather than random or consecutive recruitment. In contrast, those with unclear risk typically did not report how participants were recruited. For the objective measure domain, 4.5% of studies were classified as low risk, 22.7% as high risk, and 72.7% as unclear. Bias in this domain was assessed based on whether objective cognition measures were interpreted without knowledge of the results of the subjective measure of cognition. The high number of studies rated as unclear in this domain was due to a lack of available information in the report. Similarly, in the subjective measure domain, 13.6% of studies demonstrated low risk, 22.7% high risk, and 63.6% unclear risk of bias. Again, the high number of studies rated as unclear was due to insufficient information about whether the subjective test was scored and interpreted without knowledge of objective test results. Lastly, the timing and flow domain included 45.5% of studies rated as low risk, 13.6% as high risk, and 40.9% as unclear. This domain assessed bias regarding whether there were any gaps between the administration of the objective tests and subjective measures of cognition (ideally, these measures should be administered during the same session). A summary of the risk of bias across included studies is displayed in Figure 2.
Figure 2. Risk of bias for patient selection, objective measures, subjective measures, and timing and flow using a modified QUADAS-2 form.
Applicability concern was assessed for patient selection, measures of objective cognition, and measures of subjective cognition. For patient selection, 45.5% of studies were rated as having low applicability concern, 22.7% as high concern, and 31.8% as some concerns. In the objective measures domain, 86.4% of studies had low applicability concerns, 4.5% were rated as high concerns, and 9.1% had some concerns. The subjective measures domain showed 90.9% of studies with low applicability concern and 9.1% with some concerns. A summary of applicability across included studies is displayed in Figure 3.
Figure 3. Applicability concern for patient selection, objective measures, and subjective measures using a modified QUADAS-2 form.
Measures of subjective cognition
Twelve different measures of subjective cognition were used across the included studies. As expected, these measures varied considerably regarding psychometric quality and the aspect of subjective cognitive functioning each questionnaire measures. A summary of all measures, including scale details and test–retest reliability estimates, is shown in Table 2. For organizational purposes, the measures were grouped into broad themes based on face validity and apparent alignment with cognitive symptoms commonly reported by experiencing menopausal brain fog including general cognitive concerns, attention and concentration, and memory (Harper et al., Reference Harper, Phillips, Biswakarma, Yasmin, Saridogan, Radhakrishnan, C Davies and Talaulikar2022). This categorization was intended solely to aid interpretation and does not imply theoretical distinctions. Currently, there is limited construct validity evidence and no established factor structure to support the assumption that these self-report measures map onto distinct cognitive constructs in a manner consistent with objective cognitive domains, such as those defined in CHC theory. Specifically, confirmatory factor analysis (CFA) is considered the gold standard approach for evaluating construct validity, as it tests whether the data fit a hypothesized measurement model based on theoretical expectations (Strauss & Smith, Reference Strauss and Smith2009). The absence of such analyses during scale development limits confidence in the distinctiveness and theoretical validity of the self-report cognitive domains purportedly assessed. Without a measurement model, it is difficult to determine whether total scores and subscale scores meaningfully reflect discrete constructs, which in turn undermines attempts to evaluate convergent and discriminant validity, both between self-report measures and in relation to objective cognitive performance.
Table 2. Characteristics of measures of subjective cognition across included studies
Note: Pearson’s r test–retest coefficients and ICC values between below .75 are considered poor, .75 to .90 are considered moderate, and above .90 are considered high (Nunnally & Bernstein, Reference Nunnally and Bernstein1994; Portney & Watkins, Reference Portney and Watkins2009).
*Test–retest reliability estimate is for the overall scale.
General cognition
Three measures conceptualized subjective cognition in terms of task failures, complaints, or perceived cognitive decline. Three studies used the Cognitive Complaint Index (CCI; Conley et al., Reference Conley, Albert, Boyd, Kim, Shokouhi, McDonald, Saykin, Dumas and Newhouse2020; Dumas et al., Reference Dumas, Kutz, McDonald, Naylor, Pfaff, Saykin and Newhouse2013; Saykin et al., Reference Saykin, Wishart, Rabin, Santulli, Flashman, West, McHugh and Mamourian2006; Vega et al., Reference Vega, Zurkovsky, Albert, Melo, Boyd, Dumas, Woodward, McDonald, Saykin, Park, Naylor and Newhouse2016). The total score provides an “index” reflecting general cognitive complaints across various areas of functioning and is a compilation of 114 items from different existing measures (see Saykin et al., Reference Saykin, Wishart, Rabin, Santulli, Flashman, West, McHugh and Mamourian2006). Although the total score is estimated to have high test–retest reliability (Rattanabannakit et al., Reference Rattanabannakit, Risacher, Gao, Lane, Brown, McDonald, Unverzagt, Apostolova, Saykin and Farlow2016), this index has not been validated as a measure of subjective cognition.
The Cognitive Failures Questionnaire (CFQ), a 25-item measure of self-reported “failures” across everyday tasks such as absent-mindedness, forgetfulness, and clumsiness, was used in one study (Broadbent et al., Reference Broadbent, Cooper, FitzGerald and Parkes1982; Jenkins et al., Reference Jenkins, Ambroisine, Atkins, Cuzick, Howell and Fallowfield2008). The total score has been validated in a university sample and has moderate test–retest reliability but has not been validated in a menopausal sample (Craig Wallace, Reference Craig Wallace2004; Rast et al., Reference Rast, Zimprich, Van Boxtel and Jolles2009).
The Subjective Cognitive Decline Questionnaire (SCDQ), a 24-item measure of self-perceived cognitive decline, was used in one study (Pang & Kim, Reference Pang and Kim2021; Rami et al., Reference Rami, Mollica, García-Sanchez, Saldaña, Sanchez, Sala, Valls-Pedret, Castellví, Olives and Molinuevo2014). The scale includes questions about perceived decline in several cognitive domains, including memory, language, and executive function. The measure has moderate test–rest reliability and has been validated in a sample of individuals with mild cognitive impairment and Alzheimer’s disease (Rami et al., Reference Rami, Mollica, García-Sanchez, Saldaña, Sanchez, Sala, Valls-Pedret, Castellví, Olives and Molinuevo2014). However, dimensionality, or construct validity, was assessed using principal components analysis (PCA), which is a method of item reduction rather than factor analysis (Floyd & Widaman, Reference Floyd and Widaman1995). As such, construct validity was not evaluated according to established guidelines (Strauss & Smith, Reference Strauss and Smith2009).
Attention and concentration
Two measures focused on attention, including the 13-item Attentional Functional Index (AFI) (Cimprich et al., Reference Cimprich, Visovatti and Ronis2011), which was used in one study, and the 5-item attention and concentration subscale of the Brown Attention Deficit Disorder Scale (BADDS), which was used in two studies (Epperson et al., Reference Epperson, Pittman, Czarkowski, Bradley, Quinlan and Brown2011, Reference Epperson, Shanmugan, Kim, Mathews, Czarkowski, Bradley, Appleby, Iannelli, Sammel and Brown2015). The AFI has high test–retest reliability, but construct validity was evaluated using PCA (Cimprich et al., Reference Cimprich, Visovatti and Ronis2011). The Brown Attention-Deficit Disorder Scales are also reported to have adequate reliability and have been validated in an adult sample (Brown, Reference Brown1996; Davenport & Davis, Reference Davenport, Davis, Goldstein and Naglieri2011).
Memory functioning
Six scales were used to assess memory functioning, each with varying reliability and validity evidence. The Everyday Memory Questionnaire-Revised (EMQ-R), a 13-item scale measuring attentional and retrieval difficulties, was used in one study (Royle & Lincoln, Reference Royle and Lincoln2008; Zhu et al., 2024). While the original scale demonstrated moderate test–retest reliability, this has not been evaluated in the revised scale (Efklides et al., Reference Efklides, Yiultsi, Kangellidou, Kounti, Dina and Tsolaki2002). Moreover, the construct validity of the EMQ-R is questionable, again assessed using PCA, which identified a third factor (or component) with two unrelated items that were retained in the scale (Royle & Lincoln, Reference Royle and Lincoln2008).
Similarly, the Memory and Cognitive Confidence Scale (MACCS) is a 28-item scale that assesses various aspects of memory confidence, including general memory, decision-making, concentration ability, and cognitive perfectionism (Ballantyne et al., Reference Ballantyne, King and Green2021; Nedeljkovic & Kyrios, Reference Nedeljkovic and Kyrios2007). Test–rest reliability varies across subscales, with high reliability for the total score and decision-making scale but moderate to poor for the other subscales (Nedeljkovic & Kyrios, Reference Nedeljkovic and Kyrios2007). Validation efforts have focused on individuals with obsessive–compulsive disorder (Nedeljkovic & Kyrios, Reference Nedeljkovic and Kyrios2007).
The Memory Functioning Questionnaire (MFQ) includes four subscales, namely, Frequency of Forgetting (32 items), Seriousness of Forgetting (18 items), Mnemonics Usage (8 items), and Retrospective Forgetting (5 items; Gilewski et al., Reference Gilewski, Zelinski and Schaie1990). No total score is calculated. Two studies used all four subscales (Maki et al., Reference Maki, Gast, Vieweg, Burriss and Yaffe2007; Weber et al., Reference Weber, Mapstone, Staskiewicz and Maki2012), and one study focused solely on the Frequency of Forgetting (Grummisch et al., Reference Grummisch, Sykes Tottenham and Gordon2023). Evidence supporting the reliability and validity of these subscales is limited, with initial exploratory factor analysis providing some support for its stability (Gilewski et al., Reference Gilewski, Zelinski and Schaie1990).
The Multifactorial Memory Questionnaire (MMQ) has three subscales, including Memory Contentment (i.e., satisfaction with memory; 21 items), Memory Ability (i.e., perceived memory ability; 20 items), and Strategy Use (i.e., use of techniques to support memory; 20 items) (Troyer & Rich, Reference Troyer and Rich2002). No total score is calculated. All three subscales were used in one study (Unkenstein et al., Reference Unkenstein, Bryant, Judd, Ong and Kinsella2016). There is evidence of moderate test–retest reliability for the strategy use subscale but poor reliability for the other subscales (Yang et al., Reference Yang, Chou, Lee, Lin and Chiang2023). Similar to the EMQ-R and the MFQ, the subscales were identified using PCA (Troyer & Rich, Reference Troyer and Rich2002).
Other scales include the Subjective Memory Questionnaire (SMQ), a 43-item measure used in one study (Gorenstein et al., Reference Gorenstein, Rennó, Filho, Gianfaldoni, Gonçalves, Halbe, Fernandes and Demétrio2011), with moderate test–retest reliability but limited evidence for construct validity (Bennett-Levy & Powell, Reference Bennett-Levy and Powell1980). The Memory and Concentration subscale of the Women’s Health Questionnaire (WHQ) consists of three items and was used in one study (Hunter, Reference Hunter1992; Karossy et al., Reference Karossy, Kerekes, Horvath, Goocze and Kallai2007). The scale was developed and validated in a sample of mid-life women aged 45 to 65, with reliability estimates ranging from moderate to high (Hunter, Reference Hunter2003).
Finally, the Daily Symptom Rating Scale (DSRS) is a 17-item measure that assesses mood, behavior, and pain symptoms related to menopause but also includes single items on forgetfulness and concentration (Taylor, Reference Taylor1979). This measure was used in one study (Barnhart et al., Reference Barnhart, Freeman, Grisso, Rader, Sammel, Kapoor and Nestler1999).
Single-item measures
In addition, six studies assessed subjective cognition using a single question. For four of those studies, the question was rated on a Likert scale. For the other two studies, the question was rated on a binary scale, as “yes” or “no” (Drogos et al., Reference Drogos, Rubin, Geller, Banuvar, Shulman and Maki2013; Hogervorst et al.,Reference Hogervorst, Boshuisen, Riedel, Willeken and Jolles1999; Li et al., Reference Li, Hao, Fu, Zhou and Zhu2022; Schaafsma et al., Reference Schaafsma, Homewood and Taylor2010; Triantafyllou et al., Reference Triantafyllou, Armeni, Christidi, Rizos, Kaparos, Palaiologou, Augoulea, Alexandrou, Zalonis, Tzivgoulis and Lambrinoudaki2016). Further details are provided in Table 2.
Measures of objective cognition
A broad range of neuropsychological tests were used to measure objective cognition across included studies. The definition of each cognitive domain and a breakdown of the neuropsychological tests included in each domain are shown in Table 3. These tests were categorized into broad cognitive domains based on the CHC model. The cognitive domains assessed included learning efficiency, working memory, retrieval fluency, cognitive speed, fluid reasoning, and acquired knowledge. The most common cognitive domain tested was learning efficiency (21 studies), followed by working memory capacity (17 studies), processing speed (12 studies), and retrieval fluency (10 studies). Visual processing was measured in only four studies, comprehension knowledge was measured in two studies, and only one study measured fluid reasoning ability.
Table 3. Characteristics of measures of objective cognition classified using Cattell-Horn-Carroll (CHC) theory
Measures of other menopause symptoms
Other menopausal symptoms assessed across studies included sleep or fatigue, overall mood, depression, anxiety, stress, and vasomotor symptoms. Supplementary Table 4 summarizes the measures used to assess these symptoms.
Results of the meta-analyses
Studies that reported Pearson’s correlation or beta coefficients between measures were retained for meta-analysis. Three main meta-analyses were conducted to pool correlations, including between subjective and objective cognition (Model 1), subjective cognition and other menopausal symptoms (Model 2), and objective cognition and other menopausal symptoms (Model 3).
Model 1: subjective versus objective cognition
Higher levels of subjective cognitive complaints were expected to be associated with poorer performance on objective measures. The correlation sign for subjective measures where higher scores represent better performance was inverted to maintain consistency. As such, negative correlations reflect that higher subjective scores (representing worse subjective cognition) are associated with lower objective scores (representing worse test performance).
A forest plot of the overall model organized by objective cognitive domain is shown in Figure 4. Six studies reported a total of 24 correlations between subjective and objective cognition. A multilevel random effects meta-analysis revealed that the overall correlation was non-significant (r = .06; CI = −.06 to .19; t = 1.05; p = .303). Heterogeneity was non-significant, including for the overall model (Q = 23.74, df = 23, p = .418), within-studies (τ2 < 0.001; CI < 0.01 to 0.01), and between-studies (τ2 = 0.01; CI < 0.01 to 0.14) was minimal (I 2 = 3.1%), suggesting the effect was relatively consistent across studies. However, a subgroup analysis looking at the correlation between subjective cognition and different objective cognitive domains, as this analysis was planned a priori.
Figure 4. Forest plot of correlations between subjective and objective measures of cognition (model 1) categorized by cognitive domain.
There were 12 measures of learning efficiency and nine measures of working memory capacity. In addition, there were three measures of general cognition. A small significant correlation was observed between cognition and learning efficiency (r = .12; CI = .02 to .23). The correlation between subjective cognition and working memory capacity was non-significant (r = .05; CI = −.17 to .26), as was the correlation between subjective cognition and general cognitive ability (r = .05; CI = −.24 to .33). Heterogeneity statistics for each subgroup analysis are displayed in Table 4.
Table 4. Heterogeneity statistics for subgroup analyses for all meta-analytic models
Model 2: subjective cognition versus menopause symptoms
Higher levels of subjective cognitive complaints were expected to be associated with more severe menopausal symptoms. The correlation sign for subjective measures where higher scores represent better performance was inverted to maintain consistency. As such, positive correlations reflect that higher subjective scores (representing worse cognition) are associated with higher scores on measures of menopausal symptoms (representing worse symptoms).
Six studies reported a total of 26 correlations between subjective cognition and menopause symptoms. Positive correlations indicated that higher reports of subjective cognitive decline were associated with more menopausal symptoms. A multilevel random effects meta-analysis revealed that the overall correlation was non-significant (r = .11; CI = -.27 to .47; t = 0.60; p = .551). There was significant heterogeneity in the overall model (Q = 688.93, df = 25, p < .001), as well as significant within-study heterogeneity (τ2 = 0.38; CI = 0.21 to 0.77) and between-study variability (τ2 = 0.10; CI = 0.00 to 0.93; I 2 = 96.4%).
A subgroup analysis was conducted to investigate whether the type of menopause symptom was a significant source of heterogeneity. The correlation was non-significant across all symptom subgroups, including overall mood (r = −.19; CI = −.46 to .12), sleep or fatigue symptoms (r = .01; CI = −.57 to .59), and vasomotor symptoms (r = .63; CI = −.43 to .96), suggesting that heterogeneity in the overall model was not due to symptom type. Similarly, a subgroup analysis was conducted to investigate whether the menopause stage was a significant source of heterogeneity. The correlation was also non-significant across all subgroups, including perimenopause (r = .15; CI = −.56 to 0.74), post-menopause (r = .27; CI = −.46 to .78), and combined samples (r = −.24; CI = −.88 to .70). Heterogeneity statistics for both subgroup analyses are displayed in Table 4. A forest plot of the overall model organized by symptom type is shown in Supplementary Figure 1.
Model 3: objective cognition versus menopause symptoms
Higher levels of menopausal symptoms were expected to be associated with poorer performance on objective measures. As such, negative correlations reflect that higher scores on measures of menopausal symptoms (representing worse cognition) are associated with lower objective scores (representing worse performance).
Two studies reported a total of 60 correlations between objective cognition and menopause symptoms. Positive correlations indicated that poorer performance on objective measures was associated with more menopausal symptoms. A multilevel random effects meta-analysis revealed the overall correlation was non-significant (r = −.11; CI = −.29 to .08; t = −1.12; p = .265), with significant within-study heterogeneity (τ2 = 0.02; CI = 0.01 to 0.05) and some between-study variability (τ2 = 0.08; CI = 0.00 to 0.60; I 2 = 51.8%). Overall heterogeneity in the model was significant (Q = 122.32, df = 59, p < .001). As such, subgroup analyses were subsequently conducted to investigate potential sources of heterogeneity.
Subgroup analyses were conducted to examine whether the cognitive domain was a significant source of heterogeneity in the model. There was a significant negative correlation between menopause symptoms and tests of general intellectual ability (r = −0.25; CI = −0.43 to −0.05) and learning efficiency (r = −.19; CI = −.26 to −.12). In contrast, the correlation was non-significant for tests of visuospatial ability (r = −.04; CI = −.25 to .16) and working memory capacity (r = .13; CI = −.06 to .32, respectively). A forest plot of the overall model organized by objective cognitive domain is shown in Supplementary Figure 2. Heterogeneity statistics for both subgroup analyses are displayed in Table 4.
Sensitivity analyses
Sensitivity analyses were conducted to examine whether the difference in effect size measures was a significant source of heterogeneity. For Model 1, the effect size measure was a significant source of heterogeneity in the overall model (F 1, 22 = 8.64, p = .008). Upon further inspection in a subgroup analysis, beta effect sizes appeared to be driving the effect. Therefore, we reran the cognitive domain subgroup analysis without the beta effects (n = 1, k = 5). The resulting model showed non-significant correlations across all domains, including learning efficiency. Whilst this finding indicates that the results of this subgroup analysis should be interpreted with caution, considering that only one study used beta estimates, the results of this sensitivity analysis may be due to other study-specific effects.
To be comprehensive, sensitivity analyses were also conducted for Models 2 and 3. The results indicated that the effect size measure was not a significant source of heterogeneity in Model 2 (F 1, 33 = 0.57, p = .457) or Model 3 (F 1, 58 = 1.54, p = .219), suggesting that the inclusion of beta coefficients did not have a significant impact on the results.
Risk of bias analyses
Risk of bias analyses were conducted to assess whether studies rated as having a high risk of bias in any domain significantly impacted the results of Model 1. The results showed that variance in the overall model was not related to the risk of bias in the patient selection domain (F 1, 22 = 1.12, p = .301), the objective measure domain (F 1, 22 = 0.01, p = .920), or the timing and flow domain (F 1, 22 = 1.05, p = .316). However, variance was significantly associated with the risk of bias in the subjective measure domain (F 1, 22 = 8.64, p = .008). One of the signalling questions, which are prompts or items on quality evaluation tools to guide evaluation of the risk of bias, in this domain was whether subjective cognition was assessed using a single question. Two studies using single-item measures recruited participants based on their reported subjective cognitive decline, which may have contributed to the risk of bias in this domain.
Discussion
The current study aimed to systematically review and meta-analyze correlations between subjective and objective measures of cognition in perimenopausal and postmenopausal women. Previous research has suggested that cognitive complaints during menopause are associated with subtle declines in verbal memory, attention, and processing speed (Armeni et al., Reference Armeni, Apostolakis, Christidi, Rizos, Kaparos, Panoulis, Augoulea, Alexandrou, Karopoulou, Zalonis, Triantafyllou and Lambrinoudaki2018; Drogos et al., Reference Drogos, Rubin, Geller, Banuvar, Shulman and Maki2013; Greendale et al., Reference Greendale, Wight, Huang, Avis, Gold, Joffe, Seeman, Vuge and Karlamangla2010; Grummisch et al., Reference Grummisch, Sykes Tottenham and Gordon2023; Schaafsma et al., Reference Schaafsma, Homewood and Taylor2010; Weber & Mapstone, Reference Weber and Mapstone2009; Weber et al., Reference Weber, Rubin and Maki2013). The results of the meta-analyses revealed a small but significant correlation between subjective cognition and performance on measures of learning efficiency. Correlations between subjective cognition and all other cognitive domains were non-significant. Additionally, although subjective cognition and objective cognitive performance were expected to correlate with other menopausal symptoms, the meta-analyses did not support this prediction.
Measures of subjective cognition used in menopause research
A key finding of this study was the considerable heterogeneity in the measures of subjective cognition used across studies. To highlight current limitations in existing measures, subjective cognition was quantified in 21 different ways across the 22 included studies, including self-report questionnaires and subscales. Whilst the included studies all purported to assess “subjective cognition,” the measurement tools used in these studies varied considerably in the conceptualization of subjective cognition. For example, whilst most questionnaires provided scores aiming to quantify memory functioning in some way, this included different aspects of memory functioning, such as perceptions of the frequency of forgetting, confidence in memory ability, memory strategy use or seriousness of forgetting. Other questionnaires or scores aimed to capture related abilities, such as attention or concentration, while others aimed to capture cognition more broadly, such as those with scores reflecting cognitive complaints or cognitive failures.
Additionally, whilst most scales were validated as part of the development procedure, the method used by some researchers was PCA, which is appropriate for item reduction but is not a method of construct validation (Floyd & Widaman, Reference Floyd and Widaman1995). Only one scale, the CFQ, used a confirmatory factor analysis, which is considered the gold standard approach for evaluating construct validity (Strauss & Smith, Reference Strauss and Smith2009). Measures of subjective cognition are intended to capture perceptions of cognition in day-to-day life as well as functioning and, as such, should not be expected to map neatly onto objective cognitive domains. However, the lack of consistency raises the question of whether “subjective memory” in these studies is comparable. In terms of whether any existing measure may be appropriate for use as a scale for menopausal brain fog, the measures also varied considerably in terms of age of publication, with publication dates ranging from 1979 to 2014. Items conceptualized and validated several decades ago may be less relevant to capture functioning in the context of modern day social and occupational demands, particularly digital-based platforms, communication, and multi-tasking. Additionally, measures varied in their test–retest reliability (r = .72 to .96), ranging from poor to high reliability. While reliability estimates of .90 or above are considered acceptable for clinical decision-making, several measures fell below this threshold, suggesting limited clinical utility as potential measures of menopausal brain fog. As a result, findings from individual studies may only apply to their specific samples.
Overall, these findings emphasize the variability in the existing literature on subjective cognition in menopause, limiting the comparability of findings. In addition, the lack of validation of scales in a menopausal sample, except the EMQ-R (Zhu et al., 2024), means it remains unclear whether current research can capture the experience of “brain fog” or even a consistent concept of subjective cognition. It is uncertain to what extent these measures reflect women’s experiences of brain fog, especially as none of the studies in this review explicitly used this term. Therefore, the following findings regarding the correlations between subjective and objective cognition should be interpreted with these limitations in mind.
Subjective versus objective cognition in menopause
Despite the variability in subjective, a meta-analysis was conducted to assess whether the existing research findings provide evidence to support a correlation between objective and subjective cognition in menopause. Based on previous research suggesting that menopausal brain fog may be related to reduced attention, learning efficiency, and processing speed, it was anticipated that a small correlation would be observed between subjective and objective cognition, at least in these cognitive domains. The results of the meta-analysis were partially consistent with expectations, demonstrating a small correlation between subjective cognition and measures of learning efficiency. This finding suggests that subjective cognition during the menopause transition may be linked to a subtle decline in learning efficiency, which involves the ability to learn, store, and consolidate new information over time (Schneider & McGrew, Reference Schneider, McGrew, Flanagan and McDonough2018). Of note, the measures of subjective cognition used in the studies in this meta-analysis were the Cognitive Complaints Index (CCI), the Memory Functioning Questionnaire (MFQ), and a single-item measure of memory functioning measured on a 7-point scale (i.e., “How would you rate your memory in terms of the kinds of problems that you have?”). Thematically, these measures appear to capture broader perceptions of cognitive functioning rather than being specific to perceived memory decline. However, due to the limited construct validity evidence supporting the interpretation of these measures in terms of specific cognitive domains, this interpretation remains speculative and is a direction for future research.
Although related to memory function, it is crucial to distinguish learning efficiency as a cognitive ability distinct from the neural processes and brain areas typically associated with memory. As clinical neuropsychologists are well aware, performance on measures of learning efficiency can be influenced by numerous factors, including other cognitive abilities such as working memory, attentional capacity, and fluid reasoning (i.e., executive functioning), all of which impact how well information is learned and subsequently retrieved. Additionally, performance can be affected by non-cognitive factors, such as fatigue, effort, and distractions like pain or other physical symptoms. Thus, our findings suggest that subjective cognitive concerns are associated with poorer performance on learning efficiency tasks. However, this finding does not necessarily indicate a direct relationship between menopause and primary memory difficulties. Instead, the results imply that subjective cognitive measures aiming to capture the experience of menopausal brain fog in a reliable and meaningful way should include items not only related to memory function but also to attention, concentration, and “executive functions,” such as planning and organization. Including such items is important to ensure that subjective cognitive measures capture the full spectrum of cognitive changes experienced during menopause in order to be clinically useful and ecologically valid.
On the other hand, it is worth considering that some research findings do support an association between subjective cognitive decline during menopause and a primary memory impairment. Whilst subtle memory declines can be difficult to detect through neuropsychological tests alone due to limited sensitivity, neuroimaging studies indicate a potential link between subjective cognition and reduced grey matter volume in regions associated with memory functioning, such as the hippocampus (Conley et al., Reference Conley, Albert, Boyd, Kim, Shokouhi, McDonald, Saykin, Dumas and Newhouse2020). Other studies have reported associations between menopause stage and grey matter volume in various areas of the brain (for a review, see Ramli et al., Reference Ramli, Yahaya, Mohd Fahami, Abdul Manan, Singh and Damanhuri2023). However, findings are mixed. For instance, some studies report a decline in hippocampal volume during perimenopause that persists into postmenopause (Goto et al., Reference Goto, Abe, Miyati, Inano, Hayashi, Aoki, Mori, Kabasawa, Ino, Yano, Iida, Mima and Ohtomo2011; Mosconi et al., Reference Mosconi, Rahman, Diaz, Wu, Scheyer, Hristov, Vallabhajosula, Isaacson, de Leon, Brinton and Ginsberg2018), while others suggest that hippocampal volume decreases during perimenopause but recovers postmenopause (Mosconi et al., Reference Mosconi, Berti, Dyke, Schelbaum, Jett, Loughlin, Jang, Rahman, Hristov, Pahlajani, Andrews, Matthews, Etingin, Ganzer, de Leon, Isaacson and Brinton2021). Other studies have found minimal or no significant differences in grey matter volume across menopause stage groups (Seitz et al., Reference Seitz, Kubicki, Jacobs, Cherkerzian, Weiss, Papadimitriou, Mouradian, Buka, Goldstein and Makris2019; Sullivan et al., Reference Sullivan, Marsh and Pfefferbaum2005). These discrepancies may reflect methodological variability, such as differences in sample characteristics or menopause staging criteria. Nevertheless, the hippocampus remains a region of interest in understanding menopause-related cognitive changes and in identifying women who may be at increased risk for developing neurodegenerative disease. Given that subjective cognitive concerns are themselves a risk factor for later life dementia (Pike et al., Reference Pike, Cavuoto, Li, Wright and Kinsella2022), it is important to better understand the nature of these symptoms and to characterize their associated neuropsychological profile. Future research should prioritize the development of a clear conceptual framework and cognitive characterization of menopausal brain fog, in conjunction with the creation of a reliable self-report measure that aligns with established cognitive domains. Such work would allow self-reported symptoms to be mapped onto objective cognitive functioning, as measured using validated neuropsychological tests. Improving the alignment between theoretical models, subjective and objective cognitive measures, and neuroimaging findings is essential to clarify the nature of menopause-related cognitive changes and, ultimately, identify women at greatest risk for ongoing cognitive decline.
Limitations of the evidence
The results of the meta-analyses were limited by the small number of included studies that reported correlation matrices. Although most included studies reported outcomes for subjective cognition, objective cognition, and various other measures to assess symptoms of menopause, only six of 22 studies in total reported correlations between these measures. Evaluating correlations between measures was not the primary aim of many of these studies. However, both the American Educational Research Association (AERA, 2006) and the American Psychological Association (2020) emphasize the need for correlation matrices in published reports to facilitate secondary analysis (Zientek & Thompson, Reference Zientek and Thompson2009).
For instance, our study provides preliminary evidence supporting a correlation between subjective cognition and learning efficiency during menopause. Whilst this information may aid the development of a new scale, analyzing a broader array of correlations in the meta-analysis could enable an evaluation of which objective tests of learning efficiency may be more effective measures of menopausal brain fog. It has been suggested that associative memory, a narrow ability under learning efficiency describing the ability to form a link between previously unrelated stimuli, is closely related to hippocampal integrity (Saling, Reference Saling2009). However, due to the limited number of correlations available, an analysis of task-specific abilities was not possible in the current study. The point of raising this possibility is not to speculate on a causal relationship between menopause and cognition but to highlight the importance of researchers reporting correlation tables in their published manuscripts, even in supplementary material, to facilitate the synthesis of findings such as those presented in this review.
Limitations of the analyses
Due to the few studies included in the meta-analysis, some evaluations of heterogeneity in the results were confounded with sample effects. For instance, although there were 60 correlations included in the meta-analyses between objective cognition and menopausal symptoms, those correlations were derived from only two studies, limiting the precision of estimates of between-study heterogeneity. Similarly, sensitivity analyses were also limited, as the risk of bias across domains was assessed per study rather than effect size, meaning there was limited variability in the risk of bias rating in each meta-analysis. However, notably, the meta-analysis between objective and subjective measures included 24 effect sizes from six different studies, including 12 effect sizes across four different studies for the learning efficiency subgroup analysis. Whilst there is no strict rule for the minimum number of studies or effect sizes required for subgroup analyses, most sources indicate that a minimum of four or five studies is adequate for findings to be meaningful (Borenstein, Reference Borenstein2009; Higgins & Thompson, Reference Higgins and Thompson2002).
Implications and future directions
A key finding of this study is the inadequacy of current measures of subjective cognition during menopause. The variability in the measures used across existing research limits the comparability of findings in this area and hinders the ability to draw meaningful inferences about the experience of menopausal brain fog, particularly since none of the studies explicitly used this term with their participants. This finding highlights two main research directions for cognition during menopause.
The first direction is the need to characterize menopausal brain fog in relation to established cognitive theory. Determining how menopausal brain fog manifests across established cognitive domains (e.g., working memory, learning efficiency, processing speed, etc.) is essential for understanding how brain fog differentiates from other psychological or cognitive conditions. This characterization should be broad, aiming to capture the cognitive profile of women across all stages of the menopause transition. The focus should be on examining the covariance of changes across cognitive domains at different stages of menopause to identify specific, narrow cognitive abilities that tend to be affected. For instance, one avenue to explore is whether there is a difference between associative memory and meaningful memory and whether there is evidence of a differential demand for visual processing. Additionally, research in this area needs to account for the potential indirect effects of other menopausal symptoms, ensuring that selected measures are validated in menopausal samples. This characterization would provide the necessary conceptual foundation to guide item generation for a targeted measure of menopausal brain fog, ensuring that any new subjective tool is grounded in theory and interpretable alongside objective test performance. It would also support the identification of objective neuropsychological tests that are likely to be sensitive to menopause-related cognitive changes.
The second direction is to develop a new measure of subjective cognition designed to assess cognitive complaints experienced by perimenopausal women who describe their symptoms as “brain fog.” Currently, there is no validated measure of subjective cognition in this population. As awareness of the relationship between menopause and cognition grows, discussions of brain fog are becoming increasingly common in the community. Consequently, more women are presenting to general practitioners with complaints of “brain fog” due to menopause and are seeking treatment. Future research could be directed at developing validated scales to measure the degree of brain fog and provide a means of measuring how brain fog responds to different treatment options. In addition, clarity around the subjective and objective profile of menopause brain fog will aid discrimination between menopause-related cognitive concerns from early onset dementia as well as ADHD, which is increasingly being diagnosed in perimenopause. One final area of future research could be providing better clinical guidelines for primary care practitioners on how to assess brain fog.
Conclusion
Many women report experiencing some degree of cognitive symptoms during their menopause transition years. The severity appears to range from mild to problematic, with reports that cognitive symptoms impact work and life in general (Mitchell & Woods, Reference Mitchell and Woods2001). The present findings emphasize the need for a stronger characterization of menopausal brain fog in relation to both subjective cognitive symptoms and objective cognitive performance. Research in this area cannot progress in an empirically robust manner without a good theoretical understanding of what brain fog looks like in relation to cognition and without a valid and reliable tool to assess brain fog in individuals presenting with concerns. Ideally, these two research goals should be pursued in tandem through an iterative process. A robust characterization of brain fog is essential to guide the development of a new measure, and a robust measure itself is necessary to map objective cognitive function to various presentations and severities of subjective brain fog. Like all good scientific theories, advances in this area would self-correct through an iterative process in which a theory of brain fog is established, which guides the development and validation of a new measure. As that measure is used to collect data from larger and more diverse samples, the theory can be refined.
The results of the present study provide a lead for further research on the assessment of menopausal brain fog, suggesting that future work may begin by focusing on the factors impacting learning efficiency during menopause, such as attention or concentration, and non-cognitive factors, such as anxiety and distractions. We recommend that future research prioritize using neuropsychological tests that capture narrow abilities within the broad domain of learning efficiency, including measures of both associative memory and meaningful memory (Schneider & McGrew, Reference Schneider, McGrew, Flanagan and McDonough2018). Suitable measures of these domains include the Verbal Paired Associates (VPA) and Logical Memory subtests from the Wechsler Memory Scale-Fourth Edition (WMS-IV). Tasks from the WMS-IV or the Wechsler Adult Intelligence Scales are grounded in cognitive theory and offer high ceilings, which is key for detecting subtle cognitive changes that may occur during the menopause transition. As such, long-format tasks with robust psychometric properties are recommended. For instance, the CVLT-3 Standard Form may be a suitable choice for a word list learning task (Delis et al., Reference Delis, Kramer, Kaplan and Ober2017). However, future research is needed to explore how people experiencing cognitive symptoms during menopause perform on these tasks. Although working memory and verbal fluency were not consistently associated with subjective complaints in our review, previous research has highlighted their potential relevance. Accordingly, future studies should apply the same criteria, prioritizing tests with established construct validity and high ceilings when selecting measures for these domains. Although this study did not seek to identify causal factors between menopause and cognitive decline, exploring these avenues would help guide future research to focus on the relationship between brain changes in menopause and factors related to learning efficiency.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1355617725101306.
Data availability
The data and code used in this study will be available upon reasonable request.
Funding statement
Rachel Furey is supported by a Research Training Program Scholarship from Monash University.
Competing interests
The authors have no conflicts of interest to report.
Open access
