Sex Differences on Montreal Cognitive Assessment and Mini-Mental State Examination Scores and the Value of Self-Report of Memory Problems among Community Dwelling People 70 Years and above: The HUNT Study

With a growing number of older adults worldwide, it is estimated that in 2050, >130 million people will have dementia [1, 2]. In most cases, the first sign of dementia is a decline in cognition, most commonly a decline in episodic memory from a previous level. Today, the general approach is to identify older people with such a decline and follow them up over years because 20–50% of people with mild cognitive impairment (MCI) will develop dementia [3‒5]. As no curative treatment exists for any brain disorder leading to dementia, screening is not recommended. However, simple, cheap, and valid case-detection tools to identify persons with high risk of developing dementia are requested so that these people can make plans for the future.

For >40 years, the Mini-Mental State Examination (MMSE) has been the most used instrument to detect cognitive impairment in older adults, and in Norway, a validated version of the MMSE has been in use for the last 30 years [6, 7]. Depending on age (higher age is associated with lower score), educational level (higher education is associated with higher score), and cultural background, a score below 24 or 25 or even 26 has been recommended for the definition of cognitive impairment of significant degree [8, 9]. Using one of these cut-offs, the MMSE has been found to be poor in detecting persons with MCI but valid to detect persons with dementia [10, 11]. For this reason, and in addition that the MMSE is under copyright restriction, new case-detection tools have been developed and validated [12]. One instrument that has gained popularity is the Montreal Cognitive Assessment (MoCA) that was developed mainly as an instrument to detect MCI [13]. The general view is that a score below 26 on MoCA, adjusted to 25 among people with education of 12 years and below, will identify people with MCI. Adjustment for age could also be considered [9, 13‒15]. A substantial number of studies conducted across the world, included in the systematic reviews by Pinto et al. [16] and O’Driscoll and Shaikh [17], have shown that MoCA is superior to MMSE in detecting MCI, but not dementia. However, one should bear in mind that most of these studies used the Mayo criteria for the diagnosis of MCI [18]. One of the Mayo criteria is that a diagnosis of MCI can only be made when a person has a score of −1.5 standard deviation below what is normal for the person’s age and education. If a score on MoCA or MMSE correlates highly with the cognitive test used to diagnose MCI, this could lead to a circularity problem. To avoid circularity problems, a cognitive test should be used that correlates poorly with MoCA and MMSE.

The main differences between the 2 instruments are that the MMSE has a larger focus on verbal abilities than MoCA, and it is easier to achieve a high score, for example, 28–30 points on the MMSE compared to MoCA [16, 18]. MoCA contains more items testing executive and visuoconstructive function and covers more cognitive domains than MMSE. Both instruments are scored from 0 to 30, and many studies have explored how the test scores of the 2 tests correspond to each other [9, 15‒17, 19‒22]. The main explanations for the different scores are that MoCA is more comprehensive and challenging as it is developed to detect MCI, whereas the MMSE originally was developed to detect delirium and psychiatric condition among patients in specialist health care [6].

Studies comparing the MMSE and MoCA have to a large degree included clinical samples, not community samples, to examine the discriminatory power of the tests to separate cognitively normal older adults from those with MCI and dementia [9, 13, 15‒17, 19‒22]. In these studies, adjustments have been made according to people’s age and education, but not according to sex [9, 13, 20‒22]. A few community studies have examined whether older women and men score differently on the 2 tests, and whether sex influences differently on the scores of the 2 tests [23‒27]. These studies could report sex differences, especially by the use of MoCA, with higher scores among women [23‒26]. One larger community study reported that older women had a higher score on MMSE than men [27]. Surprisingly, few studies have examined whether sex differences can be detected using the MMSE and MoCA as it is well-known since >50 years that women outperform men on verbal tasks and men perform better than women on spatial tasks [28, 29]. Later, studies have showed that these sex differences also exist in older adults, with similar sex-associated differences seen in for example, verbal memory tests, perhaps caused by underlying pathological mechanisms, or it could indicate that the sex differences found early in life persist by increasing age [29‒33].

The association between memory complaints and significant cognitive impairment and dementia has been examined in numerous studies, but only some studies have compared self-reported cognitive complaints (SCC) with scores on the MMSE and MoCA among home-dwelling older people [34‒42]. All these studies indicate that older adults with SCC are at higher risk of developing dementia than those with no complaints, although a systematic review concludes that one should not use SCC as a predictor of future development of dementia [35]. Comparison of scores on the MMSE and MoCA with a single question of memory problems has been studied in clinical samples but less in community samples [43]. In the recent study of Wanrooji et al. [43], it was found that a single question of memory complaints was associated with recall of 3 words as measured by the MMSE and development of dementia after a mean of 6.7 years. This finding needs to be replicated because if a single question could be applied as a prescreening tool, we could better target who should be tested with the MMSE or MoCA. In addition, it adds knowledge to what has been proven useful in clinical practice and research in many areas of medicine: we should listen to the patient’s experience of own disease. Probably, a general question of any memory or other cognitive problems would not be useful, whereas a specific question related to duration and severity of memory problems or memory problems related to certain situations would be more sensitive. Further, it is suggested that older women more often complain about memory problems than men, but we do not know whether this could be due to depression, which is more prevalent in women [44, 45]. Thus, it is still of interest to examine whether older women experience more about memory problems than men and if such experiences are associated with scores on the MMSE and MoCA.

To summarize, by searching the literature, only a few community studies were found that compared the total and the domain or item scores on the MMSE and MoCA between older women and men. Although many studies have been carried out to examine associations between self-report of memory problems and future development of dementia, few community studies have been published that examined the association between a single question of memory complaint and the scores on the MMSE and MoCA in older women and men. Thus, the aims of the present study were 3-fold: (1) to examine whether sex influenced on the total and domain or item scores on the MMSE and MoCA, (2) to examine sex differences between complainers and non-complainers of memory problems, and (3) to examine whether scores on the MMSE and MoCA differ between complainers and non-complainers. Based on the literature, our hypotheses are that women compared to men will score higher on MoCA, but not on the MMSE, and we expect that self-report of memory problems is more common in women than men, whereas the question is open whether older adults with self-report of memory problems score lower on the MoCA and MMSE.

This cross-sectional study took place between April and June 2019 together with the population-based the Trøndelag Health Study in the city of Trondheim (HUNT4 Trondheim 70+). For the present study to compare the results on the MoCA and MMSE, we invited 504 home-dwelling people aged 70 years and older who had participated in the HUNT study. In all, 309 (61.3%) people consented to participate. The mean age of the 309 participants was 75.8 (SD 4.4) years; 155 were women with a mean age of 75.6 (SD 4.1) years, and 154 were men with a mean age 76.0 (SD 4.6) years. Those who declined participation were 100 women and 95 men. They did not differ from those who participated with regard to age (p = 0.91), sex (p = 0.81), or educational level (p = 0.67). As all 504 participants were tested with MoCA as part of the ongoing HUNT4 study, we could compare the scores of the 309 participants with the 195 who declined for being tested with the MMSE. Those who declined had a significantly lower score on MoCA than those who participated (23.5 [SD 3.5] vs. 24.5 [SD 3.2], p = 0.002).

All participants were tested with the 3rd Norwegian version of the MMSE and the Norwegian version (7.1) of MoCA at a test station at 2 visits within 1 week by trained health personnel. MoCA was applied at the first visit and the MMSE at the second. The MMSE is a global screening test that tests orientation, memory, language, and visuospatial function with scores from 0 to 30; a higher score indicates better cognition [6]. The MoCA scale is also a multidomain cognitive screening instrument, testing memory, visuospatial and executive functions, naming, attention, abstraction, language, and orientation with scores from 0 to 30; higher scores indicate better cognitive function [13]. In addition, we used the Word List Memory Task (WLMT) from the Consortium to Establish a Registry for Alzheimer’s Disease (CERAD) [46]. It tests memory with a list of ten words, which are repeated in a different order 3 times, that should be recalled immediately after each presentation (immediate recall or learning) and recalled after 10 min (delayed recall). The maximum score for immediate recall is 30 and for delayed recall, 10. A higher score indicates better verbal memory performance [46]. We asked the participants 2 patient-related outcome measures (PROM): a general question about memory problems: “have you experienced any memory problems?” (yes/no) and a specific question: “have you experienced significant memory problems the last 5 years?” (yes/no). The age at test date, sex, and self-report of the educational level expressed as years of schooling were recorded.

For data recording and analyses, we used SPSS, 25th edition. Before analyses, we divided the participants into 3 age groups: 70–74 years, 75–79 years, and 80+ years. The educational level was divided into 3 groups according to the Norwegian school system 50 years ago. Ten years of schooling represented primary school + middle school, 11–12 years high school, and 13 years and above academic education. The χ² test was used to test for differences in table analyses and with linear-by-linear association for 3 × 2 tables. As the normality assumption was violated for the test scores, Mann-Whitney U and Kruskal-Wallis tests were used for group comparison of results on the MMSE and MoCA. Bonferroni adjustment was used for multiple testing. Spearman’s rho, a rank correlation test, was used for correlation analyses. To explore which variables were associated with the total MMSE and total MoCA scores, and the difference between MMSE and MoCA-scores, we conducted 3 multivariate linear regression analyses including sex, age, education, and self-report of memory problems as independent variables. Enter, forward, and backward stepwise methods were performed, which did not change the results. The results of the enter method were recorded.

Personal characteristics of the participants and the results of the WLMT and the 2 PROM questions divided by sex are shown in Table 1. As can be seen, no differences were found between women and men related to age groups, but fewer women had higher academic education than men (p < 0.001). Women outperformed men on WLMT.

Results of the MMSE and MoCA broken down by sex, age, educational groups, and the 2 self-reported questions of memory problems are displayed in Table 2. To further illustrate the different distribution of the scores between the 2 tests, we added the number of people with a score below 26 on each test. It should be noted that only 5 people had a score below 24, and 6 scored 24 on the MMSE. No one scored below 21 on the MMSE, whereas 31 women and 32 men, together 63 (20.4%) reached the maximum score of 30.

Although the total score on the MMSE and MoCA did not differ between women and men (Table 2), stratifying the analysis by taking education into the analyses showed one difference. Women with 11–12 years of schooling scored significantly higher on the total MoCA than men, 24.4 (SD 3.4) versus 22.7 (SD 3.4), p = 0.005 (adjusted with Bonferroni correction). Otherwise, we found no differences in the total MMSE and MoCA scores between women and men in the various age and educational groups or according to self-report of memory problems.

Examination of the scores on the various domains/items revealed some differences between women and men. On MoCA, women scored significantly higher on the item delayed recall (see Table 3). This sex difference was also found in the age group 70–74 years, age group 80 years+, and the group with education of 11–12 years with mean scores (SD) for women and men: 3.1 (1.5) versus 2.6 (1.4), p = 0.003; 3.0 (1.7) versus 2.0, p = 0.04; and 3.2 (1.5) versus 1.8 (1.5), p < 0.001, respectively.

Men had significantly higher scores than women on the items visuoconstruction (copying a cube and clock drawing test) and serial subtraction on MoCA (see Table 3), and on the item serial sevens on the MMSE, 4.5 (0.8) versus 4.1 (1.1.), p < 0.001. Such differences between men and women were further found for some subgroups: visuocontruction (MoCA) men 70–74 years 3.9 (0.8) versus women 3.5 (1.2), p 0.01; serial subtraction (MoCA) men 75–79 years 2.8 (0.6) versus women 2.4 (0.9), p = 0.04; and serial sevens (MMSE) men with 13+ years of education 4.7 (0.6) versus women 4.2 (1.1), p = 0.001.

We found a weak to moderate rank correlation between the total score on the 2 tests (Spearman’s rho 0.36). We further calculated rho for women and men separately and for participants according to age and education. Spearman’s rho varied and was lowest for the group with ≤10 years of schooling (0.16) and highest for the group with 11–12 years of schooling (0.43).

The mean difference between the MMSE and MoCA scores was 3.7 (SD 2.7), highest on the MMSE, and varied between 3.4 (SD 2.7) for participants with 13+ years of education (lowest) and 4.3 (SD 3.9) for those with education of 10 years and below (highest). Among all, 277 (89.6%) had a higher score on the MMSE than on MoCA, 18 (5.8%) had an equal score on both tests (11 were women), and 14 (4.5%, 11 were women) scored higher on MoCA. Those who scored higher on MoCA or equal on both tests were significantly younger (p = 0.004), had higher education (p = 0.02), and were women (p = 0.02).

For the 2 self-reported questions, the prevalence of a “yes” was not different for women and men. For the question “have you experienced significant memory problems the last 5 years?” we found several significant differences in scores on both the MMSE and MoCA for people who answered “yes” (lower scores) compared to those who answered “no” (higher scores) to this question (see Table 4).

Lastly, we performed 3 linear regression analyses (Table 5). Using the MMSE score as the dependent variable, we found that a higher score on the MMSE was associated with younger age, higher education, and reporting no significant memory problems the last 5 years. A higher score on MoCA was associated to sex, age, education, and reporting no memory problems the last 5 years. These 2 regression models accounted for 15 and 14% of the explained variance, respectively. The model using the difference of MMSE-MoCA as the dependent variable explained only 5% of the variance.

To summarise, according to the adjusted regression analysis, we could confirm our hypothesis that women performed better on MoCA than men, but not on the MMSE. We further found that women had a higher score on the delay recall item on MoCA, and men outperformed women on the items serial sevens on the MMSE and serial subtraction and on the domain visuoconstruction on MoCA. Self-report of memory problems was equally common in men and women, and complaints about significant memory problems the last 5 years were associated with lower scores on the MMSE and MoCA. In addition, the participants had in general a higher score on the MMSE than on MoCA, and the rank correlation between the 2 test scores was poor to moderate. Not surprisingly, we found that education and age influenced on both test scores.

In the adjusted regression analyses (Table 5), we found that female sex was associated with a higher score on MoCA, but not on the MMSE. This finding is in line with a few community studies [23‒27]. However, this sex difference was not found by comparing the total MoCA scores among women and men, as displayed in Table 2. The explanation for this diverse result shown in Tables 2 and 5 is probably because fewer women than men had education of 13 years and above, and very few men had low education (Table 1). This imbalance could influence the total scores on both tests. In the multivariate regression analyses, the difference in education between sexes was controlled (men with education of 13 years and older had the highest score on MoCA). We did not find an association between sex and MMSE score in the adjusted regression analysis (Table 5) nor in the total MMSE score. This suggests that the score on the MMSE is not influenced by sex, possibly due to the recall item of the MMSE (recalling 3 words), and the only spatial item (drawing 2 pentagons) is not sensitive enough to separate between the women and men in a community study. Our finding is, however, in contrast to another and larger community study that reported a higher MMSE score in older women [27]. It should be noted that only 21 scored below 26 and 63 had a maximum score of 30 on the MMSE. This narrow distribution of MMSE scores, having most scores between 26 and 29, could have influenced the results.

The standardized beta in the regression analyses was about the same for the variable sex as for age and half of that of education. Comparable betas are reported by Borland et al. [25] and Thomann et al. [26], 2 community-based MoCA studies. Thomann et al. [26] estimated normative data of MoCA and could show that the score defining −1.5 SD was 1 point higher in women than in men, regardless of age and educational level [26]. We suggest that the estimated higher score in women in the study by Thomann et al. [26] can be related to better verbal memory performance in older women, but neither Borland et al. [25] nor Thomann et al. [26] reported scores on domains/items of the MoCA, so we cannot compare our findings with results from these 2 studies [25, 26]. Direct comparison is further difficult as we have no normative data on MoCA in Norway. Several studies have shown that women, also older women, performed better than men on verbal memory tests [25‒33, 47, 48]. Two larger studies could confirm that these differences were stable over a period of 9–10 years in older adults [30, 33]. The present study confirms these previous findings as we could document significantly higher scores on the item delayed recall on MoCA and WLMT in older women than in older men. However, performance on the delayed recall on the MMSE did not differ between sexes probably because it is less challenging to remember 3 words of the MMSE compared to 5 on the MoCA and ten on the WLMT. We suggest that MoCA is superior to the MMSE in this respect and is one reason why MCI is better detected by MoCA than the MMSE as most MCI cases are of amnestic nature [3]. Another explanation could be that MoCA and the MMSE are differently constructed as MoCA covers more cognitive domains than the MMSE, which could have implications when summarizing scores from sub-items into a total score. It is likely that sub-items testing many different cognitive functions (MoCA) will result in a different total score compared to sub-items testing less different cognitive functions (MMSE). For this reason, MoCA could be more sensitive to cognitive decline.

Not surprisingly, men scored better than women on serial sevens on the MMSE and serial subtraction and visuoconstruction on MoCA, as found in previous studies [27‒33]. However, these differences are judged to be of little significance for the total scores of the 2 tests, since the results had no significant effect in the adjusted regression analysis. The male sex was not associated with neither the MoCA nor MMSE total score. We suggest that the sex differences on MoCA total and item scores should be further examined, since these differences could probably have a significant impact on normative data. Our results do not support that normative data for MMSE should be adjusted for sex differences.

About half of both women and men reported significant memory problems the last 5 years, and no difference was seen between sexes. We suggest that this result is valid. A recent large epidemiological community-based study from Norway reported that 35% of adults older than 70 years had a diagnosis of MCI, and a Swiss community-based study reported a prevalence of 18.5%. In addition, a review reported a prevalence of 23.8% for the diagnosis of subjective cognitive decline [42, 49, 50]. We suggest that raised awareness about memory problems as a possible first sign of dementia and insight into their own decline of memory function by increasing age could explain the high proportion of complainers of both sexes. There is no reason to believe that raised awareness and insight into their own memory decline should be different in older women and men.

The association between scores on self-report of memory problems was not of significance for the general question of “do you experience any memory problems?” but it was for the more specific question “have you experienced significant memory problem the last 5 years?” This finding is in line with our hypothesis. In the unadjusted group comparison analysis (Table 4) and the adjusted regression (Table 5), this specific question was significantly associated with the score on both MMSE and MoCA. The standardized beta for this PROM question in the adjusted regression analyses was almost as high as for education in explaining the scores on the 2 tests, indicating that self-report of memory problem could be used as a first question to older people before performing a cognitive test. Older people answering “yes” to the question have higher probability to have a lower score on tests like MMSE and MoCA. The association between self-report of memory problems the last 5 years and scores on the 2 tests indicates that this question could be used as one of various tools in the assessment of cognition and dementia. A few standardized self-report scales have been developed for clinical use, which are useful, but more time-consuming than a single question [37, 39, 40]. One of these is the Cognitive Functional Impairment (CFI) scale, which contains 14 specific questions of cognitive problems [51]. In a recent study, we found that this scale was valid for detecting older people with MCI, but not dementia [39]. This also indicates that people with normal or subnormal cognitive performance or even MCI have good insight into their own cognitive problems [39].

We could confirm the well-known findings that the educational level and age influence on both test scores, which indicate that the data used in the present study are comparable with data of many other studies in the field [9, 13, 20‒22, 52, 53]. We could also confirm that older people score higher on the MMSE than MoCA probably because MoCA is more challenging, and lastly, we have shown that the rank correlation between the 2 tests is weak and that the scores cannot be used side by side.

Having confirmed that age and educational level influence the total scores on both tests, the novel finding in this study is that sex as well influence the MoCA score, but not on the MMSE score. This is only reported in a few previous normative studies of MoCA, and only the study by Thomann et al. [26] calculated separate normal scores for older women and men [23‒26]. The difference in the MoCA score in the Swiss normative study was small, and the influence of sex according to the linear regression analysis (Table 5) on the MoCA score in the present study was as well small. Based on the results of the present study, we cannot recommend any normative scores for older women and men. We suggest that future normative large-scale studies of MoCA should include sex, not only age and education in developing normative scores.

Although the older adults were selected randomly, we cannot claim that the sample is representative for the older population of Trondheim, since 195 of 504 persons declined participation. Those who declined had significantly lower score on MoCA, but they did not differ from the participants regarding sex, age, and education. According to the MMSE scores, few participants had severe cognitive impairment meaning that the sample is not representative for this group of older adults. But, as the main aim of the study was to examine differences between sexes with regard to scoring on the MMSE, MoCA, and self-report of memory loss, we assume that the sample is representative in that respect. Another limitation is that several health workers did the testing, and we cannot exclude that test results could have been influenced by the health worker’s various experience. To increase inter-rater reliability, all health workers did the same training before the study. Some of the items of the MMSE and MoCA are similar, but most of them are not identical. Therefore, some learning effects could not be excluded as the 2 tests were administered within 1 week. The sample is relatively small, but as we found comparable results with other larger studies examining the influence of age and education, we assume that our dataset was satisfying to answer our research questions [16, 17]. Another limitation to consider is that we did not further examine if the participants had a diagnosis of dementia, and thus, we cannot exclude that some of the participants have dementia. Nevertheless, most participants had a high score on both tests, making it unlikely that many could have dementia. Only 11 persons scored between 21 and 24 on the MMSE, a finding that also support that only a few could have had dementia.

The main results of the study are that MoCA scores are associated with younger age, higher education, and female sex, whereas sex was not associated with the MMSE score. Older women scored significantly better than men on the item recalling 5 words on MoCA, whereas men scored significantly higher on the domain visuoconstruction on MoCA and the items serial subtraction on MoCA and serial sevens on the MMSE. The question “have you experienced significant memory problems the last 5 years?” was significantly associated with both MoCA and MMSE scores and may be useful to evaluate older people’s cognition.

We thank the Trøndelag Health Study (the HUNT study) for the use of data. The HUNT study is a collaboration between the HUNT Research Centre, (Faculty of Medicine and Health Sciences, NTNU, Norwegian University of Science and Technology), Trøndelag County Council, Central Norway Regional Health Authority, and the Norwegian Institute of Public Health. We further thank the participants in the current study, the students from the Norwegian University of Science and Technology in Trondheim who participated in the data collection, and Trondheim municipality and the Norwegian National Advisory Unit for Aging and Health for co-operation in the study.

All participants received oral and written information about the study and gave informed written consent. The project was submitted to the Regional Ethics Committee for Medical and Health Research in mid-Norway (reference REK 2018/1812) and was approved. It was further approved according to the General Data Protection Regulation by the Norwegian Centre for Research Data.

None of the authors have conflicts of interest.

The Norwegian National Advisory Unit on Aging and Health funded the study.

K.E. designed the study, analysed and evaluated the data, and wrote the preliminary and final draft of the manuscript and approved it. L.G. took part in design of the study, organised data, and gave input to evaluation of the analysis and the preliminary and final draft of the manuscript, and approved it. T.C.B., P.T., G.G., and L.E. gave input to evaluation of the analysis and the preliminary and final draft of the manuscript, and approved it. G.G. took part in design of the study and gave input to evaluation of the analysis and the preliminary and final draft of the manuscript, and approved it.