Putative Dementia Cases Fluctuate as a Function of Mini-Mental State Examination Cut-Off Points

Abstract

As the population ages, there is a growing need to quickly and accurately identify putative dementia cases. Many cognitive tests are available; among those commonly used are the Cognitive Dementia Rating (CDR) and the Mini-Mental Status Examination (MMSE). The aim of this work was to compare the validity and reliability of these cognitive tests in a primary care based cohort (pcb-Cohort). The MMSE and the CDR were applied to 568 volunteers in the pcb-Cohort. Distinct cut-off points for the MMSE were considered, namely MMSE 27, MMSE 24, and MMSE PT (adapted for the Portuguese population). The MMSE 27 identified the greatest number of putative dementia cases, and, as determined by the ROC curve, it was the most sensitive and specific of the MMSE cut-offs considered. Putative predictive or risk factors identified included age, literacy, depression, and diabetes mellitus (DM). DM has previously been indicated as a risk factor for dementia and Alzheimer’s disease. Comparatively, the MMSE 27 cut-off has the greatest sensibility (94.9%) and specificity (66.3%) when compared to MMSE PT and MMSE 24. Upon comparing MMSE and CDR scores, the latter identified a further 146 putative dementia cases, thus permitting one to propose that in an ideal situation, both tests should be employed. This increases the likelihood of identifying putative dementia cases for subsequent follow up work, thus these cognitive tests represent important tools in patient care. Further, this is a significant study for Portuguese populations, where few of these studies have been carried out.

Keywords

Cognitive Dementia Rating cognitive testing pcb-Cohort primary care

INTRODUCTION

Dementia prevalence and associated pathologies increase as the population ages, consequently demanding a need for an accurate timely diagnosis to identify early onset cases. Furthermore, the increase in new disease-modifying treatments accentuates this necessity. Although molecular biomarkers are being extensively researched [1], cognitive screens are useful tools, particularly in primary care settings. Among the most commonly applied cognitive screens are the Cognitive Dementia Rating (CDR) and the Mini-Mental Status Examination (MMSE). The former is considered the gold standard for staging dementia severity [2, 3]. The CDR has been standardized for Alzheimer’s disease (AD) and is widely used in dementia research and even clinical trials [4 –6]. Further, current research continues to employ CDR, for example, in monitoring acetylcholinesterase inhibitor (AChEI) and memantine AD treatment strategies [7], furthermore shorter modified versions have also been tested (mCDR) [8]. Likewise, the MMSE continues to be an area of current research, particularly with respect to entry into AD clinical trials [9], and in efficacy and safety, for instance in studying the interaction of leuprolide acetate with AChEIs in the treatment of AD [10].

The MMSE, originally published in 1975 [11], evaluates orientation, retention of information, attention and calculation, evocation, and language. It has a maximum score of 30 points, sub-divided into two parts. The first, with a total of 21 points, evaluates verbal response on time orientation, spatial orientation, memory, attention and calculation, and memory recall. The second, with a total of 9 points, evaluates the nomination of skills, language, writing, reading, and executing tasks, as well as the ability to copy a complex polygon [11 –14]. The MMSE is commonly applied and studies have indicated that 90% of specialists use it [15 –17]. MMSE and CDR have been valuable tools for dementia screening[18, 19], with a sensitivity range between 44–100% and specificity between 46–100% [20 –24], altering in response to treatment [24, 25]. The disadvantage is mainly in relation to the second section of the test that requires writing and visual skills. Therefore, patients with low literacy, visual problems or other language disorders may not be well evaluated [11]. Of note, a report by Mitchell [26] indicated a sensitivity of 77% and a specificity of 90% for application in high specialist settings, while in primary care settings a sensitivity of 81% and a specificity of 87% were reported. Although used in primary care settings [26 –28], the MMSE is perceived as a time consuming test; thus opinion is divided regarding its adequacy as well as the optimal cut-off thresholds [23, 29]. Recommended cut-offs vary, with the literature recommending a cut-off at 24/25 [11 , 31] or 27 [14]. The German version uses the total score of the MMSE, this can be applied as a cut-off point of 27 categorizing patients “cognitively impaired” (score 27–30), “mild” (score 20–26), “moderate” (score 10–19), and “severe cognitive impairment” (score 0–9); but the authors apply the criterion of the cut-off point 27 using only two categories (with or without cognitive impairment). The MMSE was subsequently adapted for the Portuguese population, by taking into account levels of literacy. For 0–2 years of literacy, cut-off was 22; for 3–6 years of literacy, the cut-off was 24; and above 7 years of literacy, cut-off was 27 [32, 33]. The intent of the study here presented was to compare the efficacy of distinct cut-off thresholds for the MMSE and to address if the combination of MMSE and CDR increase the sensitivity and specificity in comparison to the separate use of each ofthe tests.

MATERIALS AND METHODS

Study design

A cross-sectional population-based survey of patients attending primary care health centers in the Aveiro region was carried out. The primary care based cohort (pcb-Cohort) used in this study, includes patients aged 50 years old or more and attending primary health care centers in the Aveiro region of Portugal who were invited to participate in the study. The Aveiro district had 78,450 habitants in the 2011 CENSUS (http://www.ine.pt), thus the target size should include at least 480 individuals [34]. Informed consent and inclusion and exclusion criteria were as previously described [35 –37], but to summarize, patients undergoing chemotherapy, radiotherapy, using illicit drugs, or were unable to answer the questionnaires were excluded. Of the 590 volunteers, 568 were included in the study, which was carried out in 2012.

The procedures involving data collection and blood from human subjects were carried out in accord with the ethical standards and in accord with the Helsinki Declaration of 1975. The study was approved by the ethical committee from the ‘Administração Regional de Saúde do Centro IP’ observing National guidelines ‘Lei N° 67/68 de 26 de Outubro da Assembleia da República e da Declaração N° 227/2007 da Comissão Nacional de Protecção de Dados’.

Procedures and instruments

Semi-structured questionnaires addressing sociodemographic, clinical and cognitive characteristics were applied [37]. For cognitive evaluation, the MMSE and CDR tests were implemented. The MMSE scores, were analyzed using three model ratings. The three cut-off points applied were MMSE 24 [11], MMSE 27 [14], and MMSE PT adapted for the Portuguese population with respect to literacy [32]. The pcb-Cohort was also rated according to the CDR scores (Fig. 1). The CDR applied a 0 to 3 score [2], where 0 = Normal, 0.5 = suspect, questionable or very mild dementia, and CDR≥1 [1 –3] = mild, moderate or severe dementia. Participants were also genotyped for ApoE using procedures previously described [37].

Fig.1

Cognitive evaluation tests applied to the pcb-Cohort. A pool of 568 individuals were subjected to cognitive tests; the MMSE and the CDR. Three distinct analyses were applied to the MMSE: the first MMSE classification is based on normative values for the Portuguese Population (MMSE PT); the second MMSE on the cut-off Point in the original test (Cut Point 24, included MMSE 24); and the third is based on a German Study (Cut Point 27, included MMSE 27). The numbers scoring positive in each case are indicated.

Statistical analysis

Data were analyzed using SPSS version 22. The level of statistical significance was set to p < 0.05 in all analyses. The sample size adopted fulfilled normality criteria, the central theorem. For bivariate analysis, the Student t Test and the chi-squared tests were used. Multivariate analysis applied logistic regression. The performance of different MMSE cut-off points in predicting dementia was assessed. The true patients with dementia were considered to be those with positive MMSE and positive CDR scores. Data analyses also took advantage of ROC curve and Venn diagram [38] outputs for better interpretations.

RESULTS

Characterization of the study group as a function of MMSE cut-offs

As indicated in the methodology, distinct MMSE cut-offs were employed and applied to the patient data collected for the 568 individuals (Fig. 1). Namely the cut-offs were MMSE 24, MMSE 27, and MMSE PT. Individuals were scored as positive or negative for each of the MMSE cut-offs applied.

Volunteers, 50 years old or more, were evaluated as described above and the sociodemographic results, as a function of the distinct MMSE cut-offs, are presented in Table 1. By applying bivariate analysis, a series of characteristics correlated significantly with dementia for all MMSE cut-offs, namely, age, monthly income, and educational level. With age; the older individuals were more likely to score positive, for example for the MMSE 27; in the age gap 75–90 years old 47.7% scored positive, contrasting significantly with the 15.2% positive score for the 50–64 age group. Lower income individuals as well as those with fewer years of literacy were more likely to score positive for dementia, across all MMSEs. The professional status was significant only for the MMSE 27 and MMSE 24, and the latter also identified marital status to be of significant relevance correlating with dementia. The demographic characteristic of gender and living arrangement did not correlate with the MMSE for this studygroup.

Table 1

Sociodemographic characteristics in the pcb–Cohort and MMSE performance

Demographic	n	MMSE PT			MMSE Cut point 24			MMSE Cut Point 27
Characteristics		Negative	Positive	p-value	Negative	Positive	p-value	Negative	Positive	p-value
		(n = 514)	(n = 54)		(n = 495)	(n = 73)		(n = 436)	(n = 132)
Gender^*
Male	163	147^a (28.6%)	16^a (29.6%)	0.873	143^a (28.9%)	20^a (27.4%)	0.793	124^a (28.4%)	39^a (29.5%)	0.806
Female	405	367^a (71.4%)	38^a (70.4%)		352^a (71.1%)	53^a (72.6%)		312^a (71.6%)	93^a (70.5%)
Age
Age ≥65 y^*	366	322^a (62.6%)	44^b (81.5%)	0.006	302^a (61%)	64^b (87.7%)	<0.001	254^a (58.3%)	112^b (84.8%)	<0.001
Age^*
[50–64]	202	192^a (37.4%)	10^b (18.5%)	<0.001	193^a (39.0%)	9^b (12.3%)	<0.001	182^a (41.7%)	20^b (15.2%)	<0.001
[65–74]	229	215^a (41.8%)	14^b (25.9%)		207^a (41.8%)	22^a (30.1%)		180^a (41.3%)	49^b (37.1%)
[75–90]	137	107^a (20.8%)	30^b (55.6%)		95^a (19.2%)	42^b (57.5%)		74^a (17.0%)	63^b (47.7%)
Age (y) (mean±SD)^†	–	66.9±8.8^a	74.6±10.6^b	<0.001	66.5±8.7^a	75.1±9.3^b	<0.001	65.9±8.6^a	73.5±9^b	<0.001
Marital Status^*
Living with the partner	400	368^a (71.6%)	32^a (59.3%)	0.059	356^a (71.9%)	44^b (60.3%)	0.042	312^a (71.6%)	88^a (66.7%)	0.281
Others Situation	168	146^a (28.4%)	22^a (40.7%)		139^a (28.1%)	29^b (39.7%)		124^a (28.4%)	44^a (33.3%)
Living Arrangement^*
Alone	96	87^a (16.9%)	9^a (16.7%)	0.961	83^a (16.8%)	13^a (17.8%)	0.825	71^a (16.3%)	25^a (18.9%)	0.476
Accompanied	472	427^a (83.1%)	45^a (83.3%)		412^a (83.2%)	60^a (82.2%)		365^a (83.7%)	107^a (81.1%)
Professional Status^*
Active	176	164^a (31.9%)	12^a (22.2%)	0.240	161^a (32.5%)	15^b (20.5%)	0.042	145^a (33.3%)	31^b (23.5%)	0.022
Reformed	361	321^a (62.5%)	40^b (74.1%)		305^a (61.6%)	56^b (76.7%)		264^a (60.6%)	97^b (73.5%)
Unemployed	31	29^a (5.6%)	2^a (3.7%)		29^a (5.9%)	2^a (2.7%)		27^a (6.2%)	4^a (3.0%)
Monthly Family^*
≤1 minimum wage	176	151^a (29.4%)	25^b (46.3%)	0.011	145^a (29.3%)	31^b (42.5%)	0.023	120^a (27.5%)	56^b (42.4%)	0.001
>1 minimum wage	392	363^a (70.6%)	29^b (53.7%)		350^a (70.7%)	42^b (57.5%)		316^a (72.5%)	76^b (57.6%)
Education Level^*
0–2 y of literacy	51	33^a (6.4%)	18^b (33.3%)	<0.001	21^a (4.2%)	30^b (41.1%)	<0.001	10^a (2.3%)	41^b (31.1%)	<0.001
3–6 y of literacy	402	375^a (73.0%)	27^a (50.0%)		366^a (73.9%)	36^b (49.3%)		321^a (73.6%)	81^b (61.4%)
≥7 y of literacy	115	106^a (20.6%)	9^a (16.7%)		108^a (21.8%)	7^b (9.6%)		105^a (24.1%)	10^b (7.6%)
Y of Study (mean±SD)^†	–	5.3±3.7^a	3.6±3.2^b	0.001	5.5±3.6^a	2.9±2.9^b	<0.001	5.7±3.7^a	3.1±2.6^b	<0.001

Data are presented as n (%) and % is expressed as a function of the total in each of the MMSE groups or mean±standard deviation. MMSE PT, cut-off point adapted for the Portuguese population based on literacy (22 to 0–2 years of literacy; 24 to 3–6 years of literacy; 27 to ≥7 years of literacy); MMSE cut-off point 24 based on previous studies; MMSE cut-off point 27, based on a German study (see methods). Statistical test used: (*) Chi square (χ²) test; (^†) Student t Test; ^a,bthe same subscript letter denotes a subset of MMSE categories whose column proportions do not differ significantly from each other at the p-value <0.05. Different subscript letters denote column proportions that differ significantly from each other. MMSE, Mini-Mental State Examination.

Subsequently the pcb-Cohort was analyzed with respect to comorbidities and putative dementia. A positive correlation with neuropathological disorders was obtained for the three MMSE cut-off analyses (Supplementary Table 1). In effect, one can confirm that the diagnosis of neuropathological conditions was correct in this population. With respect to other comorbidities, there was no consistency for the different MMSE cut-off points. For the MMSE PT cut-off only, a moderate correlation with depression was obtained (p-value 0.081), with the MMSE 24 cut-off a moderate correlation with hypertension was evident (p-value 0.075) and with the MMSE 27 cut-off a p-value of 0.072 was obtained for DM (Supplementary Table 1). To better address the comorbidities’ correlations, the three MMSE cut-offs were reanalyzed but each was subdivided into three age groups (Table 2). Being diagnosed with a neuropathology correlated significantly with low cognitive scores, and this was true independent of the MMSE cut-off applied. There is an exception for the younger age group and the MMSE 27 cut-off, but this may be due to the fact that we are dealing with a very small sample. In the MMSE 24 cut-off, hypertension appeared to be of significant relevance as a factor putatively contributing to dementia (Supplementary Table 1), but this was not sustained when the group was reevaluated with respect to age groups (Table 2). In a marked contrast, depression correlated with poor results in cognitive performance for all MMSE cut-offs. For the MMSE 24 cut-off in the 65–74 age group and in the MMSE PT and MMSE 27 cut-offs, significance was evident only in the older age group; more than 75 years old. In fact, consistency between MMSE PT and MMSE 27 cut-offs is also evident for DM, albeit in the 65 to 74 age group (Table 2). One can tentatively hypothesize that this is no longer evident in the older age group, as these individuals will have been lost from the population.

Table 2

Comorbidities and cognitive performance by age group

Age Group	Comorbidities	n	MMSE PT			MMSE Cut point 24			MMSE Cut Point 27
			Negative	Positive	p-value	Negative	Positive	p-value	Negative	Positive	p-value
50–64 y	DM–	164	156^a (81.3%)	8^a (80.0%)	0.921	158^a (81.9%)	6^a (66.7%)	0.254	148^a (81.3%)	16^a (80.0%)	0.886
n = 202	DM+	38	36^a (18.8%)	2^a (20.0%)		35^a (18.1%)	3^a (33.3%)		34^a (18.7%)	4^a (20.0%)
	HYP–	100	96^a (50.0%)	4^a (40.0%)	0.537	97^a (50.3%)	3^a (33.3%)	0.321	91^a (50.0%)	9^a (45.0%)	0.671
	HYP+	102	96^a (50.0%)	6^a (60.0%)		96^a (49.7%)	6^a (66.7%)		91^a (50.0%)	11^a (55.0%)
	DEP–	129	123^a (64.1%)	6^a (60.0%)	0.794	123^a (63.7%)	6^a (66.7%)	0.858	119^a(65.4%)	10^a (50.0%)	0.174
	DEP+	73	69^a (35.9%)	4^a (40.0%)		70^a (36.3%)	3^a (33.3%)		63^a (34.6%)	10^a (50.0%)
	NEURO–	198	189^a (98.4%)	9^a (90.0%)	0.062	190^a (98.4%)	8^b (88.9%)	0.044	179^a (98.4%)	19^a (95.0%)	0.307
	NEURO+	4	3^a (1.6%)	1^a (10.0%)		3^a (1.6%)	1^b (11.1%)		3^a (1.6%)	1^a (5.0%)
65–74 y	DM–	181	173^a (80.5%)	8^b (57.1%)	0.038	166^a (80.2%)	15^a (68.2%)	0.188	147^a (81.7%)	34^a (69.4%)	0.061
n = 229	DM+	48	42^a (19.5%)	6^b (42.9%)		41^a (19.8%)	7^a (31.8%)		33^a (18.3%)	15^a (30.6%)
	HYP–	78	73^a (34.0%)	5^a (35.7%)	0.893	72^a (34.8%)	6^a (27.3%)	0.480	61^a (33.9%)	17^a (34.7%)	0.916
	HYP+	151	142^a (66.0%)	9^a (64.3%)		135^a (65.2%)	16^a (72.7%)		119^a (66.1%)	32^a (65.3%)
	DEP–	151	144^a (67.0%)	7^a (50.0%)	0.194	140^a (67.6%)	11^a (50.0%)	0.097	119^a (66.1%)	32^a (65.3%)	0.916
	DEP+	78	71^a (33.0%)	7^a (50.0%)		67^a (32.4%)	11^a (50.0%)		61^a (33.9%)	17^a (34.7%)
	NEURO–	220	208^a (96.7%)	12^b (85.7%)	0.040	201^a (97.1%)	19^b (86.4%)	0.014	175^a (97.2%)	45^a (91.8%)	0.085
	NEURO+	9	7^a (3.3%)	2^b (14.3%)		6^a (2.9%)	3^b (13.6%)		5^a (2.8%)	4^a (8.2%)
≥75 y	DM–	108	85^a (79.4%)	23^a (76.7%)	0.743	76^a (80.0%)	32^a (76.2%)	0.615	60^a (81.1%)	48^a (76.2%)	0.485
n = 137	DM+	29	22^a (20.6%)	7^a (23.3%)		19^a (20.0%)	10^a (23.8%)		14^a (18.9%)	15^a (23.8%)
	HYP–	39	29^a (27.1%)	10^a (33.3%)	0.504	27^a (28.4%)	12^a (28.6%)	0.986	19^a (25.7%)	20^a (31.7%)	0.433
	HYP+	98	78^a (72.9%)	20^a (66.7%)		68^a (71.6%)	30^a (71.4%)		55^a (74.3%)	43^a (68.3%)
	DEP–	106	88^a (82.2%)	18^b (60.0%)	0.010	77^a (81.1%)	29^a (69.0%)	0.122	62^a (83.8%)	44^a (69.8%)	0.052
	DEP+	31	19^a (17.8%)	12^b (40.0%)		18^a (18.9%)	13^a (31.0%)		12^a (16.2%)	19^a (30.2%)
	NEURO–	128	105^a (98.1%)	23^b (76.7%)	<0.001	94^a (98.9%)	34^b (81.0%)	<0.001	73^a (98.6%)	55^b (87.3%)	0.008
	NEURO+	9	2^a (1.9%)	7^b (23.3%)		1^a (1.1%)	8^b (19.0%)		1^a (1.4%)	8^b (12.7%)

The analysis was carried out for each age group, taking into consideration the cognitive performance in patients with different comorbidities. Data are presented as n (%) and % is expressed as a function of the total in each of the MMSE groups. MMSE PT, cut-off point adapted for the Portuguese population based on literacy (22 to 0–2 years of literacy; 24 to 3–6 years of literacy; 27 to ≥7 years of literacy); MMSE cut-off point 24 based on previous studies; MMSE cut-off point 27, based on a German study (see methods). HYP, arterial hypertension; DEP, depression; NEURO, neurologic disease. Statistical test used: Chi square (χ²) test; ^a,bthe same subscript letter denotes a subset of MMSE categories whose column proportions do not differ significantly from each other at the p-value<0.05. Different subscript letters denote column proportions that differ significantly from each other at the p-value <0.05.

Fig.2

Receiver operating characteristic (ROC) curve for MMSE different cut-offs to predict dementia. Receiver operating characteristic curves for MMSE 27, MMSE 24, and MMSE PT cut-off points in screening for dementia. AUCs: MMSE 27 = 0.879 (CI 95% 0.807–0.952); MMSE24 = 0.882 (CI 95%:0.815–0.948); MMSE PT = 0.880 (CI 95%: 0.816–0.944). AUC, area under curve; CI, confidence interval.

ApoE genotyping was carried out for the pcb-Cohort, given that it has been described as the main genetic risk factor for dementia [37 , 39–41]. ApoE genotyping as a predictive factor (Supplementary Table 2) was not consistent for all the MMSE cut-offs; in particular, no correlation was obtained with MMSE PT. Previous studies have shown that ɛ4 allele carriers are at increased risk of developing AD [39]. In the pcb-Cohort a 0.064 p-value was obtained for ɛ4 allele carriers with poor cognitive performance, as evaluated with the MMSE 27 cut-off, but a similar correlation was obtained for ɛ3 allele carriers. The ɛ2 allele revealed no correlation withMMSE.

Comparative accuracy of putative dementia as a function of MMSE cut-offs

In order to compare across all the characteristics multivariate analysis was carried out using a Regression Logistic Model (Table 3). The likelihood ratio test was used to assess the significance of the fitted model. As the p-value≤0.000, for a significance level of 0.05, one can conclude that the model is statistically significant, that is, at least one of the explanatory variables influences the determination of cognitive impairment for each of the three MMSE cut-offs. The Hosmer and Lemeshow Test was used to evaluate the fit of the model. As p-value >0.05, for a significance level of 0.05, do not reject the hypothesis that the model fits the data. To summarize, it was possible to confirm that age and educational level were significant for all three MMSE cut-offs (Table 3), and are therefore strong risk factors for cognitive deficits. However, with respect to comorbidities strong correlations were obtained for neuropathologies for all MMSE cut-offs and DM but only for the MMSE 24 and MMSE 27 cut-offs (Table 3). Thus it would appear that the MMSE 24 and MMSE 27 cut-off points readily identify further risk factors. In fact, the bivariate and multivariate analysis largely reveal consistent data regarding risk factors for dementia as determined by the MMSE.

Table 3

Multivariate analysis based on regression logistics to different MMSE cut-off points

Characteristics	MMSE PT			MMSE Cut Point 24			MMSE Cut Point 27
	p-value	OR	CI	p-value	OR	CI	p-value	OR	CI
ApoE
ɛ2 carriers	0.328	0.470	(0.104–2.135)	0.169	0.382	(0.097–1.506)	0.499	0.692	(0.239–2.008)
ɛ3 carriers	0.998	0.000	–	0.278	3.123	(0.399–24.466)	0.274	2.411	(0.497–11.689)
ɛ4 carriers	0.776	0.865	(0.319–2.347)	0.550	0.738	(0.272–2.001)	0.169	0.618	(0.312–1.226)
Female	0.866	0.920	(0.350–2.421)	0.282	0.602	(0.239–1.517)	0.760	0.904	(0.473–1.728)
Age
50–64 y	0.042	–	–	<0.001	–	–	<0.001	–	–
65–74 y	0.530	1.421	(0.475–4.252)	0.039	3.433	(1.065–11.069)	0.006	2.853	(1.355–6.008)
≥75 y	0.023	3.499	(1.189–10.293)	<0.001	10.134	(3.041–33.767)	<0.001	7.542	(3.397–16.744)
Education Level
0–2 y of Literacy	0.008	–	–	<0.001	–	–	<0.001	–	–
3–6 y of Literacy	0.230	2.220	(0.603–8.170)	<0.001	12.282	(3.168–47.625)	<0.001	26.446	(8.216–85.124)
≥7 y of Literacy	0.207	0.503	(0.173–1.461)	0.819	0.871	(0.268–2.831)	0.143	1.934	(0.801–4.673)
Marital Status	0.838	1.100	(0.442–2.737)	0.571	1.292	(0.533–3.135)	0.225	1.475	(0.787–2.765)
Monthly income	0.903	0.946	(0.388–2.306)	0.355	0.659	(0.272–1.597)	0.699	1.125	(0.618–2.047)
Dependent Katz	<0.001	7.428	(2.417–22.829)	<0.001	11.077	(3.176–38.636)	0.012	4.635	(1.396–15.386)
Dependent IADL	<0.001	6.160	(2.536–14.959)	<0.001	6.247	(2.749–14.195)	<0.001	3.541	(2.015–6.223)
Comorbidities
HYP	0.999	1.001	(0.418–2.393)	0.252	1.662	(0.696–3.966)	0.724	0.900	(0.502–1.615)
GID	0.671	0.818	(0.324–2.066)	0.560	0.769	(0.318–1.860)	0.285	1.387	(0.761–2.529)
DEP	0.347	1.529	(0.631–3.703)	0.619	1.239	(0.531–2.891)	0.488	1.235	(0.680–2.246)
DM	0.104	2.177	(0.853–5.556)	0.051	2.411	(0.995–5.840)	0.085	1.743	(0.926–3.282)
NEURO	0.009	6.228	(1.576–24.608)	0.004	8.170	(1.937–34.466)	0.098	2.855	(0.824–9.897)
Constant.	0.003	0.032	–	<0.001	0.012	–	<0.001	0.025	–

Comparative Risk Factor based on different MMSE cut-off points. (–) This parameter is set to zero because it is redundant. It is the reference class of the parameters studied. OR, odds ratio; CI, confidence interval; ApoE ɛ2, ɛ3, ɛ4, Apolipoprotein E allele ɛ2, ɛ3, ɛ4 carriers, respectively; Katz, Katz Index; IADL, Instrumental Activities Daily Life; HYP, arterial hypertension; GID, gastrointestinal disease; DEP, depression; DM, diabetes mellitus; NEURO, neurologic disease. NOTE: OR represents risk or protective values in each of the parameters studied. CI is the confidence interval of OR. If OR = 1 the CI is not statistically significant. If OR = 0 there is no CI, as for ApoE ɛ3 carriers in MMSE PT.

Consequently, a receiver operating characteristic curve was employed to check the model’s ability (with respect to the distinct MMSE cut-offs) in identifying patients with cognitive impairment (Fig. 2). The value of the area of this curve is used to evaluate the discriminant power of the regression model. It is clear from this analysis that the increasing discriminatory capacity for the three MMSE cut-offs is MMSE PT, MMSE 24, and MMSE 27, respectively. In other words, comparatively, the MMSE 27 cut-off has the greatest sensibility (94.9%) and specificity (66.3%) when compared to MMSE PT and MMSE 24 (Fig. 2), in the pcb-Cohort group.

Comparison of MMSE cut-offs with CDR scores

Relevant to the volunteers themselves, is an individual’s potential to be appropriately diagnosed independent of the MMSE cut-off. As such the MMSE cut-offs were compared with the aid of a Venn diagram (Fig. 3A). The MMSE PT, MMSE 24, and MMSE 27 identified 54, 73 and 132 positive cases, respectively. Within test pool 51, individuals (Fig. 3A) were always diagnosed with dementia, however in the MMSE PT, two more volunteers scored positive only when this cut-off was applied. Further, it is clear that the MMSE PT does not identify many of the putative dementia cases identified when the MMSE 24 or the MMSE 27 are applied. As expected, the latter identified far more putative dementia cases, but the MMSE 24 is clearly a subgroup, as all the individuals scoring positive in the latter fall within the MMSE 27 cut-off.

Fig.3

Overlap between MMSE and CDR. The overlap in the number of individuals scoring positive between the different MMSE cut-offs (A) is shown and the overlap between MMSE cut-offs and CDR positive scores is also indicated (B). The number of individuals in each of the classes is indicated in the Venn diagrams.

Another test commonly used for dementia diagnosis is the CDR test. The pcb-Cohort was also evaluated with the CDR test and individuals scored as ≥0.5 or ≥1 [37]. Subsequently, individuals scoring positive for cognitive deficits, across the different tests, were all compared (Fig. 3B). The total pool size of putative dementia cases is largely different for each of the tests, namely 267 individuals for CDR ≥0.5 and 132 individuals for the MMSE 27. That is, in this study, the CDR ≥0.5 identifies twice as many putative dementia cases when compared to the MMSE 27. Independent of the test applied 36 individuals always scored positive for dementia (Fig. 3B). In contrast a large variability is evident, in what one can assume to be putative dementia cases. It is encouraging that there is some overlap between MMSE and CDR score, for instance: 27 individuals positive in the MMSE 27 and CDR ≥0.5; 10 individuals positive in the MMSE 27, CDR ≥0.5, and CDR ≥1; 5 individuals positive in the MMSE 27, MMSE 24, CDR ≥0.5 and CDR ≥1 and 14 individuals positive in the MMSE 27, MMSE 24, and CDR ≥0.5. However, 21 individuals scored positive only with the MMSE 27 and 146 by the CDR ≥0.5. Thus it would appear that the ideal situation would be for individuals to be evaluated by applying the MMSE 27 (which is more sensitive and specific), as well as the CDR cognitive tests (Fig. 3B).

DISCUSSION

For the three distinct MMSE cut-off points, good correlations were obtained with respect to age, monthly income, and educational level. This is consistent with previous findings [42]. Analyses of correlations with comorbidities, with respect to age groups, showed greater consistency with MMSE PT and MMSE 27 cut-offs. In summary, DM appears to be a good predictive factor for dementia, particularly in the age group 65–74 years old and likewise for depression, particularly in the most elderly (more than 75 years old). The relevance of the above-mentioned characteristics as putative factors contributing to dementia was confirmed when multivariate analysis was applied; consistent with this study, DM has previously been indicated as a risk factor for dementia and also AD [43 –45]. Thus, one should perhaps consider testing putative dementia cases for DM or even evaluating patients with diabetes for dementia. In fact it is perhaps not surprising that AD has been referred to as diabetes type 3 [46].

Taken together the data here presented revealed MMSE 27 to be a more sensitive and specific cut-off point when compared to the other two MMSE cut-off points tested, this was confirmed by the ROC curve. In contrast, and based on the Venn diagram analysis of the MMSE tests; the MMSE PT diagnosed fewer potential dementia cases, and care should be taken when applying this test in a clinical setting. Care should also be observed when the standard international cut-offs are not employed. When applying the MMSE to population studies it is important to be aware of these limitations, but these tests are nonetheless useful screening tools. For instance, in another study, based on a MMSE score of ≤23 point (comparative to the MMSE 24 used in the study here presented), 7.7% of the Australian elderly population screened positive for cognitive impairment, indicating putative dementia [42]. Applying the MMSE in these settings identifies individuals who should be tagged for follow up studies in the possibility of being potential dementia cases. However, it should be noted that putative dementia cases might be missed when only the MMSE is applied as a diagnostic tool.

Comparison of the MMSE with the CDR test revealed the former to be less sensitive. Even so, 21 putative dementia cases were identified in the MMSE 27 but missed with the CDR. In contrast the CDR ≥0.5 identified 146 putative dementia cases, which were not at all recognized as putative dementia in the other tests applied. Hence it is reasonable to deduce that in an ideal situation both cognitive tests should be carried out. It is also noteworthy that when the CDR data is evaluated on its own [37], gastrointestinal disease is a putative risk factor in dementia for the pcb-Cohort. This was not identified when the MMSE data is analyzed because the above-mentioned gastrointestinal disease association lies in the 146 individuals not identified in the MMSE analysis but were so in the CDR ≥0.5. Although not a direct diagnosis, the CDR has been shown to be a reliable tool in AD diagnosis. It is also particularly noteworthy that Perneczky et al. [12] showed MMSE to be a useful CDR surrogate measure when staging AD related dementia.

Conclusion

The MMSE and CDR result in useful predictive scores for dementia using low cost interventions. Further, given that the tests are independent of technical aids, they may be used for large-scale preliminary screening programs. This work clearly showed that different MMSE cut-offs result in identifying distinct ‘at risk’ groups and the CDR identifies a higher number of possible dementia cases. This should be borne in mind when applying these cognitive tests.

Risk factors in this population include, age, literacy, hypertension, depression, and DM. Thus, in this Portuguese primary care based population (the pcb-Cohort); few studies of this type have been carried out in Portugal, risk factors for dementia previously identified are sustained. The study also demonstrated the usefulness of carrying out internationally validated cognitive testing in a younger population, for instance correlation of younger DM patients with cognitive deficits was evident and deserves follow up studies. In closing, the results here presented are relevant and should have clinical impact, contributing to the diagnosis of cognitive disorders and in identifying putative dementia cases in a primary care setting.

Footnotes

ACKNOWLEDGMENTS

We would like to thank the volunteers and their families as well as all the health professionals across the health care centers in the Aveiro district who made this study possible. This work was financed by PTDC/DTP-PIC/5587/2014 and supported by UID/BIM/04501/2013, iBiMED, University of Aveiro and the Fundação para a Ciência e Tecnologia of the Ministério da Educação e Ciência, and by the project JPND/0006/2011-BIOMARKAPD.

Authors’ disclosures available online ().

The supplementary material is available in the electronic version of this article: .

References

Henriques

, Müller

, Oliveira

, Cova

, da Cruz e Silva

OAB

(2016) Altered protein phosphorylation as a resource for potential AD biomarkers. Sci Rep 6, 30319.

Morris

(1993) The Clinical Dementia Rating (CDR): Current version and scoring rules. Neurology 43, 2412–2414.

Perneczky

(2003) The appropriateness of short cognitive tests for the identification of mild cognitive impairment and mild dementia. Aktuel Neurol 30, 114–117.

Petersen

(2000) Mild cognitive impairment or questionable dementia? Arch Neurol 57, 643–644.

Hänninen

, Hallikainen

, Tuomainen

, Vanhanen

, Soininen

(2002) Prevalence of mild cognitive impairment: A population-based study in elderly subjects. Acta Neurol Scand 106, 148–154.

, Ng

, Kua

, Ko

(2006) Brief informant screening test for mild cognitive impairment and early Alzheimer’s disease. Dement Geriatr Cogn Disord 21, 392–402.

McClendon

, Hernandez

, Smyth

, Lerner

(2009) Memantine and acetylcholinesterase inhibitor treatment in cases of CDR 0.5 or questionable impairment. J Alzheimers Dis 16, 577–583.

Duara

, Loewenstein

, Greig-Custo

, Raj

, Barker

, Potter

, Schofield

, Small

, Schinka

, Wu

, Potter

(2010) Diagnosis and staging of mild cognitive impairment, using a modification of the clinical dementia rating scale: The mCDR. Int J Geriatr Psychiatry 25, 282–289.

Chapman

, Bing-Canar

, Alosco

, Steinberg

, Martin

, Chaisson

, Kowall

, Tripodis

, Stern

(2016) Mini Mental State Examination and Logical Memory scores for entry into Alzheimer’s disease trials. Alzheimers Res Ther 8, 9.

10.

Bowen

, Perry

, Xiong

, Smith

, Atwood

(2015) A clinical study of lupron depot in the treatment of women with Alzheimer’s disease: Preservation of cognitive function in patients taking an acetylcholinesterase inhibitor and treated with high dose lupron over 48 weeks. J Alzheimers Dis 44, 549–560.

11.

Folstein

, Folstein

, McHugh

(1975) “Mini-mental state”. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12, 189–198.

12.

Perneczky

, Wagenpfeil

, Komossa

, Grimmer

, Diehl

, Kurz

(2006) Mapping scores onto stages: Mini-Mental State Examination and Clinical Dementia Rating. Am J Geriatr Psychiatry 14, 139–144.

13.

Beyermann

, Trippe

, Bähr

, Püllen

(2013) Mini-Mental-Status-Test im stationären geriatrischen Bereich: Eine Evaluation der diagnostischen Qualität. Z Gerontol Geriatr 46, 740–747.

14.

Eichler

, Thyrian

, Hertel

, Michalowsky

, Wucherer

, Dreier

, Kilimann

, Teipel

, Hoffmann

(2015) Rates of formal diagnosis of dementia in primary care: The effect of screening. Alzheimers Dement (Amst) 1, 87–93.

15.

Davey

, Jamieson

(2004) The validity of using the Mini Mental State Examination in NICE dementia guidelines. J Neurol Neurosurg Psychiatry 75, 343–344.

16.

Shulman

, Herrmann

, Brodaty

, Chiu

, Lawlor

, Ritchie

, Scanlan

(2006) IPA survey of brief cognitive screening instruments. Int Psychogeriatr 18, 281–294.

17.

Reilly

, Challis

, Burns

, Hughes

(2004) The use of assessment scales in Old Age Psychiatry Services in England and Northern Ireland. Aging Ment Health 8, 249–255.

18.

Petersen

, Smith

, Waring

, Ivnik

, Tangalos

, Kokmen

(1999) Mild cognitive impairment: Clinical characterization and outcome. Arch Neurol 56, 303–308.

19.

Ousset

, Andrieu

, Reynish

, Puel

, Vellas

(2003) [Clinical evaluation of dementia in a cohort of 358 patients with the French version of the Clinical Dementia Rating (CDR) scale]. Rev Med Interne 24(Suppl 3), 283s–287s.

20.

Winblad

, Viramo

, Remes

, Manninen

, Jokelainen

(2010) Prevalence of dementia - A rising challenge among ageing populations. Eur Geriatr Med 1, 330–333.

21.

Kaufer

, Williams

, Braaten

, Gill

, Zimmerman

, Sloane

(2008) Cognitive screening for dementia and mild cognitive impairment in assisted living: Comparison of 3 tests. J Am Med Dir Assoc 9, 586–593.

22.

Pendlebury

, Mariz

, Bull

, Mehta

, Rothwell

(2012) MoCA, ACE-R, and MMSE versus the National Institute of Neurological Disorders and Stroke-Canadian Stroke Network Vascular Cognitive Impairment Harmonization Standards Neuropsychological Battery after TIA and stroke. Stroke 43, 464–469.

23.

Tombaugh

, McIntyre

(1992) The Mini-Mental State Examination: A comprehensive review. J Am Geriatr Soc 40, 922–935.

24.

Feldman

, Jacova

, Robillard

, Garcia

, Chow

, Borrie

, Schipper

, Blair

, Kertesz

, Chertkow

(2008) Diagnosis and treatment of dementia: 2. Diagnosis. CMAJ 178, 825–836.

25.

DeKosky

, Fitzpatrick

, Ives

, Saxton

, Williamson

, Lopez

, Burke

, Fried

, Kuller

, Robbins

, Tracy

, Woolard

, Dunn

, Kronmal

, Nahin

, Furberg

, GEMS Investigators (2006) The Ginkgo Evaluation of Memory (GEM) study: Design and baseline data of a randomized trial of Ginkgo biloba extract in prevention of dementia. Contemp Clin Trials 27, 238–253.

26.

Mitchell

(2009) A meta-analysis of the accuracy of the Mini-Mental State Examination in the detection of dementia and mild cognitive impairment. J Psychiatr Res 43, 411–431.

27.

Glasser

(1993) Alzheimer’s disease and dementing disorders: Practices and experiences of rural physicians. Am J Alzheimers Dis Other Demen 8, 28–35.

28.

Brodaty

, Howarth

, Mant

, Kurrle

(1994) General practice and dementia. A national survey of Australian GPs. Med J Aust 160, 10–14.

29.

Mossello

, Boncinelli

(2006) Mini-Mental State Examination: A 30-year story. Aging Clin Exp Res 18, 271–273.

30.

Blesa

, Pujol

, Aguilar

, Santacruz

, Bertran-Serra

, Hernández

, Sol

, Peña-Casanova

, NORMACODEM Group, NORMAlisation of Cognitive and Functional Instruments for DEMentia (2001) Clinical validity of the “mini-mental state” for Spanish speaking communities. Neuropsychologia 39, 1150–1157.

31.

Plassman

, Germain

, Hayden

, Potter

, Romero

, Welsh-Bohmer

, Burke

(2013) Misdiagnosis using brief cognitive screening instruments in a population representative sample. Alzheimers Dement 9, P784–P785.

32.

Morgado

, Rocha

, Maruta

, Guerreiro

, Martins

(2009) Novos valores normativos do Mini-Mental State Examination. Sinapse 9, 10–16.

33.

Morgado

, Rocha

, Maruta

, Guerreiro

, Martins

(2010) Cut-off scores in MMSE: A moving target? Eur J Neurol 17, 692–695.

34.

Schulz

, Grimes

(2005) Sample size calculations in randomised trials: Mandatory and mystical. Lancet 365, 1348–1353.

35.

Rosa

, Cavaleiro

, Ferreira

, Cova

, Afreixo

, da Cruz e Silva

OAB

(2013) Cognitive characterization of a pilot study group. Acta Gerontol (Milano) 1, 195.

36.

Rosa

, Henriques

, da Cruz E Silva

OAB

(2015) Cognitive performance of primary care based cohort using Mini Mental State Examination. In Congresso de Neurologia 2015, At Lisboa, Volume: 15.

37.

Rosa

, Henriques

, Oliveira

, da Cruz E Silva

OAB

(2017) Screening younger individuals in a primary care setting flags putative dementia cases and correlates gastrointestinal diseases with poor cognitive performance. Dement Geriatr Cogn Disord 43, 15–28.

38.

Heberle

, Meirelles

, da Silva

, Telles

, Minghim

(2015) InteractiVenn: A web-based tool for the analysis of sets through Venn diagrams. BMC Bioinformatics 16, 169.

39.

Verghese

, Castellano

, Holtzman

(2011) Apolipoprotein E in Alzheimer’s disease and other neurological disorders. Lancet Neurol 10, 241–252.

40.

Dufouil

, Richard

, Fiévet

, Dartigues

, Ritchie

, Tzourio

, Amouyel

, Alpérovitch

(2005) APOE genotype, cholesterol level, lipid-lowering treatment, and dementia: The Three-City Study. Neurology 64, 1531–1538.

41.

Chaudhry

, Hasnain

, Snitz

, Wang

, Rosenthal

, Demirci

, Kamboh

(2014) Association of APOE polymorphisms and stressful life events with dementia in a Pakistani population. Neurosci Lett 570, 42–46.

42.

Anderson

, Sachdev

, Brodaty

, Trollor

, Andrews

(2007) Effects of sociodemographic and health variables on Mini-Mental State Exam scores in older Australians. Am J Geriatr Psychiatry 15, 467–476.

43.

Baumgart

, Snyder

, Carrillo

, Fazio

, Kim

, Johns

(2015) Summary of the evidence on modifiable risk factors for cognitive decline and dementia: A population-based perspective. Alzheimers Dement 11, 718–726.

44.

Biessels

, Staekenborg

, Brunner

, Brayne

, Scheltens

(2006) Risk of dementia in diabetes mellitus: A systematic review. Lancet Neurol 5, 64–74.

45.

Peila

, Rodriguez

, Launer

; Honolulu-Asia Aging Study (2002) Type 2 diabetes, APOE gene, and the risk for dementia and related pathologies: The Honolulu-Asia Aging Study. Diabetes 51, 1256–1262.

46.

de la Monte

(2014) Type 3 diabetes is sporadic Alzheimer’s disease: Mini-review. Eur Neuropsychopharmacol 24, 1954–1960.