Neurol Sci DOI 10.1007/s10072-014-1711-y

ORIGINAL ARTICLE

Clustering mild cognitive impairment by mini-mental state examination So Young Kim • Tae Sung Lim • Hyun Young Lee So Young Moon



Received: 5 August 2013 / Accepted: 5 March 2014 Ó Springer-Verlag Italia 2014

Abstract We aimed to evaluate whether the performance of the mini-mental state examination (MMSE) could identify risky mild cognitive impairment (MCI). We recruited 122 amnestic MCI-single domain (ASM), 303 amnestic MCI-multiple domains (AMM), and 94 nonamnestic MCI (NAM). Two-step cluster and linear discriminant analyses were used for identifying the clusters of the MMSE with age and education, as well as establishing prediction models for each cluster. Conversion into dementia was compared among clusters. Cluster analyses revealed the following three: cluster 1 = 205 AMM (100 %); cluster 2 = 61 NAM (33.3 %) and 122 ASM (66.7 %); and cluster 3 = 33 NAM (25.2 %) and 98 AMM (74.8 %). Cluster 3 showed a significantly lower ability with regards to orientation to time and place, registration of three words, attention/calculation, language, and copying interlocking pentagons, than clusters 1 and 2. However, for delayed recall, cluster 1 was significantly more impaired than cluster 2. Patients in the cluster 1 showed the most common conversion into dementia [odds ratio (OR) = 2.940 vs. cluster 2, OR = 2.271 vs. cluster 3].

S. Y. Kim  T. S. Lim  S. Y. Moon (&) Department of Neurology, School of Medicine, Ajou University, 5 San, Woncheon-dong, Yongtong-gu, Suwon-si, Kyunggi-do 442-749, Republic of Korea e-mail: [email protected] H. Y. Lee Regional Clinical Trial Center, Ajou University Medical Center, Suwon, Republic of Korea H. Y. Lee Department of Biostatistics, Yonsei University College of Medicine, Seoul, Republic of Korea

This study showed that clustering by performance in MMSE could help define groups at higher risk for conversion to dementia. Therefore, MMSE can be considered as a promising screening tool including subtyping for MCI when detailed neuropsychological tests are not feasible. Keywords Mild cognitive impairment  Alzheimer’s disease  Mini-mental state examination  Amnestic  Nonamnestic  Clusters

Introduction Mild cognitive impairment (MCI) is a disorder that has been associated with risk for dementia [1]. By defining the subtype of MCI, the clinician can make a reliable prediction regarding the outcome of the MCI syndrome. Although detailed neuropsychological testing is the most desirable tool to define the subtypes of MCI, it is often not feasible to be performed by primary care physicians or in community studies. However, few studies attempted to evaluate the performances of more feasible tests for defining MCI subtypes. The mini-mental state examination (MMSE) is the most widely used cognitive screening tool [2]. However, it is controversial on the utility of MMSE in MCI patients. Whereas some studies showed a good diagnostic accuracy and predictive value of MMSE [3–6], other reports did not suggest MMSE as screening instrument for MCI [7, 8]. In our study, we aimed to compare the pattern of cognitive deficits on the MMSE with subtypes of MCI. In addition, we attempted to evaluate whether understanding the natural structure of MMSE performance by MCI group can be used to identify those at greater risk for converting to dementia.

123

Neurol Sci

Methods Patients From November 2005 to December 2008, we recruited 519 patients who had been newly diagnosed with MCI from a nationwide multicenter study of dementia, the Clinical Research for Dementia of South Korea (CREDOS) study. Since 2005, the CREDOS study had aimed to understand the characteristics of Korean patients with dementia. It had developed common protocols, including its own ischemic scale, and recruited patients with subjective memory impairment, MCI, MCI of a subcortical vascular type, AD, and subcortical vascular dementia. A total of 31 university and general hospitals in South Korea participated in this study. Details of diagnostic evaluation of patients and inclusion/exclusion criteria were published previously [9]. The MCI patients were diagnosed according to the criteria proposed by Peterson et al. [1]. This study was approved by the Institutional Review Boards of all participating hospitals, and written informed consent was obtained from patients and their caregivers after receiving a complete description of the study. Classification of MCI On the basis of the results of the Seoul Neuropsychological Screening Battery [10], patients were divided into three groups: amnestic MCI-single domain (ASM), amnestic MCI-multiple domains (AMM), or non-amnestic MCI (NAM). Classification was based on four cognitive domains: memory, language, visuospatial, and frontal functions. The scores on cognitive tests were classified as abnormal when they were below the 16th percentile of the norms for the age-, sex-, and education-matched normal subjects. Memory function was evaluated by the delayed recall on the Seoul Verbal Learning Test or Rey–Osterrieth Complex Figure Test (RCFT). Language was assessed by the Korean version of the Boston Naming Test. Visuospatial function was evaluated by the copying score of the RCFT. Finally, frontal/executive tests were classified into three groups: motor executive function (contrasting program, go/ no-go, fist–edge–palm, alternating hand movement, alternative square and triangle, and Luria loop), Controlled Oral Word Association Test (COWAT), and Stroop Test. Impaired frontal/executive function was operationally defined as impairment in at least two of the three groups. K-MMSE The K-MMSE contains 30 items, with higher scores indicating better cognitive functioning. Items comprise seven areas: orientation to time (5), orientation to place (5),

123

registration of three words (3), attention and calculation (5), delayed recall of three words (3), language (8), and copying interlocking pentagons (1). Following a careful review by experts, who reached a consensus, and a subsequent validation process [11], the K-MMSE is now widely used in Korea. Follow-up Of the 519 MCI patients, 176 (33.9 %) patients completed at least one follow-up visit with the same interview and neuropsychological tests for the baseline evaluation from November 2005 to February 2011. The mean duration of follow-up was 23.9 ± 11.4 months. There was no significant difference in their age, sex, education, and scores of baseline MMSE between patients who performed at least one follow-up neuropsychological test and those who did not. The diagnosis of dementia was based on criteria from the Diagnostic and Statistical Manual of Mental Disorders (4th edition) and required clinical evidence of cognitive deficits confirmed by neuropsychological tests, as well as evidence of impairment in social or occupational functions confirmed by activities of daily living scales. We determined the presence of dementia by using DSM-IV because our analyses did not include specific causes of dementia. Among 176 MCI subjects, 61 (34.7 %) progressed onto dementia and 18 (10.2 %) reverted to a state of normal cognition. Data analysis and statistics First, we compared patient demographic findings and performance on the K-MMSE among three subtypes of MCI. Pearson’s Chi-square (v2) test was used to examine trends in categorical data and the analysis of variance (ANOVA) was used for continuous variables with the Tukey method. For comparing the total K-MMSE scores or score of each domain across the three groups, we used the univariate general linear model with adjustments for age and education. Ability in relation to the K-MMSE was obtained with a fraction of an amount expressed as a particular score of hundredths of that amount. And then, we attempted to identify the natural structure (clusters) based on multivariate profiles including education, age, and seven domain scores of K-MMSE by two-step cluster analysis. In the first step, the algorithm performed a procedure that creates pre-clusters. Based on these results, the second step conducts a modified hierarchical agglomerative clustering procedure that combines the objects sequentially to form homogenous clusters. The number of clusters was determined as part of the clustering algorithm. For predicting each cluster, linear discriminant analysis was applied, and two discriminant equations were obtained. To verify the predicting ability of the two equations, sensitivity and specificity were evaluated.

Neurol Sci Table 1 Demographic characteristics and scores of each area in the K-MMSE among subtypes of MCI

Age

Amnestic MCI-single domain (n = 122)

Amnestic MCI-multiple domains (n = 303)

Non-amnestic MCI (n = 94)

66.6 ± 8.1

69.0 ± 7.8

69.2 ± 7.7

P value

Adjusted P value

0.012 ,}

Female, n (%)

79 (65)

200 (66)

70 (74)

0.249

Education

9.3 ± 5.0

8.2 ± 5.0

6.7 ± 5.6

0.001},à

Total K-MMSE scores

26.4 ± 2.5

24.6 ± 3.3

24.8 ± 3.7

\0.001 ,}

Orientation to time (5)

4.2 ± 1.0

4.1 ± 1.0

4.3 ± 0.8

0.392

0.089

Orientation to place (5)

4.8 ± 0.4

4.7 ± 0.5

4.7 ± 0.6

0.059

0.216

Registration of three words (3) Attention and calculation (5)

3.0 ± 0.0 4.0 ± 1.1

2.9 ± 0.2 3.2 ± 1.5

2.9 ± 0.1 3.0 ± 1.7

0.127 \0.001 ,}

0.131 \0.001 ,}

Delayed recall of three words (3)

1.6 ± 1.1

1.4 ± 1.1

1.8 ± 0.9

0.001à

0.001à

Language (8) Copying interlocking pentagons (1)

7.7 ± 0.4 0.9 ± 0.2

7.3 ± 0.9 0.7 ± 0.4

7.2 ± 0.9 0.7 ± 0.4

 ,}

\0.001 ,}

 ,}

\0.001 ,}

\0.001 \0.001

Data for each area except for sex proportion: mean ± SD. Adjusted P values are those P values resulting from comparisons among three groups after adjustment for both age and education. P values \0.05 resulting from comparisons among the three groups have been italicized K-MMSE Korean version of mini-mental state examination, MCI mild cognitive impairment P values \0.05 resulting from post hoc analyses are as follows:   between amnestic MCI-single domain and amnestic MCI-multiple domains, between amnestic MCI-single domain and non-amnestic MCI, à between amnestic MCI-multiple domains and non-amnestic MCI

}

Kaplan–Meier survival analyses with log rank test were used to illustrate the difference in converting to dementia among clusters. Time to the event was defined as the time from study entry to the follow-up visit at which a first-time diagnosis of dementia was made. Subjects that did not convert to dementia were treated as censored observations from the time of their final follow-up evaluation. By the binary logistic regression, we obtained odds ratio (OR) for dementia or normal cognition between clusters. P \ 0.05 was considered significant. All statistical analyses were performed using SPSS version 20.0 (Chicago, IL, USA).

Results Characteristics of the patients Study subjects consisted of 122 patients with ASM, 303 with AMM, and 94 with NAM (Table 1). All patients scored 0.5 on the Clinical Dementia Rating (CDR) scale. Whereas the three groups did not differ in sex, they showed differences in age, education, and total MMSE scores. After adjustments for age and education, the total K-MMSE scores were still significantly higher in the ASM group than in AMM and NAM groups (P \ 0.001). Clusters from 519 subjects according to age, education, and areas of K-MMSE We obtained three clusters. Cluster 1 consisted of 205 patients with AMM (100 %). Cluster 2 was composed of 61 NAM (33.3 %) and 122 ASM (66.7 %). Finally, cluster

Fig. 1 Distribution of clusters according to age, education, and seven area scores of the K-MMSE. Patients were divided into three clusters by two-step cluster analysis according to age, education, and the seven area scores of K-MMSE. Cluster 1 consisted of only amnestic MCI-multiple domains (100 %). Cluster 2 was composed of non-amnestic (33.3 %) and amnestic MCI-single domain (66.7 %), and cluster 3 consisted of non-amnestic (25.2 %) and amnestic MCImultiple domains (74.8 %)

3 consisted of 33 NAM (25.2 %) and 98 AMM (74.8 %) (Fig. 1). Cluster 3 was significantly older and had a lower educational level than clusters 1 and 2 (age 68.2 ± 8.1, 67.0 ± 8.3, and 71.2 ± 6.6 in clusters 1, 2, and 3 respectively, P \ 0.001; education 10.3 ± 4.3, 9.5 ± 4.9, and 3.4 ± 3.7 in clusters 1, 2, and 3 respectively, P \ 0.001). The ability to achieve seven areas of K-MMSE was significantly different among clusters 1, 2, and 3. Cluster 3 showed significantly lower ability for orientation to time and place, registration of three words, attention/calculation,

123

Neurol Sci Fig. 2 Ability to achieve seven areas of K-MMSE among clusters. Difference in ability to achieve seven areas of K-MMSE was significant for recall between clusters 1 and 2, and for orientation to time and place, registration of three words, attention/calculation, language, and copying interlocking pentagons between clusters 1 (or 2) and cluster 3.   P \ 0.05 in cluster 1 vs. cluster 2, àP \ 0.05 in cluster 1 vs. cluster 3, }P \ 0.05 in cluster 2 vs. cluster 3 with post hoc analysis. K-MMSE Korean version of mini-mental state examination

language, and copying interlocking pentagons than clusters 1 and 2 (P \ 0.001; Fig. 2). However, at delayed recall, cluster 1 was significantly more impaired than cluster 2. Predicting each cluster using age, education, and seven area scores of the K-MMSE We obtained two equations for predicting each cluster by linear discriminant analysis using age, education, and seven area scores of the K-MMSE (equation 1 = -0.061 9 education ? 0.017 9 age ? 0.020 9 time - 0.376 9 place 2.891 9 registration - 0.178 9 calculation ? 0.037 9 recall - 0.406 9 language - 3.325 9 pentagon ? 15.717; and equation 2 = 0.560 9 recall ? 1.032 9 language 3.015 9 pentagon - 5.925). Next, we constructed flowcharts for classifying clusters using the two equations (Fig. 3). We obtained 152 patients having positive equation 1 scores of which all 152 were predicted to be in cluster 3. Of these, 129 patients were correctly predicted and 23 patients were incorrectly predicted. The remaining 367 subjects were examined with equation 2. If they had a positive score, they were sorted into cluster 2; otherwise, they were sorted into cluster 1. In this step, 205 patients were correctly categorized, but 162 patients were incorrectly grouped. As a result, we obtained 64.4 % overall accuracy from the two equations. In particular, clusters 1, 2, and 3 obtained 60.0, 44.8, and 98.5 % sensitivity and 72.0, 78.0, and 94.1 % specificity, respectively.

123

Fig. 3 Flowchart of predicting each cluster using age, education, and seven area scores of the K-MMSE. Using two equations including age, education, and seven area scores of the K-MMSE, 64.4 % overall accuracy was obtained. K-MMSE Korean version of mini-mental state examination

Conversion to dementia among clusters There was no significance in the mean duration of followup among clusters (Table 2). The conversion into dementia was revealed most commonly in the cluster 1 (OR = 2.940 vs. cluster 2, OR = 2.271 vs. cluster 3) whereas reversion

Neurol Sci Table 2 Follow-up results among clusters Cluster 1 (n = 72)

Cluster 2 (n = 58)

Cluster 3 (n = 46)

P value

25.6 ± 12.2

24.1 ± 11.9

21.9 ± 10.2

0.244

Normal cognition, n (%)

4 (5.6)

12 (20.0)

2 (4.3)

0.002

Mild cognitive impairment, n (%)

34 (47.2)

34 (56.7)

31 (67.4)

Dementia, n (%)

34 (47.2)

14 (23.3)

13 (28.3)

Duration of follow-up (months) Diagnosis at follow-up

Table 3 Relative risks of conversion to dementia and reversion to normal between clusters B

SE

P value

Odds ratio (95 % CI)

Cluster 1 vs. 2

1.078

0.386

0.005

2.940 (1.38, 6.263)

Cluster 1 vs. 3

0.820

0.404

0.042

2.271 (1.030, 5.010)

MCI to dementia

MCI to normal cognition Cluster 2 vs. 1

1.447

0.607

0.017

4.250 (1.292, 13.975)

Cluster 2 vs. 3

1.705

0.792

0.031

5.500 (1.165, 25.961)

into normal cognition occurred most commonly in the cluster 2 (OR = 4.250 vs. cluster 1, OR = 5.500 vs. cluster 3; Table 3). However, survival analyses showed no significant difference among clusters (P [ 0.05).

Discussion Our study showed that we isolated the natural structure in three clusters for 519 patients with MCI according to age, education, and seven area scores of the MMSE and established prediction models for each cluster. In addition, although only part of the patients was followed up, the longitudinal observation revealed that the most risky group for dementia was the cluster 1 which consisted of 100 % of AMM. Recent studies showed that patients who died while their clinical classification was amnestic, MCI had neuropathologic features intermediate between changes of normal aging and Alzheimer’s disease (AD) [12]. Therefore, it is meaningful to differentiate between amnestic MCI and NAM. Our results showed that the K-MMSE could help distinguish between amnestic MCI and NAM similar to the delayed memory test in the quick cognitive screening test (QSCT) [13]. Patients who have greater memory impairment and probably those who have AMM will progress more rapidly than those with less severe or single domain impairment. A recent study indicated that those patients with AMM actually had worse survival rates than patients with ASM [14]. Therefore, it is clinically important to

identify patients with AMM. The cluster analysis in our study showed that patients were sorted into three clusters by age, education, and seven area scores of the K-MMSE. Cluster 1 consisted totally of AMM with more impaired performance in delayed recall than the cluster 2. Considering previous studies which showed that patients who have greater memory impairment and probably those who have AMM will progress more rapidly than those with less severe or single domain impairment; cluster 1 may represent a more risky group for conversion to dementia, which was shown by our follow-up study. Cluster 3, which consisted of AMM and NAM, was the oldest in patient age and lowest in education level. Except for delayed recall, they showed the worst performance in all remaining areas. Therefore, cluster 3 also appeared to have more risks for converting to dementia. However, our longitudinal observation showed that the conversion rate into dementia in the cluster 3 was similar with one in the cluster 2 but lower than one in the cluster 1. The worst performance of the K-MMSE by the cluster 3 seemed to be affected by their age and education but not related to the risk for the conversion to the dementia. Previous reports in the literature suggested that widely used cognitive screening tests for detecting MCI were the Montreal Cognitive Assessment (MoCA) [15] and QCST [13, 14, 16]. A recent study showed that the Consortium to Establish a Registry for Alzheimer Disease neuropsychological battery (CERAD–NP) also has modest discrimination ability between normal controls and MCI [17]. On the other hand, although the MMSE is the most widely used cognitive screening tool, it has been controversial when used for detecting MCI. However, a recent study suggested that only variables independently associated with MCI in medical inpatients were considered in the total score of the MMSE [3]. In addition, our study showed that performance in the K-MMSE by subtype of MCI reflected the original definition well. Clustering by performance in the K-MMSE could help define more risky groups for conversion to dementia. Therefore, the MMSE can be considered as a promising screening tool including subtyping for MCI when detailed neuropsychological tests are not feasible.

123

Neurol Sci

This study had several limitations. First, although our follow-up study showed the relative risk to dementia among clusters, the survival analyses did not reveal the significant difference. Second, two equations for predicting each cluster by linear discriminant analysis actually had low accuracy. Acknowledgments This study was supported by a grant of the Korea Healthcare technology R&D Project, Ministry of Health and Welfare, Republic of Korea (HI10C2020).

References 1. Petersen RC, Smith GE, Waring SC, Ivnik RJ, Tangalos EG, Kokmen E (1999) Mild cognitive impairment: clinical characterization and outcome. Arch Neurol 56:303–308 2. Folstein MF, Folstein SE, McHugh PR (1975) ‘‘Mini-mental state’’. A practical method for grading the cognitive state of patients for the clinician. J Psychiatr Res 12:189–198 3. De Marchis GM, Foderaro G, Jemora J, Zanchi F, Altobianchi A, Biglia E, Conti FM, Monotti R, Mombelli G (2010) Mild cognitive impairment in medical inpatients: the mini-mental state examination is a promising screening tool. Dement Geriatr Cogn Disord 29:259–264 4. Moretti F, De Ronchi D, Palmer K, Forlani C, Morini V, Ferrari B, Dalmonte E, Atti AR (2013) Prevalence and characteristics of mild cognitive impairment in the general population. Data from an Italian population-based study: the Faenza project. Aging Ment Health 17:267–275 5. Palmqvist S, Hertze J, Minthon L, Wattmo C, Zetterberg H, Blennow K, Londos E, Hansson O (2012) Comparison of brief cognitive tests and CSF biomarkers in predicting Alzheimer’s disease in mild cognitive impairment: six-year follow-up study. PLoS One 7:e38639 6. Xie H, Mayo N, Koski L (2011) Predictors of future cognitive decline in persons with mild cognitive impairment. Dement Geriatr Cogn Disord 32:308–317 7. Mitchell AJ (2009) A meta-analysis of the accuracy of the minimental state examination in the detection of dementia and mild cognitive impairment. J Psychiatr Res 2009(43):411–431

123

8. Wind AW, Schellevis FG, Van Staveren G, Scholten RP, Jonker C, Van Eijk JT (1997) Limitations of the mini-mental state examination in diagnosing dementia in general practice. Int J Geriatr Psychiatry 12:101–108 9. Moon SY, Na DL, Seo SW, Lee JY, Ku BD, Kim SY, Park KW, Shim YS, Youn YC, Chung CS, Cheong HK, Choi SH, Cha KR, Kim JE, Jeong JH (2011) Impact of white matter changes on activities of daily living in mild to moderate dementia. Eur Neurol 65:223–230 10. Ahn HJ, Chin J, Park A, Lee BH, Suh MK, Seo SW, Na DL (2010) Seoul neuropsychological screening battery-dementia version (SNSB-D): a useful tool for assessing and monitoring cognitive impairments in dementia patients. J Korean Med Sci 25:1071–1076 11. Oh E, Kang Y, Shin JH, Yeon BK (2010) A validity study of K-MMSE as a screening test for dementia: comparison against a comprehensive neuropsychological evaluation. Dement Neurocogn Disord 9:8–12 12. Petersen RC, Parisi JE, Dickson DW, Johnson KA, Knopman DS, Boeve BF, Jicha GA, Ivnik RJ, Smith GE, Tangalos EG, Braak H, Kokmen E (2006) Neuropathologic features of amnestic mild cognitive impairment. Arch Neurol 63:665–672 13. Guo QH, Cao XY, Zhou Y, Zhao QH, Ding D, Hong Z (2010) Application study of quick cognitive screening test in identifying mild cognitive impairment. Neurosci Bull 26:47–54 14. Hunderfund AL, Roberts RO, Slusser TC, Leibson CL, Geda YE, Ivnik RJ, Tangalos EG, Petersen RC (2006) Mortality in amnestic mild cognitive impairment: a prospective community study. Neurology 67:1764–1768 15. Nasreddine ZS, Phillips NA, Bedirian V, Charbonneau S, Whitehead V, Collin I, Cummings JL, Chertkow H (2005) The Montreal cognitive assessment, MoCA: a brief screening tool for mild cognitive impairment. J Am Geriatr Soc 53:695–699 16. Mate-Kole CC, Conway J, Catayong K, Bieu R, Sackey NA, Wood R, Fellows R (2009) Validation of the revised quick cognitive screening test. Arch Phys Med Rehabil 90:1469–1477 17. Seo EH, Lee DY, Lee JH, Choo IH, Kim JW, Kim SG, Park SY, Shin JH, Do YJ, Yoon JC, Jhoo JH, Kim KW, Woo JI (2010) Total scores of the CERAD neuropsychological assessment battery: validation for mild cognitive impairment and dementia patients with diverse etiologies. Am J Geriatr Psychiatry 18:801–809

Clustering mild cognitive impairment by mini-mental state examination.

We aimed to evaluate whether the performance of the mini-mental state examination (MMSE) could identify risky mild cognitive impairment (MCI). We recr...
421KB Sizes 0 Downloads 0 Views