Send Orders for Reprints to
[email protected] Current Alzheimer Research, 2017, 14, 1-12
1
RESEARCH ARTICLE
Predicting Stability of Mild Cognitive Impairment (MCI): Findings of a Community Based Sample Sinika Ellendt1, Bianca Voß1, Nils Kohn1,3, Lisa Wagels1, Katharina S. Goerlich1, Eva Drexler1, Frank Schneider1,2 and Ute Habel1,2,* 1
Department of Psychiatry, Psychotherapy and Psychosomatics, Medical School, RWTH Aachen, University, Aachen, Germany; 2JARA - Translational Brain Medicine, Aachen, Germany; 3Donders Institute, Centre for Cognitive Neuroimaging, Nijmegen, Netherlands Abstract: Background: Mild Cognitive Impairment (MCI) is a risk factor for Alzheimer’s disease (AD) and other forms of dementia. However, much heterogeneity concerning neuropsychological measures, prevalence and progression rates impedes distinct diagnosis and treatment implications. Objective: Aim of the present study was the identification of specific tests providing a high certainty for stable MCI and factors that precipitate instability of MCI in a community based sample examined at three measurement points.
ARTICLE HISTORY Received: April 27, 2016 Revised: November 20, 2016 Accepted: December 01, 2016 DOI: 10.2174/15672050146661612131 20807
Method: 130 participants were tested annually with an extensive test battery including measures of memory, language, executive functions, intelligence and dementia screening tests. Exclusion criteria at baseline comprised, severe cognitive deficits (e.g. diagnosis of dementia, psychiatric or neurological disease). Possible predictors for stability or instability of MCI-diagnosis were analyzed using Regression and Receiver Operating Characteristic (ROC) curve analysis. Age, IQ and APOE status were tested for moderating effects on the interaction of test performances and group membership. Results: A high prevalence of MCI (49%) was observed at baseline with a reversion rate of 18% after two years. Stability of MCI was related to performances in four measures (VLMT: delayed recall, CERAD: recall drawings, CERAD: Boston Naming Test, Benton Visual Retention Test: number of mistakes). Conversion to MCI is associated with language functions. Reversion to ‘normal’ was primarily predicted by single domain impairment. There was no significant influence of demographic, medical or genetic variables. Conclusion: The results highlight the role of repeated measurements for a reliable identification of functional neuropsychological predictors and better diagnostic reliability. In cases of high uncertainty close monitoring over time is needed in order of estimating outcome.
Keywords: Mild Cognitive Impairment (MCI), Alzheimer Dementia (AD), cognitive functions, memory, longitudinal survey, neuropsychology. 1. INTRODUCTION The life expectancy of the world’s population is constantly rising and it is estimated that the proportion of people over the age of 60 will double and grow extensively to 115.4 million until the year 2050 [1]. Accordingly, the number of people affected by age-related diseases, such as neurodegenerative disorders like Alzheimer’s disease (AD) will increase and require vast resources in order to provide adequate treatment and care. In 2007, a study by Brookmeyer et al. estimated that 1 in 85 persons will live with AD by the year 2050 [2]. Currently, no cure for AD exists and medical *Address correspondence to this author at the Department of Psychiatry, Psychotherapy and Psychosomatics, Medical School, RWTH Aachen University, Aachen, Germany; Tel: +49 (0)241 80 80386; Fax: +49 (0)241 80 82401; E-mail:
[email protected];
[email protected] 1567-2050/17 $58.00+.00
therapy is limited to symptomatic treatment with oftentimes only low to moderate benefits [3, 4]. Because of its threatening characteristics, an increasing fear of AD in the older population is not surprising [5], leading to a danger of misinterpreting age-appropriate symptoms like mild forgetfulness or confusion as beginning AD. But which symptoms actually are predictive for the risk of dementia? Attempts of identifying and defining the gray zone between age-appropriate cognitive functioning and dementia are not new. In the early 1960’s, Kral identified a concept named ‘Malignant Senescent Forgetfulness’ (MSF) in order to describe early pathological memory decline with ageing [6]. Many alternative and revised concepts have been proposed since then, though most with insufficient ability to distinguish between normal ageing and the risk of developing a neurodegenerative disorder [7]. © 2017 Bentham Science Publishers
2 Current Alzheimer Research, 2017, Vol. 14, No. 6
One of the first constructs successfully identifying persons with an elevated risk for dementia, specifically for AD, was ‘Mild Cognitive Impairment’ (MCI) [8]. Originally, the criteria for MCI focused primarily on memory complaints, yet further research suggested that other cognitive domains are involved, too [7, 9]. After thorough debate and revision, MCI is now understood as a decline of cognitive functions greater than expected for normally ageing people, but not large enough for a diagnosis of dementia and without a significant impact on daily functioning [7, 10]. Today, the concept is widely accepted as a transitional state between healthy cognition and dementia and has proven to be valuable in estimating the risk of pathological cognitive decline [11]. This is highlighted by the inclusion of ‘mild neurocognitive disorders’ as a new category in the Diagnostic and Statistical Manual-5 (DSM-5), which derived mainly from findings and insights on MCI [12]. At the same time and even after extensive research, MCI is a heterogeneous condition and studies vary widely in applied criteria, cognitive domains assessed and sampling methods (clinical vs. population based). Accordingly, prevalence rates have ranged from 3.2% to 20-30% [13-16]. Similar differences have been found concerning progression rates to dementia and AD [16-18]. Various studies frequently illustrate, that clinically based samples usually show slightly higher conversion rates, while reversion rates to ‘normal’ cognition are often higher in population based samples [19, 20]. The first might concentrate more on specific pathological characteristics of AD and numerous studies point to varying and more liberal diagnostic criteria as one important source of heterogeneity in reported prevalence and progression rates [21]. Above that, there is no consensus on specific tests when assessing MCI, yet [10]. Various studies have focused on memory functions [22] and highlight the significance of delayed recall tasks in predicting a progression to AD and in discriminating healthy individuals from MCI patients [23, 24]. This is supported by a review article, reporting a worsened performance in delayed recall tasks for twothirds of the converters from MCI to AD [21]. However, for the very early preclinical phase of AD, the importance of further impaired domains and tasks, such as visuospatial processing or letter fluency has been demonstrated as well [25]. Finally, age, hypertension, diabetes, ApoE 4 allele, gender, depressive symptoms and education have frequently been shown to influence the risk of cognitive decline [2628]. However, some of these factors are mentioned inconsistently, with several studies even reporting that no demographical or medical factor influences the stability of MCI diagnosis [29, 30]. Especially in doubtful or instable cases of MCI, it is not clear yet, if and how these factors represent reliable predictors. MCI does not seem to be a ‘one-way-path’ to dementia. Numerous studies have reported that a transition to normal cognition is fairly common, especially in population-based samples [20, 31]. Factors associated with a reversion to a ‘normal’ or healthy state could comprise disparities in criteria and sampling, as well as the presence or absence of a second condition (e.g. depression), or the number of impaired cognitive domains. Especially the latter has been pointed out by some researchers as an important predictor of instability; subjects with a single-domain
Ellendt et al.
impairment are more likely to revert to age-appropriate cognitive functioning [30-32]. In spite of these heterogeneous results, evidence suggests that MCI can be considered a risk factor for a progression to dementia [21]. A reliable diagnosis of MCI, as well as a reliable prognosis of pathological cognitive decline is important, as premature diagnosis could be incriminating for those affected. Moreover, identification of factors that can predict stability or instability of MCI will enable clinicians to intervene in an early stage and distinguish between those with pathological decline and those who merely show normal signs of ageing (e.g. fluctuating cognitive functions). In order to identify these factors at the very earliest stage, they should preferably be investigated in a sample not yet in need of treatment in a memory clinic. Hence, based on previously reported results [33], one aim of this study was to examine the prevalence and course of MCI in a population-based cohort over a three-year follow-up period with annual measurement points. In a second step, our main objective was to identify which factors can predict a progression to MCI, which can predict stability or reversion of cognitive impairment and which tests can reliably distinguish between stable MCI patients and healthy subjects in this sample. This aim was pursued by predicting outcome at measurement time point three (TP3) based on neuropsychological test results at baseline measurement (TP1). Furthermore, in order to avoid a number of the above named disparities, relatively strict diagnostic criteria as well as an extensive test battery involving multiple cognitive domains were adopted. Additionally, we examined several frequently reported risk factors for MCI and dementia (e.g. age, APOE status, hypertension, diabetes) with the objective of determining their impact and importance for a clinical diagnosis and future progression of cognitive functioning. 2. MATERIALS AND METHOD 2.1. Participants Subjects were recruited as part of the HelMA (Helmholtz Alliance for Mental Health in an Aging Society) longitudinal study on MCI that started in 2010 with an initial sample of 185 volunteers completing the first measurement point. Participants were recruited using newspaper advertisement and via visitations to charity organizations and citizen centres. The study was approved by the Institutional Review Board (IRB) of the Medical Faculty, RWTH Aachen University. Subjects were initially screened for exclusion criteria during a short telephone interview. Participants were included if they were 50 years or older, had sufficient German language and adequate visual performance abilities. Exclusion criteria comprised a diagnosis of dementia, the presence of any neurological, psychiatric or physical disease that could interfere with cognitive performance, lifetime or current drug addiction, seriously reduced vision, inability to follow the protocol and medication use with possible cognitive side effects. At the same time, subjects with diabetes or hypertension were not excluded, for the purpose of examining a possible influence of these diseases. If included, subjects participated in a number of different examinations involving annual neuropsychological testing, structural magnetic resonance imaging as well as functional magnetic reso-
Predicting Stability of Mild Cognitive Impairment (MCI)
Current Alzheimer Research, 2017, Vol. 14, No. 6
3
nance imaging (fMRI) and genetic analyses. Results on fMRI measurements, using only a subgroup of the present sample are reported elsewhere [34]. Examinations have reached the third measurement time point (TP3) and due to drop-out a total of 134 participants completed the neuropsychological examination. Four participants were additionally excluded due to elevated depression scores (BDI-II>16), these subjects were referred to a medical expert. The final sample comprised a total of 130 participants.
ecutive functions, psychomotor speed, visuoconstruction and visuo-spatial abilities, attention and language (supplementary Table 5 and details [33]). Even though olfaction represents no actual cognitive domain, it has been assessed using the Sniffin’ Sticks test (SIT) [35] , as some studies suggest that olfactory deficits might be a possible marker for early AD [36-38].
Depending on the results of neuropsychological testing subjects were divided into MCI-patients and healthy controls, as further described in the following. Based on their classification at all measurement points, participants were subdivided into five distinct groups: 1) “stable controls”, subjects classified as healthy controls during all three measurement points 2) “stable MCI-patients”, subjects classified as MCI-patients during all three measurement points 3) “converters”, subjects who were classified as healthy controls at baseline and as MCI-patients at all following measurement points 4) “re-converters”, participants who were diagnosed with MCI at baseline and returned to normal cognition after that and 5) “changing diagnosis”, participants who had a different diagnosis at each measurement point. In order to counteract a learning effect, parallel test versions were used wherever possible or necessary. Participants were paid for their participation and received a refund for travel expenses.
Next to the neuropsychological examination each year all subjects completed two questionnaires assessing depressive symptoms and management of everyday activities. Specifically, the Beck’s Depression Inventory, German version (BDI-II) was administered in order to capture current mood and depressive symptoms [45]. The Bayer Activities of Daily Living Scale, German version (B-ADL) evaluated coping with day-to-day activities [46]. Above that, the Hamilton Rating Scale for Depression, German version (HAM-D) was applied [47].
All participants gave written informed consent at baseline measurement. The study was performed in accordance with the Declaration of Helsinki. 2.1.1. Classification at Follow-Up 16 control subjects and 35 MCI-patients did not complete either first or second follow-up measurements. One subject was excluded at baseline for not fulfilling the study criteria. 22 subjects could not be contacted after the first or second measurement. 13 persons refused further participation at first or second follow-up, 7 subjects cancelled participation for health reasons, 3 participants died and 5 subjects classified for probable dementia at baseline measurement and were referred to a medical expert. After exclusion, the remaining 130 participants were classified as shown in Fig. (1).
2.3. Psychopathological Evaluation
2.4. MCI Criteria Diagnostic criteria were generally based on Winblad et al. (2004). Participants were diagnosed with MCI if there was 1) impairment of at least -1,5 SD in at least one cognitive test, 2) no dementia, and 3) preserved activities of daily living. Since some studies showed that the subjective estimation of one’s own level of functioning is not always reliable as a criterion for MCI or dementia [39, 40], the presence of subjective cognitive complaints was not included as such. As it is important to avoid false positive diagnosis of MCI, specifications were made regarding the first criterion. If the only deficit was seen in one of the subtests of the Consortium to Establish a Registry for Alzheimer's Disease (CERAD-Plus) or the California Verbal Learning Test (German version, VLMT) participants were only diagnosed with MCI if they also showed deficits in at least one other test (below 1.5 SD). A German dementia screening test (Test zur Früherkennung von Demenzen, TFDD) and the German Syndrom-Kurztest (SKT) were used with the objective of eliminating dementia rather than diagnosing MCI. Likewise, the SIT was included into the test battery but not relevant for a diagnosis of MCI. Additionally, if identified as MCIpatient, allocation into four clinical subtypes followed at each measurement point based on Petersen [7]. For the purpose of this study, only the kind of impairment (single/multiple) was entered into statistical analysis. 2.6. Statistical Analysis
Fig. (1). Classification of subjects at time point three (TP3).
2.2. Neuropsychological Assessment In order to assess a broad range of different cognitive domains all participants completed an extensive neuropsychological test battery, evaluating memory, intelligence, ex-
Neuropsychological, demographical and medical data were analysed using data from baseline (TP1) and third measurement time point (TP3) by SPSS for Windows (Version 22.0). Demographic data are presented in means and standard deviations (SD), medical data as frequencies (n) and percentages. As previously reported, results comprised only part of the final sample of 185 participants [33], neuropsychological and demographic data from TP1 were analyzed again for MCI-patients and controls by means of independent two-sample t-tests for continuous variables and chisquare tests for categorical variables. For group comparison
4 Current Alzheimer Research, 2017, Vol. 14, No. 6
at TP3, univariate analysis of variance (ANOVA) and chisquare tests were conducted comparing five groups; stable patients, stable controls, converter, re-converter and changing diagnosis. The Bonferroni method was used in order to adjust for multiple comparisons. So as to identify the best predictor variables for outcome of diagnosis, logistic regression analysis was performed using test data from TP1. In a first step, multinomial logistic regression was conducted, thereby entering four groups (stable controls, stable MCI, converter and re-converter), using a stepwise method (forward). In a second step, binary logistic regression analysis was performed in order to determine predictor variables for a reversion of diagnosis, again using a stepwise method (forward, p for inclusion = 0.01, p for exclusion = 0.1). For regression analysis all neuropsychological test-scores were entered as predictor variables and in order to avoid multicollinearity, but simultaneously ensure completeness of the model, sex, age, history of hypertension and diabetes were entered as demographic variables. Prior to this, neuropsychological test scores were also examined by correlation analysis, showing only low and medium correlations. Binary logistic regression analysis additionally included variables on single or multiple domain impairment at TP1. Based on results of regression analysis, a standardized sum-score of significant predictive test scores was calculated for each participant. In order to test the discriminant ability of this composite score, univariate analysis of variance (ANOVA) was performed comparing the four relevant groups (stable controls, stable MCI converter, re-converter). Sensitivity and specificity of the composite score alone and compared to predictive single measures, the VLMT-delayed recall (VLMT-DR), the CERAD-recall drawings (CERADRD), the CERAD-Boston Naming Test (CERAD-BNT) and the Benton Test: number of mistakes (BRVT:mistakes) were further analyzed by means of a Receiver Operating Characteristic (ROC) Curve for all groups. Cut-off points were determined by choosing the area under the ROC curve that maximized sensitivity and specificity. To test for further differences and the effect of some possibly influential factors for MCI, moderation analysis with group membership as predictor, test results on the identified predictive tests (VLMT-DR, BVRT: mistakes, CERAD-RD, CERAD-BNT) as outcome and age, IQ as measured by the MWT-B or APOE status as moderator variables were performed. Moderation by each of these three variables was tested separately for each test significantly predicting group membership. The PROCESS tool [41] implemented in SPSS was used and the first model, i.e. testing a simple moderation, was chosen. Data on APOE genotypes was not available for all participants, the influence of genetic status on outcome of diagnosis at follow-up was analyzed for a total of 78 subjects. These subjects were divided into two groups based on their APOE genotype; ε4 carriers (N=21, 26.9%) and ε4 non-carriers (N= 57, 73.1%). 3. RESULTS
Ellendt et al.
jects were men (N=68), 63% women (N= 117). Nearly 9% (N= 16) reported a history of diabetes and around 47% (N=86) reported a history of hypertension. There are missing values regarding demographic data for a small number of participants (HAM-D: N
