1
Journal of Neuropsychology (2015) © 2015 The British Psychological Society www.wileyonlinelibrary.com
Neuropsychological predictors of conversion to probable Alzheimer disease in elderly with mild cognitive impairment Sara Garcıa-Herranz*, M. Carmen Dıaz-Mardomingo and Herminia Peraita Department of Psychology Basic I, National University of Distance Education (UNED), Madrid, Spain In the field of neuropsychology, it is essential to determine which neuropsychological tests predict Alzheimer’s disease (AD) in people with mild cognitive impairment (MCI) and which cut-off points should be used to identify people at greater risk for converting to dementia. The aim of the present study was to analyse the predictive value of the cognitive tests included in a neuropsychological battery for conversion to AD among MCI participants and to analyse the influence of some sociodemographic variables – sex, age, schooling – and others, such as follow-up time and emotional state. A total of 105 participants were assessed with a neuropsychological battery at baseline and during a 3-year follow-up period. For the present study, the data were analysed at baseline. During the follow-up period, 24 participants (22.85%) converted to dementia (2.79 1.14 years) and 81 (77.14%) remained as MCI. The logistic regression analysis determined that the long delay cued recall and the performance time of the Rey figure test were the best predictive tests of conversion to dementia after an MCI diagnosis. Concerning the sociodemographic factors, sex had the highest predictive power. The results reveal the relevance of the neuropsychological data obtained in the first assessment. Specifically, the data obtained in the episodic verbal memory tests and tests that assess visuospatial and executive components may help to identify people with MCI who may develop AD in an interval not longer than 4 years, with the masculine gender being an added risk factor.
Alzheimer’s disease (AD) is a neurodegenerative pathology characterized by progressive cognitive decline that affects neuropsychological functioning in various cognitive processes (attention, memory, executive functions, visuospatial functioning, language, etc.). AD is the most prevalent dementia in our society and will become increasingly frequent due to longer life expectancy in the population in general, and in the older population in particular. Although its aetiology is unknown, it has been confirmed that age is one of the main risk factors that influence its development (Fellows, Bergman, Wolfson, & Chertkow, 2008; Visser, Kester, Jolles, & Verhey, 2006). According to the World Health Organization (WHO, 2008), the global prevalence of dementia due to AD in persons over 60 years will increase to 21% in the year 2050. In Spain, the prevalence of dementia in
*Correspondence should be addressed to Sara Garcıa-Herranz, Juan del Rosal, 10, 28040 Madrid, Spain (email: sgarcia@ bec.uned.es) DOI:10.1111/jnp.12067
2
S. Garcıa-Herranz et al.
general population for people over 70 years is estimated at about 8.2%, and of this percentage, 57.2% are identified as AD (De Pedro-Cuesta et al., 2009). Mild cognitive impairment (MCI) is the priority target of study in the sphere of the neurocognitive sciences due to the interest it evokes for being considered a transitory state between normal ageing and dementia (Peraita, Garcıa-Herranz, & Dıaz-Mardomingo, 2011; Petersen et al., 2001). Mild cognitive impairment is a heterogeneous syndrome, which sometimes represents a prior cognitive condition in the course of the clinical manifestation of dementia, especially AD (Aretouli, Tsilidis, & Brandt, 2013). This syndrome is characterized by presenting problems of episodic memory and/or alterations in other cognitive domains: Attention, language, executive functions, etc. In fact, the term MCI is used in the literature to describe people with cognitive impairment of a heterogeneous aetiology in the absence of significant functional impairment (Winblad et al., 2004). Impairment defined as a function of the affected cognitive domain or domains and the analysis of the evolution of each one of the MCI subtypes has led to the analysis of the relation between MCI and the development of possible neurodegenerative dementias. In fact, the prediction of which people will develop dementia after an MCI diagnosis is essential in the setting of gerontology and clinical neuropsychology in order to apply a palliative and/or preventive treatment as well as to establish cognitive therapeutic strategies that are adapted and specific to early phases to delay or slow down the development of the disease. Different studies have verified that people with deficits in multiple cognitive domains (multidomain amnestic and multidomain non-amnestic MCI) have a higher risk of evolving to dementia than people with impairment in a single domain (monodomain amnestic and monodomain non-amnestic MCI) (Aretouli, Okonkwo, Samek, & Brandt, 2011; Dıaz-Mardomingo, Garcıa-Herranz, & Peraita-Adrados, 2010; Jessen, 2014; Loewenstein et al., 2009; Rasquin, Lodder, Visser, Lousberg, & Verhey, 2005; Summers & Saunders, 2012). The practical and functional nature of cognitive assessment is essential for the early detection of AD. Dubois et al. (2007) and Albert et al. (2011) proposed deterioration of the episodic memory as the core of the diagnostic criteria of prodromal AD. The determination of which cognitive tests are sensitive to identify the first symptoms associated with cognitive impairment of a possible degenerative origin would help to differentiate people with incipient AD from people with MCI. In fact, neuropsychological assessment is still one of the most sensitive and specific markers in the detection of prodromal AD. Specifically, the tests to assess episodic memory and executive functioning have shown their utility in the early detection of MCI, and their predictive value in the diagnosis of incipient AD has been contrasted (Aretouli et al., 2013; B€ackman, 2009; Belleville, Gauthier, Lepage, Kergoat, & Gilbert, 2014; Summers & Saunders, 2012). Regarding episodic memory, long delayed cued recall tests are considered more useful and the best predictors of conversion to AD among MCI patients (Dierckx et al., 2009; Dubois et al., 2014; Sarazin et al., 2007; Wagner et al., 2012). Klages, Fisk, and Rockwood (2005) reported that delayed memory tests were a more robust predictor of the risk of developing AD in the next 5 years than the presence of the alleles 3 or 4 of the ApoE gene. Other authors consider executive functioning tests to be the best predictors of conversion (Aretouli et al., 2013; Brandt et al., 2009; Clark et al., 2012). Manly et al. (2008) showed that people with MCI and isolated executive deficits were less prone to develop dementia than people with MCI and deficits of episodic memory or language alterations. Other authors have attributed
Neuropsychological predictors of conversion to AD
3
the visuospatial deficits in initial stages of AD to deterioration resulting from problems in the executive system (Freeman et al., 2000). In addition to the cognitive variables, other sociodemographic variables have been identified as risk factors for the development of AD. Different studies have verified that age, feminine gender, low educational level, and the duration of evolution until development of the disease, among others, increase the risk (Launer et al., 1999; Lindsay et al., 2002; Mangialasche, Kivipelto, Solomon, & Fratiglioni, 2012; Visser et al., 2006). Other studies indicate that depressive states are a particularly high risk factor for the development of AD (Ownby, Crocco, Acevedo, John, & Loewenstein, 2006; Rushing, Sachs-Ericsson, & Steffens, 2014). Nevertheless, there is some controversy about the influence of each one of these variables on the evolution of dementia. In fact, methodological differences, such as duration of follow-up, and selection and number of study participants may account for these discrepancies. Longitudinal investigations related to the neurocognitive changes that characterize ageing and neurodegenerative diseases such as that of Loewenstein et al. (2009), Lopez et al. (2012) and that carried out in Spain by Peraita et al. (2011) have revealed the great heterogeneity and variability of the cognitive profile of people with MCI. These studies have confirmed that some people with MCI are stable over time, whereas others develop AD, some revert to cognitive normality, and still other people who had reverted to normality revert back to an MCI diagnosis, and finally develop dementia. The prevalence of MCI is very variable; it has been reported that between 30% and 50% of the cases return to a normal cognitive state during follow-up (Garcıa-Herranz, DıazMardomingo, & Peraita, 2014; Lopez et al., 2012; Peraita et al., 2011), although the percentage of people with MCI who develop dementia, especially AD, between 2 and 5 years after diagnosis is between 10% and 20% (Dıaz-Mardomingo et al., 2010; Lonie et al., 2010; Petersen et al., 1999). The possibility of identifying people at high risk of incipient dementia long before their cognitive deficits become clinically evident, through their performance in certain cognitive tests, represents a great advance in the research of dementias. Therefore, the main goal of this study was to analyse the predictive value of the cognitive tests included in the neuropsychological battery used for conversion to AD after an MCI diagnosis and, accordingly, to determine the cut-off point in each of the predictive tests. In addition, we also want to analyse the influence of some sociodemographic variables – sex, age, schooling – and other variables such as follow-up time and emotional state.
Methods Participants Our sample (N = 105 participants) was recruited from a larger sample of 217 participants in an ongoing longitudinal study (ref. SEJ 2004-04233 and SEJ 2007-63325) focused on determining the prevalence the different MCI subtypes (Dıaz & Peraita, 2008; Peraita et al., 2011). The participants were recruited in the Autonomous Community of Madrid (ACM, Spain). The following inclusion and exclusion criteria were used in the general longitudinal project on which this work is based: Inclusion criteria: Individuals aged between 60 and 90 years living in two towns of the ACM who volunteered to participate because they were interested in having an assessment of their cognitive processes, either because they expressed subjective
4
S. Garcıa-Herranz et al.
complaints about the functioning of these processes or simply because they were interested in having information on these processes. They were to be informed about their cognitive processes at the end of the study. Exclusion criteria: (1) previously diagnosed MCI or neurodegenerative disease; (2) disabling chronic disease; (3) psychiatric disorder (major depression. . .); (4) marked neurological abnormality, such as aphasia, agraphia, alexia, and/or apraxia; (5) severe sensory deficit (blindness, hearing impairments. . .); (6) diabetes; (7) cerebrovascular accident; and (8) loss of consciousness. These criteria were confirmed by a neurologist. All participants were informed about the general aim of the study and their involvement therein. The subjects provided written consent to perform the neuropsychological battery in accordance with the guidelines of the Universidad Nacional de Educaci on a Distacia (UNED). The study was approved by the Ethics and Research Committee of the UNED. According to specific goals of this work, only the data obtained from 105 participants – 102 MCIs and three people who were healthy at their first assessment, but who subsequently evolved to AD – were taken into account. The mean age of the 105 participants was 72.36 (6.62) years in the 60–89 range; 73 women (69.50%) and 32 men (30.50%), with a mean of 8.06 years of formal studies (6.13%). They were assessed longitudinally during an average period of 3.22 (1.38) years. During the follow-up period, 24 people (22.85%) evolved into some sort of dementia (2.79 1.14 years): The most prevalent, AD dementia (n = 22; 91.66%), dementia due to normotensive hydrocephalia and mixed dementia (AD with vascular component) (n = 2; 8.34%). Of these, 21 were diagnosed as MCI both at the first assessment and the last evaluation prior to the diagnosis of dementia, and only three people (two men and one woman) were considered healthy at their first assessment, after which they evolved to a diagnosis of MCI, subsequently developing dementia. Eighty-one non-converter individuals (77.15%) remained as MCI over the entire follow-up.
Sociodemographic and clinical characteristics Subjects were interviewed to collect personal information and sociodemographic data – gender, age, education level – as well as information on their lifestyle and habits – tobacco – and clinical condition – hypertension, cholesterol, diabetes, and thyroid problems. Prior to the neuropsychological assessment, the Yesavage scale (reduced version of the Geriatric Depression Scale; Yesavage, Brink, Rose, & Lum, 1982) was applied to assess the participant’s emotional state, as well as a personal autonomy scale (the Blessed Dementia Scale; Blessed, Tomlinson, & Roth, 1968). Specifically, we used part A of the Blessed Scale, which evaluates the performance of daily life activities. Participants’ general cognitive state was first measured by the Spanish version of the Mini-Mental State Examination (MMSE), which scores over 35 points instead of over 30 (Lobo, Ezquerra, G omez, Sala, & Seva, 1979).
Neuropsychological assessment The neuropsychological battery was applied at baseline and at yearly intervals for the next 3 years. The tests included in the neuropsychological battery assess performance in different cognitive domains: Episodic memory, executive function, attention, language, visuospatial constructional skill, praxis constructive graphics,
Neuropsychological predictors of conversion to AD
5
and ideomotor praxis. Episodic memory and learning were assessed with the Verbal Learning Test of the Complutense University (TAVEC) (Spanish version of the Californian Verbal Learning Test [CVLT]) (Benedet & Alejandre, 1998). The TAVEC battery includes the following tests: Immediate free recall (lists A and interference list B), short delay free recall, short delay cued recall, long delay free recall, long delay cued recall, and recognition. Processing speed, attention, and executive function were assessed with Trail Making Tests A and B (Reitan & Wolfson, 1993) and alternating graphs and loops (Pe~ na-Casanova, 1991). Language: Phonemic fluency (P) (Pe~ na-Casanova, 1991) and semantic fluency (animals, plants, clothes, and vehicles) (Peraita, Gonzalez-Labra, Sanchez-Bernardos, & Galeote, 2000). Only the correct number of words produced was computed. Copying the Rey–Osterrieth complex figure (Rey, 2003) allowed assessment of organization and planning ability, as well as visuospatial constructional ability. Time and quality in the reproduction of the Rey figure were computed. Ideomotor praxis was assessed through mimicking the use of objects and symbolic communication gestures (Pe~ na-Casanova, 1991) (Table 1). The neuropsychological assessments were carried out by psychologists specialized in ageing and neuropsychology, at cultural and educational centres for older people of the municipalities of Las Rozas and Pozuelo de Alarc on (ACM). The assessment lasted approximately 90 min.
Participant classification procedure Neuropsychological evaluations were conducted every 12 2 months during a 3-year follow-up. The scores obtained in each of the tests of the neuropsychological assessment were interpreted according to standardized norms in each one of the tests. Scores that were 1.5 SD below standardized norms for age and education reflect a deficient performance. In Trail Making Tests A and B and in the Rey figure (variable ‘time’), the criterion followed was 1.5 SD above the mean (Dıaz & Peraita, 2008; Peraita et al., 2011). Mild cognitive impairment was defined as having a score 1.5 SD below the mean in at least 2 of the tests applied. As shown in Table 2, depending on the data obtained through the different neuropsychological assessments, the participants were classified in one of the following cognitive profiles: Healthy individuals – expected performance according to references scales – or MCIs – different subtypes: Amnestic MCI (aMCI), no amnestic MCI (naMCI) and multidomain MCI (mMCI): (1). aMCI: 1.5 SD below the mean in at least two tests of the TAVEC battery. (2). naMCI: 1.5 SD below the mean in two or more tests, none of which are from the TAVEC. (3). mMCI: 1.5 SD below the mean in a TAVEC test and in at least one other test (different from the TAVEC test). The diagnosis of AD was made and confirmed by the neurologists of the research team according to the DSM IV criteria and the NINCDS-ADRDA (American Psychiatric Association, 1994) criteria after neuropsychological assessment. The people who converted to dementia over the 3-year follow-up were considered as converters, and the people diagnosed as MCI who remained in that same cognitive profile without conversion to dementia during follow-up were considered as non-converters (see Table 2).
6
S. Garcıa-Herranz et al.
Table 1. Neuropsychology battery tests and range of scores in each one of the tests Assessment cognitive domain Emotional state Daily life activities General cognitive state Verbal Episodic Memory
Tests Yesavage Blessed
Total score (0–15) Total score (part A) (0–4)
MMSE (Spanish version)
Total score (0–35)
TAVEC total immediate free recall (list A) TAVEC total immediate free recall (list B) TAVEC short delay free recall
Total number of correct responses in five trials (0–80) Total number of correct responses recalled (0–16) Total number of correct responses recalled (0–16) Total number of correct responses recalled (0–16) Total number of correct responses recalled (0–16) Total number of correct responses recalled (0–16) Total number of correct responses recognized (0–16) Total number of correct responses produced in 3 min Total number of correct responses produced for category (1 min) Total time in the numerical implementation Total time in the alpha-numerical implementation Total quality (0–4) Total quality (0–36) Time Total quality (0–12)
TAVEC short delay cued recall TAVEC long delay free recall TAVEC long delay cued recall Recognition Language, attention, and executive function
Verbal fluency/p/ Category fluency/animals, cloths, plants, vehicles/ Trail Making Test A Trail Making Test B
Visuospatial constructional capacity and praxis constructive graphics Ideomotor praxis
Measure
Alternating graphs and loops Rey–Osterrieth complex figure Rey–Osterrieth complex figure Visuoconstructive praxis
Mimicking the use of objects and symbolic communication gestures
Total quality (0–20)
Note. TAVEC, Verbal Learning Test of the Complutense University.
Statistical analysis We performed a descriptive analysis of all the variables of interest. We performed nonparametric Mann–Whitney U (z) and Wilcoxson (z) techniques because the scores of the tests used were not distributed normally (Kolmogorov–Smirnov and Shapiro–Wilks test). The level of significance was .05. To identify the tests that could be clinically useful in the prediction of conversion to probable AD, we carried out a logistic regression model. The goal was to analyse the prediction of conversion to AD based on the results of the neuropsychological tests at the
Neuropsychological predictors of conversion to AD
7
Table 2. Cognitive profiles of converters and non-converters. First evaluation Converters
Non-converters
Cognitive profiles
n
Male
Female
n
Male
Female
Healthy aMCI naMCI mMCI Total
3 4 3 14 24
2 1 0 7 10
1 3 3 7 14
0 6 42 33 81
0 4 4 21 29
0 2 38 12 52
Note. aMCI, amnestic MCI; naMCI, no amnestic MCI; mMCI, multidomain MCI.
first assessment. Standardized regression coefficients were used to compare the relative strength between the independent variables and the dependent variable, because the independent variables were measured with different scales (Menard, 2010). In order to determine the sensitivity and specificity, in addition to the positive predictive value (PPV) and negative predictive value (NPV) of the predictive tests in the binary logistic regression, we analysed the receiver operating characteristic (ROC) curves, accepting an explained area of >0.70. Lastly, to analyse the influence or effect of different sociodemographic variables – sex, age, education – and other variables such as follow-up time and emotional state in the development of conversion to dementia, we also applied binary logistic regression analysis.
Results Baseline sociodemographic data of converters and non-converters There were significant differences between converters and non-converters in gender (v2 = 4.223, p < .040) and with regard to the functional activity assessed by means of the Blessed Scale (z = 2.958, p < .003). Women predominated in the group of nonconverters, and functional competence was more compromised in the group of converters versus the non-converters (Table 3).
Neuropsychological data Mann–Whitney’s U-test showed the existence of significant differences between the converters and non-converters in the following memory tests: Learning list A of the TAVEC (z = 4.579, p < .001), short delay free recall (z = 4.444, p < .001), short delay cued recall (z = 3.361, p < .001), long delay free recall (z = 4.424, p < .001), long delay cued recall (z = 4.795, p < .001), and in execution time of the Rey figure (z = 3.106, p < .002). The scores of the converters were significantly worse than those of the nonconverters in each of the above-mentioned cognitive tests (Table 3). Figure 1 presents the performance of the converters versus the non-converters.
Predictors of conversion To determine the predictive value of the neuropsychological tests used, we performed a binary logistic regression analysis in which we included all the measures of the
8
S. Garcıa-Herranz et al.
Table 3. Baseline sociodemographic and clinical data and neuropsychological performance of converters and non-converters
Sociodemographic variables Gender (female) Age (years) Formal education (years) Follow-up time (years) Geriatric Depression Scale, mean (GDS, Yesavage) Functional level: Blessed Dementia Rating (Blessed) Neuropsychological tests Cognitive status (MMSE Spanish version) Fluency Verbal fluency/p/ Category fluency/animals, clothing, plants, vehicles/ TAVEC Immediate free recall (list A) Immediate free recall (list B) Short delay free recall Short delay cued recall Long delay free recall Long delay cued recall Recognition TMT TMT A TMT B Alternating graphs and loops Quality Rey figure Time Quality Constructive praxias Quality Ideomotor praxias Quality
Mean (SD) Converters n = 24
Mean (SD) Non-converters n = 81
p-Value
12 (50.00%) 74.17 (5.89) 8.38 (6.39) 2.79 (1.14) 4.21 (3.40)
61 (75.30%) 71.83 (6.76) 7.96 (6.10) 3.35 (1.42) 3.52 (2.61)
.040 ns ns ns ns
1.45 (1.06)
0.83 (0.98)
.003
29.46 (3.83)
29.79 (3.61)
ns
22.00 (9.41) 46.54 (9.49)
22.30 (8.93) 48.63 (11.99)
ns ns
30.54 (11.72) 3.92 (1.90) 4.04 (2.75) 6.92 (2.85) 4.33 (3.52) 5.96 (3.04) 13.00 (2.41)
42.23 (10.51) 4.70 (2.05) 7.81 (3.34) 8.98 (2.90) 8.16 (3.59) 9.14 (2.80) 13.84 (2.27)
.001 ns .001 .001 .001 .001 ns
84.00 (38.32) 189.23 (98.96)
88.16 (51.28) 180.20 (86.05)
ns ns
2.58 (1.17)
2.47 (1.23)
ns
329.87 (140.22) 23.34 (9.24)
237.68 (100.90) 22.07 (9.75)
.002 ns
8.96 (2.67)
8.66 (2.42)
ns
17.33 (2.94)
17.37 (2.36)
ns
Note. TAVEC, Verbal Learning Test of the Complutense University. Statistically significant (p ≤ .05).
neuropsychological tests that found significant differences between converters and nonconverters at the first assessment. The analysis showed that only long delay cued recall (p < .001, OR = 0.586, 95% CI: 0.448–0.767) and execution time of the Rey figure (p < .005, OR = 1.008, 95% CI: 1.003–1.014) were statistically significant predictor variables; that is, these tests contribute the most to the prediction of conversion to dementia after an MCI diagnosis (Table 4). The effect size was very large for the delayed cued recall test (d = 1.326) and for the execution time of the Rey figure (d = 0.958). The
Neuropsychological predictors of conversion to AD 55
Non-converters
9
Converters
T-scores
50 45 40 35 30
Figure 1. Cognitive performance of converters versus non-converters. First evaluation.
Table 4. Multiple logistic regression model for prediction: Conversion/non-conversion Included variables Long delay cued recall Rey figure execution time
B coefficient
B (SE)
OR
95% CI of OR
Wald
df
p-significance
.535
.137
0.586
0.448–0.767
15.17
1
.001**
.008
.003
1.008
1.003–1.014
9.72
1
.005*
Excluded variables
Wald
df
p-significance
Short delay cued recall Total immediate free recall (list A) Long delay free recall Short delay free recall Blessed Dementia Rating (part A)
3.45
1
ns
1.27
1
ns
0.72 0.26 0.12
1 1 1
ns ns ns
Summary of the model Significant Omnibus test model: Hosmer-Lemeshow test: R2 Nagelkerke: .512 Total percentage correct classification: 87.9%
v2(2) = 36.89, p = .001** v2(8) = 7.07, p = .529 ns
Note. ns, non-significant (p > .50), *p < .005; **p < .001.
percentage of correct classification was 87.9%. So, almost 90% of the people were correctly diagnosed. As a measure of the diagnostic precision of the tests, we studied the ROC curves to identify the cognitive tests with greater predictive capacity, sensitivity, and specificity, and the optimal value at which one could consider deterioration compatible with dementia or probable AD. We also analysed the PPV and NPV. Figure 2 shows the ROC curves for long delay cued recall and execution time of the Rey figure, in which diagnostic groups studied – converters versus non-converters – are
10
S. Garcıa-Herranz et al.
Figure 2. Receiver operating characteristic (ROC) curves. Long delay cued recall (AUC = 0.839) and execution time of Rey figure test (AUC = 0.728).
Table 5. Sensitivity, specificity, positive (PPV), and negative predictive values (NPV) and cut-off points of long delay cued recall and execution time of Rey figure test Cognitive tests
Cut-off points Sensitivity (%) Specificity (%) Positive predictive value (PPV) (%) Negative predictive value PPN (%)
Long delay cued recall
Execution time of Rey figure
247 70 62 34.1 88
contrasted. In Table 5, we present the values of sensitivity and specificity, as well as the PPV and NPV for the different optimal cut-points for each of the above tests. The results of these tests are very sensitive. For long delay cued recall, we observe an area under the curve and a highly significant reference diagonal line (AUC = 0.839, 95% CI: 0.731–0.946, p < .001). For execution time of the Rey figure, we observe an area under the curve and a diagonal reference line (AUC = 0.728, 95% CI: 0.597–0.859), in this case, also very significant (p < .002). This indicates that the tests have good predictive or diagnostic capacity. With regard to the results obtained with the regression model using the sociodemographic predictors, follow-up time, and emotional state, bivariate analysis revealed the variables sex, v2(1) = 5.570, p < .018, and follow-up time (z = 1.979, p < .048) to be significant. However, in the logistic regression model factor, gender was the only significant factor (p < .021; OR = 3.050, 95% CI: 1.184–7.856), such that the risk of converting to AD is higher in men than in women.
Neuropsychological predictors of conversion to AD
11
Discussion In this study, the main goals were to analyse the predictive value of the cognitive tests included in the neuropsychological battery used for conversion to AD after an MCI diagnosis and, accordingly, to determine the cut-off point in each of the predictive tests. Our third goal was to analyse the influence of some sociodemographic variables – sex, age, schooling – and other variables such as follow-up time and emotional state. Regarding the first objective, our findings showed that some of the tests that assess episodic memory and visuospatial ability were the most predictive of conversion to dementia, specifically, the long delay cued recall test and the execution time of the Rey figure. Of these two tests, the most sensitive test with the greatest predictor capacity for conversion to dementia was long delay cued recall. The results found in the present work regarding the predictive value of episodic memory tests are consistent with other research (Albert, Blacker, Moss, Tanzi, & McArdle, 2007; Dierckx et al., 2009; Dubois et al., 2014; Egli et al., 2014; Hanseeuw & Ivanoiu, 2011; Ivanoiu et al., 2005; Sarazin et al., 2007; Silva et al., 2012; Wagner et al., 2012), which used similar tests and in the same way as those used in this work, showing that deterioration of the episodic memory is associated with probable AD. Dierckx et al. (2009) concluded that the delayed cued recall task was the best predictor of conversion to AD among MCI patients. Hanseeuw and Ivanoiu (2011) showed that the cued recall test could predict dementia between 5 and up to 10 years before its onset. Sarazin et al. (2007) showed that impairment of free recall, total recall, and index of sensitivity of cueing could identify prodromal AD in patients with MCI. In a similar vein, Wagner et al. (2012) found that deficits in memory tests are very closely related to the biomarkers indicative of AD in subjects with MCI and conclude that this finding complements the results of prospective clinical studies and provides empirical support ratifying that cued recall tasks can be considered an appropriate specific indicator of prodromal AD. Our results ratify what has already been described in the literature: Years before the evolution of AD, due to the atrophy of the temporal structures, people with MCI who evolve to dementia present memory deficits due to their greater difficulty in information coding and organization processes, which are revealed in the recall process, regardless of the conditions of facilitation through semantic cues, as reported by Carlesimo, Perri, and Caltagirone (2011). Moreover, visuospatial ability-related research has in general shown that the deterioration of this capacity is even evident years before the diagnosis of AD (B€ackman, Jones, Berger, Laukka, & Small, 2005; Laukka, Macdonald, Fratiglioni, & B€ackman, 2012). We also highlight that we observed significant differences between the groups studied in the time required to execute the Rey figure, a variable related to visuospatial skill and to different components of the executive function, such as planning, organization, and attentional capacity (Somerville, Tremont, & Stern, 2000; Watanabe et al., 2005). This suggests that the increase in the execution time of the Rey figure reflected in the converters may be related not only to visuospatial problems but also to attentional, planning, and organizational problems, related to a generalized decrease of information processing, rather than to specific deficits in certain aspects of visual processing determined by the quality of the execution of the Rey figure. Rapp and Reischies (2005) verified that older individuals with early dementia exhibit slower information-processing speed in attentional and executive tests. The increase in execution time reveals that alterations compatible with initial dementia are not limited only to mnestic performance, but instead visuospatial, executive, and attentional components are also affected (Salmon
12
S. Garcıa-Herranz et al.
& Bondi, 2009). It should not be forgotten that some people included in our study diagnosed as AD by the neurologist presented mixed pathology (probable AD with an associated vascular component), and, as pointed out by Bastos-Leite et al. (2007), in these cases, the executive function tests and delayed memory tests are affected. There is increasing evidence showing that people with memory deficits and alterations in other cognitive functions are at greater risk of developing AD than are people who present isolated memory deficits (Albert et al., 2011; Brandt et al., 2009; Dıaz-Mardomingo et al., 2010; Lonie, Herrmann, Donaghey, & Ebmeier, 2008; Ritchie, Artero, & Touchon, 2001; Sacuiu et al., 2009; Tabert et al., 2006). The results of this work reveal that visuospatial skills are still an important component of neuropsychological assessment to detect AD, not only because of their sensitivity for the detection of this neurocognitive pathology, but also because this domain is associated with other cognitive domains that are altered in probable AD. With regard to the study of the predictive validity of long delay cued recall, we obtained a cut-off point of
Journal of Neuropsychology (2015) © 2015 The British Psychological Society www.wileyonlinelibrary.com
Neuropsychological predictors of conversion to probable Alzheimer disease in elderly with mild cognitive impairment Sara Garcıa-Herranz*, M. Carmen Dıaz-Mardomingo and Herminia Peraita Department of Psychology Basic I, National University of Distance Education (UNED), Madrid, Spain In the field of neuropsychology, it is essential to determine which neuropsychological tests predict Alzheimer’s disease (AD) in people with mild cognitive impairment (MCI) and which cut-off points should be used to identify people at greater risk for converting to dementia. The aim of the present study was to analyse the predictive value of the cognitive tests included in a neuropsychological battery for conversion to AD among MCI participants and to analyse the influence of some sociodemographic variables – sex, age, schooling – and others, such as follow-up time and emotional state. A total of 105 participants were assessed with a neuropsychological battery at baseline and during a 3-year follow-up period. For the present study, the data were analysed at baseline. During the follow-up period, 24 participants (22.85%) converted to dementia (2.79 1.14 years) and 81 (77.14%) remained as MCI. The logistic regression analysis determined that the long delay cued recall and the performance time of the Rey figure test were the best predictive tests of conversion to dementia after an MCI diagnosis. Concerning the sociodemographic factors, sex had the highest predictive power. The results reveal the relevance of the neuropsychological data obtained in the first assessment. Specifically, the data obtained in the episodic verbal memory tests and tests that assess visuospatial and executive components may help to identify people with MCI who may develop AD in an interval not longer than 4 years, with the masculine gender being an added risk factor.
Alzheimer’s disease (AD) is a neurodegenerative pathology characterized by progressive cognitive decline that affects neuropsychological functioning in various cognitive processes (attention, memory, executive functions, visuospatial functioning, language, etc.). AD is the most prevalent dementia in our society and will become increasingly frequent due to longer life expectancy in the population in general, and in the older population in particular. Although its aetiology is unknown, it has been confirmed that age is one of the main risk factors that influence its development (Fellows, Bergman, Wolfson, & Chertkow, 2008; Visser, Kester, Jolles, & Verhey, 2006). According to the World Health Organization (WHO, 2008), the global prevalence of dementia due to AD in persons over 60 years will increase to 21% in the year 2050. In Spain, the prevalence of dementia in
*Correspondence should be addressed to Sara Garcıa-Herranz, Juan del Rosal, 10, 28040 Madrid, Spain (email: sgarcia@ bec.uned.es) DOI:10.1111/jnp.12067
2
S. Garcıa-Herranz et al.
general population for people over 70 years is estimated at about 8.2%, and of this percentage, 57.2% are identified as AD (De Pedro-Cuesta et al., 2009). Mild cognitive impairment (MCI) is the priority target of study in the sphere of the neurocognitive sciences due to the interest it evokes for being considered a transitory state between normal ageing and dementia (Peraita, Garcıa-Herranz, & Dıaz-Mardomingo, 2011; Petersen et al., 2001). Mild cognitive impairment is a heterogeneous syndrome, which sometimes represents a prior cognitive condition in the course of the clinical manifestation of dementia, especially AD (Aretouli, Tsilidis, & Brandt, 2013). This syndrome is characterized by presenting problems of episodic memory and/or alterations in other cognitive domains: Attention, language, executive functions, etc. In fact, the term MCI is used in the literature to describe people with cognitive impairment of a heterogeneous aetiology in the absence of significant functional impairment (Winblad et al., 2004). Impairment defined as a function of the affected cognitive domain or domains and the analysis of the evolution of each one of the MCI subtypes has led to the analysis of the relation between MCI and the development of possible neurodegenerative dementias. In fact, the prediction of which people will develop dementia after an MCI diagnosis is essential in the setting of gerontology and clinical neuropsychology in order to apply a palliative and/or preventive treatment as well as to establish cognitive therapeutic strategies that are adapted and specific to early phases to delay or slow down the development of the disease. Different studies have verified that people with deficits in multiple cognitive domains (multidomain amnestic and multidomain non-amnestic MCI) have a higher risk of evolving to dementia than people with impairment in a single domain (monodomain amnestic and monodomain non-amnestic MCI) (Aretouli, Okonkwo, Samek, & Brandt, 2011; Dıaz-Mardomingo, Garcıa-Herranz, & Peraita-Adrados, 2010; Jessen, 2014; Loewenstein et al., 2009; Rasquin, Lodder, Visser, Lousberg, & Verhey, 2005; Summers & Saunders, 2012). The practical and functional nature of cognitive assessment is essential for the early detection of AD. Dubois et al. (2007) and Albert et al. (2011) proposed deterioration of the episodic memory as the core of the diagnostic criteria of prodromal AD. The determination of which cognitive tests are sensitive to identify the first symptoms associated with cognitive impairment of a possible degenerative origin would help to differentiate people with incipient AD from people with MCI. In fact, neuropsychological assessment is still one of the most sensitive and specific markers in the detection of prodromal AD. Specifically, the tests to assess episodic memory and executive functioning have shown their utility in the early detection of MCI, and their predictive value in the diagnosis of incipient AD has been contrasted (Aretouli et al., 2013; B€ackman, 2009; Belleville, Gauthier, Lepage, Kergoat, & Gilbert, 2014; Summers & Saunders, 2012). Regarding episodic memory, long delayed cued recall tests are considered more useful and the best predictors of conversion to AD among MCI patients (Dierckx et al., 2009; Dubois et al., 2014; Sarazin et al., 2007; Wagner et al., 2012). Klages, Fisk, and Rockwood (2005) reported that delayed memory tests were a more robust predictor of the risk of developing AD in the next 5 years than the presence of the alleles 3 or 4 of the ApoE gene. Other authors consider executive functioning tests to be the best predictors of conversion (Aretouli et al., 2013; Brandt et al., 2009; Clark et al., 2012). Manly et al. (2008) showed that people with MCI and isolated executive deficits were less prone to develop dementia than people with MCI and deficits of episodic memory or language alterations. Other authors have attributed
Neuropsychological predictors of conversion to AD
3
the visuospatial deficits in initial stages of AD to deterioration resulting from problems in the executive system (Freeman et al., 2000). In addition to the cognitive variables, other sociodemographic variables have been identified as risk factors for the development of AD. Different studies have verified that age, feminine gender, low educational level, and the duration of evolution until development of the disease, among others, increase the risk (Launer et al., 1999; Lindsay et al., 2002; Mangialasche, Kivipelto, Solomon, & Fratiglioni, 2012; Visser et al., 2006). Other studies indicate that depressive states are a particularly high risk factor for the development of AD (Ownby, Crocco, Acevedo, John, & Loewenstein, 2006; Rushing, Sachs-Ericsson, & Steffens, 2014). Nevertheless, there is some controversy about the influence of each one of these variables on the evolution of dementia. In fact, methodological differences, such as duration of follow-up, and selection and number of study participants may account for these discrepancies. Longitudinal investigations related to the neurocognitive changes that characterize ageing and neurodegenerative diseases such as that of Loewenstein et al. (2009), Lopez et al. (2012) and that carried out in Spain by Peraita et al. (2011) have revealed the great heterogeneity and variability of the cognitive profile of people with MCI. These studies have confirmed that some people with MCI are stable over time, whereas others develop AD, some revert to cognitive normality, and still other people who had reverted to normality revert back to an MCI diagnosis, and finally develop dementia. The prevalence of MCI is very variable; it has been reported that between 30% and 50% of the cases return to a normal cognitive state during follow-up (Garcıa-Herranz, DıazMardomingo, & Peraita, 2014; Lopez et al., 2012; Peraita et al., 2011), although the percentage of people with MCI who develop dementia, especially AD, between 2 and 5 years after diagnosis is between 10% and 20% (Dıaz-Mardomingo et al., 2010; Lonie et al., 2010; Petersen et al., 1999). The possibility of identifying people at high risk of incipient dementia long before their cognitive deficits become clinically evident, through their performance in certain cognitive tests, represents a great advance in the research of dementias. Therefore, the main goal of this study was to analyse the predictive value of the cognitive tests included in the neuropsychological battery used for conversion to AD after an MCI diagnosis and, accordingly, to determine the cut-off point in each of the predictive tests. In addition, we also want to analyse the influence of some sociodemographic variables – sex, age, schooling – and other variables such as follow-up time and emotional state.
Methods Participants Our sample (N = 105 participants) was recruited from a larger sample of 217 participants in an ongoing longitudinal study (ref. SEJ 2004-04233 and SEJ 2007-63325) focused on determining the prevalence the different MCI subtypes (Dıaz & Peraita, 2008; Peraita et al., 2011). The participants were recruited in the Autonomous Community of Madrid (ACM, Spain). The following inclusion and exclusion criteria were used in the general longitudinal project on which this work is based: Inclusion criteria: Individuals aged between 60 and 90 years living in two towns of the ACM who volunteered to participate because they were interested in having an assessment of their cognitive processes, either because they expressed subjective
4
S. Garcıa-Herranz et al.
complaints about the functioning of these processes or simply because they were interested in having information on these processes. They were to be informed about their cognitive processes at the end of the study. Exclusion criteria: (1) previously diagnosed MCI or neurodegenerative disease; (2) disabling chronic disease; (3) psychiatric disorder (major depression. . .); (4) marked neurological abnormality, such as aphasia, agraphia, alexia, and/or apraxia; (5) severe sensory deficit (blindness, hearing impairments. . .); (6) diabetes; (7) cerebrovascular accident; and (8) loss of consciousness. These criteria were confirmed by a neurologist. All participants were informed about the general aim of the study and their involvement therein. The subjects provided written consent to perform the neuropsychological battery in accordance with the guidelines of the Universidad Nacional de Educaci on a Distacia (UNED). The study was approved by the Ethics and Research Committee of the UNED. According to specific goals of this work, only the data obtained from 105 participants – 102 MCIs and three people who were healthy at their first assessment, but who subsequently evolved to AD – were taken into account. The mean age of the 105 participants was 72.36 (6.62) years in the 60–89 range; 73 women (69.50%) and 32 men (30.50%), with a mean of 8.06 years of formal studies (6.13%). They were assessed longitudinally during an average period of 3.22 (1.38) years. During the follow-up period, 24 people (22.85%) evolved into some sort of dementia (2.79 1.14 years): The most prevalent, AD dementia (n = 22; 91.66%), dementia due to normotensive hydrocephalia and mixed dementia (AD with vascular component) (n = 2; 8.34%). Of these, 21 were diagnosed as MCI both at the first assessment and the last evaluation prior to the diagnosis of dementia, and only three people (two men and one woman) were considered healthy at their first assessment, after which they evolved to a diagnosis of MCI, subsequently developing dementia. Eighty-one non-converter individuals (77.15%) remained as MCI over the entire follow-up.
Sociodemographic and clinical characteristics Subjects were interviewed to collect personal information and sociodemographic data – gender, age, education level – as well as information on their lifestyle and habits – tobacco – and clinical condition – hypertension, cholesterol, diabetes, and thyroid problems. Prior to the neuropsychological assessment, the Yesavage scale (reduced version of the Geriatric Depression Scale; Yesavage, Brink, Rose, & Lum, 1982) was applied to assess the participant’s emotional state, as well as a personal autonomy scale (the Blessed Dementia Scale; Blessed, Tomlinson, & Roth, 1968). Specifically, we used part A of the Blessed Scale, which evaluates the performance of daily life activities. Participants’ general cognitive state was first measured by the Spanish version of the Mini-Mental State Examination (MMSE), which scores over 35 points instead of over 30 (Lobo, Ezquerra, G omez, Sala, & Seva, 1979).
Neuropsychological assessment The neuropsychological battery was applied at baseline and at yearly intervals for the next 3 years. The tests included in the neuropsychological battery assess performance in different cognitive domains: Episodic memory, executive function, attention, language, visuospatial constructional skill, praxis constructive graphics,
Neuropsychological predictors of conversion to AD
5
and ideomotor praxis. Episodic memory and learning were assessed with the Verbal Learning Test of the Complutense University (TAVEC) (Spanish version of the Californian Verbal Learning Test [CVLT]) (Benedet & Alejandre, 1998). The TAVEC battery includes the following tests: Immediate free recall (lists A and interference list B), short delay free recall, short delay cued recall, long delay free recall, long delay cued recall, and recognition. Processing speed, attention, and executive function were assessed with Trail Making Tests A and B (Reitan & Wolfson, 1993) and alternating graphs and loops (Pe~ na-Casanova, 1991). Language: Phonemic fluency (P) (Pe~ na-Casanova, 1991) and semantic fluency (animals, plants, clothes, and vehicles) (Peraita, Gonzalez-Labra, Sanchez-Bernardos, & Galeote, 2000). Only the correct number of words produced was computed. Copying the Rey–Osterrieth complex figure (Rey, 2003) allowed assessment of organization and planning ability, as well as visuospatial constructional ability. Time and quality in the reproduction of the Rey figure were computed. Ideomotor praxis was assessed through mimicking the use of objects and symbolic communication gestures (Pe~ na-Casanova, 1991) (Table 1). The neuropsychological assessments were carried out by psychologists specialized in ageing and neuropsychology, at cultural and educational centres for older people of the municipalities of Las Rozas and Pozuelo de Alarc on (ACM). The assessment lasted approximately 90 min.
Participant classification procedure Neuropsychological evaluations were conducted every 12 2 months during a 3-year follow-up. The scores obtained in each of the tests of the neuropsychological assessment were interpreted according to standardized norms in each one of the tests. Scores that were 1.5 SD below standardized norms for age and education reflect a deficient performance. In Trail Making Tests A and B and in the Rey figure (variable ‘time’), the criterion followed was 1.5 SD above the mean (Dıaz & Peraita, 2008; Peraita et al., 2011). Mild cognitive impairment was defined as having a score 1.5 SD below the mean in at least 2 of the tests applied. As shown in Table 2, depending on the data obtained through the different neuropsychological assessments, the participants were classified in one of the following cognitive profiles: Healthy individuals – expected performance according to references scales – or MCIs – different subtypes: Amnestic MCI (aMCI), no amnestic MCI (naMCI) and multidomain MCI (mMCI): (1). aMCI: 1.5 SD below the mean in at least two tests of the TAVEC battery. (2). naMCI: 1.5 SD below the mean in two or more tests, none of which are from the TAVEC. (3). mMCI: 1.5 SD below the mean in a TAVEC test and in at least one other test (different from the TAVEC test). The diagnosis of AD was made and confirmed by the neurologists of the research team according to the DSM IV criteria and the NINCDS-ADRDA (American Psychiatric Association, 1994) criteria after neuropsychological assessment. The people who converted to dementia over the 3-year follow-up were considered as converters, and the people diagnosed as MCI who remained in that same cognitive profile without conversion to dementia during follow-up were considered as non-converters (see Table 2).
6
S. Garcıa-Herranz et al.
Table 1. Neuropsychology battery tests and range of scores in each one of the tests Assessment cognitive domain Emotional state Daily life activities General cognitive state Verbal Episodic Memory
Tests Yesavage Blessed
Total score (0–15) Total score (part A) (0–4)
MMSE (Spanish version)
Total score (0–35)
TAVEC total immediate free recall (list A) TAVEC total immediate free recall (list B) TAVEC short delay free recall
Total number of correct responses in five trials (0–80) Total number of correct responses recalled (0–16) Total number of correct responses recalled (0–16) Total number of correct responses recalled (0–16) Total number of correct responses recalled (0–16) Total number of correct responses recalled (0–16) Total number of correct responses recognized (0–16) Total number of correct responses produced in 3 min Total number of correct responses produced for category (1 min) Total time in the numerical implementation Total time in the alpha-numerical implementation Total quality (0–4) Total quality (0–36) Time Total quality (0–12)
TAVEC short delay cued recall TAVEC long delay free recall TAVEC long delay cued recall Recognition Language, attention, and executive function
Verbal fluency/p/ Category fluency/animals, cloths, plants, vehicles/ Trail Making Test A Trail Making Test B
Visuospatial constructional capacity and praxis constructive graphics Ideomotor praxis
Measure
Alternating graphs and loops Rey–Osterrieth complex figure Rey–Osterrieth complex figure Visuoconstructive praxis
Mimicking the use of objects and symbolic communication gestures
Total quality (0–20)
Note. TAVEC, Verbal Learning Test of the Complutense University.
Statistical analysis We performed a descriptive analysis of all the variables of interest. We performed nonparametric Mann–Whitney U (z) and Wilcoxson (z) techniques because the scores of the tests used were not distributed normally (Kolmogorov–Smirnov and Shapiro–Wilks test). The level of significance was .05. To identify the tests that could be clinically useful in the prediction of conversion to probable AD, we carried out a logistic regression model. The goal was to analyse the prediction of conversion to AD based on the results of the neuropsychological tests at the
Neuropsychological predictors of conversion to AD
7
Table 2. Cognitive profiles of converters and non-converters. First evaluation Converters
Non-converters
Cognitive profiles
n
Male
Female
n
Male
Female
Healthy aMCI naMCI mMCI Total
3 4 3 14 24
2 1 0 7 10
1 3 3 7 14
0 6 42 33 81
0 4 4 21 29
0 2 38 12 52
Note. aMCI, amnestic MCI; naMCI, no amnestic MCI; mMCI, multidomain MCI.
first assessment. Standardized regression coefficients were used to compare the relative strength between the independent variables and the dependent variable, because the independent variables were measured with different scales (Menard, 2010). In order to determine the sensitivity and specificity, in addition to the positive predictive value (PPV) and negative predictive value (NPV) of the predictive tests in the binary logistic regression, we analysed the receiver operating characteristic (ROC) curves, accepting an explained area of >0.70. Lastly, to analyse the influence or effect of different sociodemographic variables – sex, age, education – and other variables such as follow-up time and emotional state in the development of conversion to dementia, we also applied binary logistic regression analysis.
Results Baseline sociodemographic data of converters and non-converters There were significant differences between converters and non-converters in gender (v2 = 4.223, p < .040) and with regard to the functional activity assessed by means of the Blessed Scale (z = 2.958, p < .003). Women predominated in the group of nonconverters, and functional competence was more compromised in the group of converters versus the non-converters (Table 3).
Neuropsychological data Mann–Whitney’s U-test showed the existence of significant differences between the converters and non-converters in the following memory tests: Learning list A of the TAVEC (z = 4.579, p < .001), short delay free recall (z = 4.444, p < .001), short delay cued recall (z = 3.361, p < .001), long delay free recall (z = 4.424, p < .001), long delay cued recall (z = 4.795, p < .001), and in execution time of the Rey figure (z = 3.106, p < .002). The scores of the converters were significantly worse than those of the nonconverters in each of the above-mentioned cognitive tests (Table 3). Figure 1 presents the performance of the converters versus the non-converters.
Predictors of conversion To determine the predictive value of the neuropsychological tests used, we performed a binary logistic regression analysis in which we included all the measures of the
8
S. Garcıa-Herranz et al.
Table 3. Baseline sociodemographic and clinical data and neuropsychological performance of converters and non-converters
Sociodemographic variables Gender (female) Age (years) Formal education (years) Follow-up time (years) Geriatric Depression Scale, mean (GDS, Yesavage) Functional level: Blessed Dementia Rating (Blessed) Neuropsychological tests Cognitive status (MMSE Spanish version) Fluency Verbal fluency/p/ Category fluency/animals, clothing, plants, vehicles/ TAVEC Immediate free recall (list A) Immediate free recall (list B) Short delay free recall Short delay cued recall Long delay free recall Long delay cued recall Recognition TMT TMT A TMT B Alternating graphs and loops Quality Rey figure Time Quality Constructive praxias Quality Ideomotor praxias Quality
Mean (SD) Converters n = 24
Mean (SD) Non-converters n = 81
p-Value
12 (50.00%) 74.17 (5.89) 8.38 (6.39) 2.79 (1.14) 4.21 (3.40)
61 (75.30%) 71.83 (6.76) 7.96 (6.10) 3.35 (1.42) 3.52 (2.61)
.040 ns ns ns ns
1.45 (1.06)
0.83 (0.98)
.003
29.46 (3.83)
29.79 (3.61)
ns
22.00 (9.41) 46.54 (9.49)
22.30 (8.93) 48.63 (11.99)
ns ns
30.54 (11.72) 3.92 (1.90) 4.04 (2.75) 6.92 (2.85) 4.33 (3.52) 5.96 (3.04) 13.00 (2.41)
42.23 (10.51) 4.70 (2.05) 7.81 (3.34) 8.98 (2.90) 8.16 (3.59) 9.14 (2.80) 13.84 (2.27)
.001 ns .001 .001 .001 .001 ns
84.00 (38.32) 189.23 (98.96)
88.16 (51.28) 180.20 (86.05)
ns ns
2.58 (1.17)
2.47 (1.23)
ns
329.87 (140.22) 23.34 (9.24)
237.68 (100.90) 22.07 (9.75)
.002 ns
8.96 (2.67)
8.66 (2.42)
ns
17.33 (2.94)
17.37 (2.36)
ns
Note. TAVEC, Verbal Learning Test of the Complutense University. Statistically significant (p ≤ .05).
neuropsychological tests that found significant differences between converters and nonconverters at the first assessment. The analysis showed that only long delay cued recall (p < .001, OR = 0.586, 95% CI: 0.448–0.767) and execution time of the Rey figure (p < .005, OR = 1.008, 95% CI: 1.003–1.014) were statistically significant predictor variables; that is, these tests contribute the most to the prediction of conversion to dementia after an MCI diagnosis (Table 4). The effect size was very large for the delayed cued recall test (d = 1.326) and for the execution time of the Rey figure (d = 0.958). The
Neuropsychological predictors of conversion to AD 55
Non-converters
9
Converters
T-scores
50 45 40 35 30
Figure 1. Cognitive performance of converters versus non-converters. First evaluation.
Table 4. Multiple logistic regression model for prediction: Conversion/non-conversion Included variables Long delay cued recall Rey figure execution time
B coefficient
B (SE)
OR
95% CI of OR
Wald
df
p-significance
.535
.137
0.586
0.448–0.767
15.17
1
.001**
.008
.003
1.008
1.003–1.014
9.72
1
.005*
Excluded variables
Wald
df
p-significance
Short delay cued recall Total immediate free recall (list A) Long delay free recall Short delay free recall Blessed Dementia Rating (part A)
3.45
1
ns
1.27
1
ns
0.72 0.26 0.12
1 1 1
ns ns ns
Summary of the model Significant Omnibus test model: Hosmer-Lemeshow test: R2 Nagelkerke: .512 Total percentage correct classification: 87.9%
v2(2) = 36.89, p = .001** v2(8) = 7.07, p = .529 ns
Note. ns, non-significant (p > .50), *p < .005; **p < .001.
percentage of correct classification was 87.9%. So, almost 90% of the people were correctly diagnosed. As a measure of the diagnostic precision of the tests, we studied the ROC curves to identify the cognitive tests with greater predictive capacity, sensitivity, and specificity, and the optimal value at which one could consider deterioration compatible with dementia or probable AD. We also analysed the PPV and NPV. Figure 2 shows the ROC curves for long delay cued recall and execution time of the Rey figure, in which diagnostic groups studied – converters versus non-converters – are
10
S. Garcıa-Herranz et al.
Figure 2. Receiver operating characteristic (ROC) curves. Long delay cued recall (AUC = 0.839) and execution time of Rey figure test (AUC = 0.728).
Table 5. Sensitivity, specificity, positive (PPV), and negative predictive values (NPV) and cut-off points of long delay cued recall and execution time of Rey figure test Cognitive tests
Cut-off points Sensitivity (%) Specificity (%) Positive predictive value (PPV) (%) Negative predictive value PPN (%)
Long delay cued recall
Execution time of Rey figure
247 70 62 34.1 88
contrasted. In Table 5, we present the values of sensitivity and specificity, as well as the PPV and NPV for the different optimal cut-points for each of the above tests. The results of these tests are very sensitive. For long delay cued recall, we observe an area under the curve and a highly significant reference diagonal line (AUC = 0.839, 95% CI: 0.731–0.946, p < .001). For execution time of the Rey figure, we observe an area under the curve and a diagonal reference line (AUC = 0.728, 95% CI: 0.597–0.859), in this case, also very significant (p < .002). This indicates that the tests have good predictive or diagnostic capacity. With regard to the results obtained with the regression model using the sociodemographic predictors, follow-up time, and emotional state, bivariate analysis revealed the variables sex, v2(1) = 5.570, p < .018, and follow-up time (z = 1.979, p < .048) to be significant. However, in the logistic regression model factor, gender was the only significant factor (p < .021; OR = 3.050, 95% CI: 1.184–7.856), such that the risk of converting to AD is higher in men than in women.
Neuropsychological predictors of conversion to AD
11
Discussion In this study, the main goals were to analyse the predictive value of the cognitive tests included in the neuropsychological battery used for conversion to AD after an MCI diagnosis and, accordingly, to determine the cut-off point in each of the predictive tests. Our third goal was to analyse the influence of some sociodemographic variables – sex, age, schooling – and other variables such as follow-up time and emotional state. Regarding the first objective, our findings showed that some of the tests that assess episodic memory and visuospatial ability were the most predictive of conversion to dementia, specifically, the long delay cued recall test and the execution time of the Rey figure. Of these two tests, the most sensitive test with the greatest predictor capacity for conversion to dementia was long delay cued recall. The results found in the present work regarding the predictive value of episodic memory tests are consistent with other research (Albert, Blacker, Moss, Tanzi, & McArdle, 2007; Dierckx et al., 2009; Dubois et al., 2014; Egli et al., 2014; Hanseeuw & Ivanoiu, 2011; Ivanoiu et al., 2005; Sarazin et al., 2007; Silva et al., 2012; Wagner et al., 2012), which used similar tests and in the same way as those used in this work, showing that deterioration of the episodic memory is associated with probable AD. Dierckx et al. (2009) concluded that the delayed cued recall task was the best predictor of conversion to AD among MCI patients. Hanseeuw and Ivanoiu (2011) showed that the cued recall test could predict dementia between 5 and up to 10 years before its onset. Sarazin et al. (2007) showed that impairment of free recall, total recall, and index of sensitivity of cueing could identify prodromal AD in patients with MCI. In a similar vein, Wagner et al. (2012) found that deficits in memory tests are very closely related to the biomarkers indicative of AD in subjects with MCI and conclude that this finding complements the results of prospective clinical studies and provides empirical support ratifying that cued recall tasks can be considered an appropriate specific indicator of prodromal AD. Our results ratify what has already been described in the literature: Years before the evolution of AD, due to the atrophy of the temporal structures, people with MCI who evolve to dementia present memory deficits due to their greater difficulty in information coding and organization processes, which are revealed in the recall process, regardless of the conditions of facilitation through semantic cues, as reported by Carlesimo, Perri, and Caltagirone (2011). Moreover, visuospatial ability-related research has in general shown that the deterioration of this capacity is even evident years before the diagnosis of AD (B€ackman, Jones, Berger, Laukka, & Small, 2005; Laukka, Macdonald, Fratiglioni, & B€ackman, 2012). We also highlight that we observed significant differences between the groups studied in the time required to execute the Rey figure, a variable related to visuospatial skill and to different components of the executive function, such as planning, organization, and attentional capacity (Somerville, Tremont, & Stern, 2000; Watanabe et al., 2005). This suggests that the increase in the execution time of the Rey figure reflected in the converters may be related not only to visuospatial problems but also to attentional, planning, and organizational problems, related to a generalized decrease of information processing, rather than to specific deficits in certain aspects of visual processing determined by the quality of the execution of the Rey figure. Rapp and Reischies (2005) verified that older individuals with early dementia exhibit slower information-processing speed in attentional and executive tests. The increase in execution time reveals that alterations compatible with initial dementia are not limited only to mnestic performance, but instead visuospatial, executive, and attentional components are also affected (Salmon
12
S. Garcıa-Herranz et al.
& Bondi, 2009). It should not be forgotten that some people included in our study diagnosed as AD by the neurologist presented mixed pathology (probable AD with an associated vascular component), and, as pointed out by Bastos-Leite et al. (2007), in these cases, the executive function tests and delayed memory tests are affected. There is increasing evidence showing that people with memory deficits and alterations in other cognitive functions are at greater risk of developing AD than are people who present isolated memory deficits (Albert et al., 2011; Brandt et al., 2009; Dıaz-Mardomingo et al., 2010; Lonie, Herrmann, Donaghey, & Ebmeier, 2008; Ritchie, Artero, & Touchon, 2001; Sacuiu et al., 2009; Tabert et al., 2006). The results of this work reveal that visuospatial skills are still an important component of neuropsychological assessment to detect AD, not only because of their sensitivity for the detection of this neurocognitive pathology, but also because this domain is associated with other cognitive domains that are altered in probable AD. With regard to the study of the predictive validity of long delay cued recall, we obtained a cut-off point of
