Master Thesis (MT) Final Version – February 2018
Department of Psychology, University of Amsterdam
Semantic fluency performance in presymptomatic frontotemporal dementia: the prognostic value of a qualitative analysis
Student
Name : J. J. Altelaar
Address : Mauvestraat 46-2
Zip code and residence : 1073 RN Amsterdam
Telephone number : 06 - 46 25 12 70 Student ID Card number :10145052
E-mail address : j.j.altelaar@gmail.com Supervisors
Within program : Tim Ziermans
Specialization : Klinische neuropsychologie External supervisor(s), : Esther van den Berg
Lize Jiskoot
Second assessor : Thelma Schilt
Abstract
Frontotemporal dementia (FTD) is a neurodegenerative disease associated with behavioural disturbances, executive deficits, and language impairment. Familial FTD can be used to investigate the phase before the onset of clinical symptoms, the presymptomatic phase, and evaluate progression of the disease. A qualitative analysis of verbal fluency (VF)
performance has proved to be able to predict disease progression in Alzheimer Dementia (AD), and differentiate FTD phenotypes. In this study, we examined participants of the Frontotemporal Dementia Risk Cohort (FTD-RisC), a cohort consisting of 50 percent at-risk mutation carriers, by exploring qualitative performance in VF as stated by Ledoux et al. (2014; clustering of words, switching between clustered groups). We investigated qualitative decline between mutation carriers and healthy controls, and decline from the presymptomatic to the symptomatic stage in a four-year follow-up study. Presymptomatic MAPT mutation carriers (n= 20), GRN mutation carriers (n= 33) and healthy controls received VF as part of neuropsychological assessment test every two years. During the FTD-RisC project, there have been eight cases of conversion (5 MAPT, 3 GRN mutation carriers) to the symptomatic stage. Between group performance decline in healthy controls, mutation carriers and
converters were examined. There were no differences between healthy controls and mutation carriers concerning qualitative VF abilities. Converters’ performance declined compared to non-converters; they produced fewer words, a smaller number of clusters, less words within clusters, and a smaller percentage of words within clusters. Qualitative fluency scores failed to predict future conversion between two and four years before symptom onset. Number of generated words and clusters, as well as the total number of words within clusters, were predictive of conversion between two years before symptom onset and actual symptom onset. This study shows that qualitative fluency scores differed between presymptomatic mutation carriers and converters. Results from the VF task can be used to distinguish between healthy FTD mutation carriers and converters, but this study found no added clinical value of
qualitative subscores in addition to traditional fluency scores.
Table of Content ABSTRACT ... 2 INTRODUCTION ... 4 FAMILIAL FTD ... 4 PRESYMPTOMATIC PHASE ... 5 VERBAL FLUENCY ... 6
QUALITATIVE FLUENCY TASK ... 7
RESEARCH QUESTIONS ... 8 METHODS ... 10 PARTICIPANTS ... 10 MATERIALS ... 12 STATISTICAL ANALYSIS ... 13 RESULTS ... 15 DEMOGRAPHICS ... 15 INTERRATER RELIABILITY ... 15
MUTATION CARRIERS AND HEALTHY CONTROLS ... 16
CONVERTERS AND NON-CONVERTERS ... 16
CONVERSION CLASSIFICATION ... 22
DISCUSSION ... 24
Introduction
Frontotemporal dementia (FTD) is a neurodegenerative disease associated with asymmetrical cerebral hypometabolism and atrophy of the frontal and temporal lobes (Seelaar, Rohrer, Pijnenburg, Fox & van Swieten, 2010; Warren, Rohrer, & Rossor, 2013). The heterogeneous clinical spectrum includes behavioural disturbances, executive deficits, and language impairment (Seelaar, Rohrer, Pijnenburg, Fox, & van Swieten, 2010). There are multiple clinical variants of FTD: the behavioural variant of FTD (bvFTD), and the language variants of FTD, referred to as primary progressive aphasia (PPA). bvFTD accounts for about half of the cases and is characterised by, among other, disinhibition, disturbed social conduct, loss of empathy, emotional blunting, apathy, and disturbed executive functioning; as well as relatively intact memory and visuoconstruction (Bathgate, Snowden, Varma,
Blackshaw, & Neary, 2001; Rascovsky et al., 2011; Seelaar et al., 2010). The language variants of FTD are categorized into a semantic (svPPA) and non-fluent (nfvPPA) variant. Together with the logopenic (lvPPA) variant, which has been linked to Alzheimer (AD) pathology (Gorno-Tempini et al., 2011), these syndromes are referred to as PPA. Patients with svPPA mainly show impairments in confrontation naming and single-word
comprehension; nfvPPA patients primarily show agrammatism and effortful, halting speech, while persons with lvPPA have an impaired repetition of sentences and tend to have a slow rate of speech (Gorno-Tempini et al., 2011).
Familial FTD
A significant portion of patients with FTD have a positive family history with an autosomal dominant mode of inheritance. Although the degree of heritability widely differs between studies, most find a rate between 20 and 50 percent (Rohrer et al., 2015; Rohrer & Warren, 2011; Seelaar et al., 2010; Seelaar et al., 2008). This percentage differs between
FTD subtypes, whereby bvFTD is the most heritable (Pickering-Brown et al., 2008; Rohrer & Warren, 2011). Over 50 percent of patients with familial FTD have Progranulin (GRN), microtubule-associated protein tau (MAPT) mutations, or – the more recently discovered – chromosome 9 open reading frame 72 (C9orf72) repeat expansion (Boxer et al., 2009;
Dejesus-Hernandez et al., 2011; Pottier, Ravenscroft, Sanchez, & Rademakers, 2016; Seelaar et al., 2011; Snowden et al., 2015; van Swieten & Spillantini, 2007). Patients with a MAPT mutation have a mean age onset of 55 years and are most often characterised by behavioural changes (disinhibition, obsessive-compulsive, and stereotyped behaviours) and/or
parkinsonism (Ghetti et al., 2015; Seelaar et al., 2011; Snowden et al., 2015). The mean age of onset for GRN mutation carriers is around 60 years and varies between 35 and 89 years. These mutation carriers tend to show more apathetic behavioural changes (e.g., social
withdrawal), non-fluent aphasia, and possible hallucinatory or delusional symptoms (Seelaar et al., 2011; Snowden et al., 2015). C9orf72 patients, who have a disease onset around 55 years, show disinhibited behaviour as a primary feature. Furthermore, co-occurrence of motoric symptoms linked to amyotrophic lateral sclerosis (ALS) is a distinguishing feature (Dejesus-Hernandez et al., 2011; Hsiung et al., 2011; van Langenhove et al., 2013; Snowden et al., 2015).
Presymptomatic phase
Since mutation carriers ultimately will develop the disease, these persons can be screened in their presymptomatic phase to look for structural and functional changes in the brain (Dopper et al., 2013; Rohrer et al., 2015), changes in metabolism (Jacova et al., 2013), or cognitive distortions (Rohrer et al., 2015). In a longitudinal study, Jiskoot et al. (2016) found a presymptomatic decline in the cognitive domains of social cognition and memory, with the start of cognitive decline to occur eight years before estimated symptom onset. Since
the presymptomatic phase is the time when irreversible neuronal loss is minimal and cognitive functioning is still relatively intact (Rohrer et al., 2015), this would be the ideal moment for (preventative) treatment (Reiman et al., 2011). Furthermore, more knowledge about disease progression could be beneficial to developing treatment and diagnostics as well. It therefore becomes more and more important to accurately screen the presymptomatic phase and predict future symptom onset. Finding of markers which define this phase, are therefore of key importance. Since neuropsychological assessment is a relatively cheap and non-invasive method, there would be much benefit to gain from the establishment of a cognitive marker.
Verbal fluency
One of these cognitive markers may be found in the VF task. The VF task is important for serving as an indicator for the level of language, executive function, processing speed, and working memory (Auriacombe et al., 2006; Azuma, 2004). Since language and executive dysfunction distortions are important features of FTD pathology, this test could be beneficial to (FTD) diagnostics. For this task, participants are asked to generate as many words as possible in 60 seconds. Two versions of the task are used; a semantic version in which semantically related words like animals or professions have to be named; and a phonological version, in which all words start with a given letter. The task is easy to administer and takes approximately five minutes. Because of the fast-paced administration, low costs, and diagnostic utility (Auriacombe et al., 2006; Azuma, 2004) this task is widely used in both clinical practice and research settings.
Analysis of the fluency tasks by Marczinski and Kertesz (2006) showed a significant difference in number of words generated in a semantic fluency task between control
index, which measures the relative difference in letter fluency and semantic fluency scores, is able to differentiate between AD and FTD patients (Rascovsky, Salmon, Hansen, Thal, & Galasko; 2007).
In AD, the fluency task scores have been identified as an indicator of conversion to the symptomatic phase. Auriacombe et al. (2006) looked at the presymptomatic phase in AD and found that the number of words generated in the semantic fluency task is already at a
significantly lower level five years before disease onset. Amieva et al. (2005) found a declining performance in a four-component semantic fluency task (Isaac Set Test) starting at nine years before the clinical diagnosis, implicating that performance drops years before clinical symptom onset. In an FTD cohort, semantic fluency performance also showed presymptomatic decline in MAPT mutation carriers between baseline and two-year follow-up (Jiskoot et al., 2016).
Qualitative fluency task
Traditionally, studies on VF focus on the total number of words generated. However, there are also qualitative aspects of fluency results which may have additional diagnostic value. Qualitative aspects as ‘number of errors’ and ‘repeated words’ helped Marczinski and Kertesz (2006) to distinguish different dementia syndromes. Recent research by Van den Berg et al. (2017) showed promising results by using a qualitative analysis (based on Ledoux et al., 2014) of the fluency tasks. When performing a fluency task, people do not generate random words, but tend to produce spurts of words within semantically related subcategories (Troyer, 1997). By focusing on the number of subcategories people create (clustering), and how many times they switch between those categories (switching), the researchers were able to differentiate between bvFTD and PPA subtypes. PPA patients tended to generate a smaller number of clusters than bvFTD patients, while svPPA patients used more switches than
nfvPPA and lvPPA patients (Van den Berg et al., 2017). By using this method, Clark et al. (2016) were able to predict conversion from mild cognitive impairment (MCI) to clinical AD. Raoux et al. (2008) showed disease progression in AD is associated with a lower switching index, and were able to predict presymptomatic decline five years before conversion into AD by the amount of switches. The fact that this qualitative analysis shows predictive qualities in AD, and contributes to differential diagnostics in FTD, makes it an interesting approach for the characterization of the presymptomatic phase in FTD. Consequently, subtle changes in qualitative fluency scores could be a predictor of future conversion and give insight in the possible phenotype in which the disease will manifest itself.
Research Questions
Our aim in this paper is to establish a cognitive marker to predict future conversion from the presymptomatic phase to the symptomatic FTD phase by exploring qualitative performance in VF. Research has shown cognitive decline starts 5-10 years before estimated symptom onset (Jiskoot et al., 2016; Rohrer et al., 2015). Since earlier research was able to predict future conversion in AD by using qualitative performance in VF (Clark et al., 2016), this could also be the case for presymptomatic FTD. In this study, we aim to find parameters of future conversion in FTD mutation carriers via a qualitative analysis of the semantic fluency task. Therefore, we will investigate fluency performance on a presymptomatic cohort.
Firstly, we will investigate differences between mutation carriers and healthy controls. We will use variables derived from the semantic fluency task at baseline and two follow-up measurements (two and four years after baseline) to detect deterioration. Based on the results of earlier research by Jiskoot et al. (2016), where cognitive decline was shown years before
disease onset, we expect qualitative VF-performance decline in mutation carriers compared to healthy controls. Additionally, we will perform explorative analyses to look for gene-specific differences.
Secondly, we will we will investigate differences between converters and
non-converters (presymptomatic mutation carriers). Baseline and follow-up scores will be used to map fluency deterioration over time. Gene-specific differences will be evaluated. Semantic fluency abilities in converters are expected to show significantly more deterioration over time compared to non-converters. Fluency abilities which prove to be most vulnerable to disease progression could be potential indicators of future conversion.
Thirdly, we will examine whether we can use VF subscores to differentiate between converting and non-converting mutation carriers before symptom onset – therefore
predicting disease progression. Cluster size, number of clusters, and number of switches are expected to be predictors of conversion (i.e., will be able to differentiate between converters and non-converters at least two years before symptom onset).
Methods Participants
This retrospective study included 120 participants, participating in the FTD-RisC study at the Erasmus Medical Center between 2009 and 2017. Participants come from
pathologically conformed FTD families (MAPT, GRN, or C9orf72 mutations). These 50 percent at-risk mutation carriers underwent standardised assessment, consisting of neuropsychological testing, neurological examination, and magnetic resonance imaging (MRI) of the brain. DNA sequencing was performed at study entry. A history of neurological (n = 1), neurodegenerative (other than FTD) (n = 2) or major psychiatric disorders (n = 3), as well as a contraindication for MRI (n = 2) are considered exclusion criteria for participation. Furthermore, participants needed to be asymptomatic and at least 18 years at the start of the study. After exclusion, a total of 114 participants remained in this study, of which 74
participants participated in baseline and two follow-up measurements. The other participants were included in a later time period, and therefore did not take the follow-up 2 measurement yet. The total dataset consisted of 53 non-mutation carriers (46.5%; depicted as healthy controls) and 61 mutation carriers (53.5%). The latter group contains of 20 MAPT (32.8%), 33 GRN (54.1%), and 8 C9orf72 (13.1%) mutation carriers. Since the start of this project eight participants have converted to the symptomatic stage of the disease. Converters are 5 MAPT mutation carriers, who have all developed bvFTD. The remaining converters have a GRN mutation, one of them has also developed bvFTD, two have developed nfvPPA. Since all c9orf72 mutation carriers were included relatively late in the FTD-RisC project, none performed a follow-up 2 measurement or conversed to the symptomatic stage of the disease. Considering this, all c9orf72 mutation carriers were not included in our analyses. After this, a total of 103 participants remained in the study (Figure 1). At study entry, all participants
gave written informed consent. The study was approved by the Medical and Ethical Review Committee of the Erasmus Medical Center.
Figure 1. Classification of participants
Most participants underwent three measurements for the FTD-RisC cohort. The first measurement will be referred to as the ‘baseline-measurement’ (BL), the second
measurement as ‘Follow-up 1’ (FU-1) and the third measurement as ‘Follow-up 2’ (FU-2).
Converters.
Conversion is established when participants start developing symptoms based on the diagnostic criteria for PPA (Gorno-Tempini et al., 2011) or bvFTD (Rascovsky et al., 2011). Since participants are tested every two years, symptom onset develops somewhere between those measurements. However, the moment of diagnosis will be referred to as ‘symptom onset’. In this way, there will be three relevant measurement points for the converters.
• Symptom onset: the data of the visit the diagnosis was made. Eight converters are included. This measurement corresponds to the data of FU-2 for non-converters and healthy controls1.
• 2 years before symptom onset: the data of the visit two years before diagnosis. Eight converters are included. This measurement corresponds to the data of FU-1 for non-converters and healthy controls.
1 Since two participants converted between baseline and FU-1 measurement, their ‘symptom onset’ corresponded with ‘FU-1’, and ‘baseline measurement’ corresponded with ‘2 years before symptom onset’.
• 4 years before symptom onset: the data of the visit four years before diagnosis. Only six of the eight converters are included, since two converters converted between baseline and FU-1. This measurement corresponds to the baseline for non-converters and healthy controls
Materials
Semantic fluency measures were obtained through the VF task, which is part of the FTD-RisC neuropsychological protocol. For this task, participants are asked to generate as many words as possible in the category ‘animals’. The time limit for this task is 60 seconds. All scoring forms will be anonymously rated using the scoring method as developed by Ledoux et al. (2014), deducing six subscores from the raw data (Table 1). Each form will be scored by two independent raters to determine interrater reliability. Scoring forms with more than three unrecognizable words or symbols (e.g. checkmark instead of actual words) were excluded.
Table 1 Qualitative fluency subscores with description
Total number of words Sum of all correct words produced Number of clusters Number of multiword strings.2
Number of switches Number of transitions between clustered or nonclustered words.3
Total cluster size Sum of all clustered words
Mean cluster size Total cluster size divided by the number of clusters Percent words in clusters Percentage of total words4 produced that are members of
clusters
2 Every cluster contains at least two words; Smaller clusters within larger clusters aren’t added to the total number of clusters.
3 Switches can occur from one associative strategy or to another or to none at all. 4 Errors (non-animal words, repetitions, rule breaks) are included.
Research shows high interrater reliability in qualitative ratings of the semantic fluency, with correlations – dependent on evaluated subscore - between 0.896 and 0.998 (Ledoux et al., 2014). Research by Van den Berg et al. (2017) replicated this reliability, with a range between 0.90 and 1.0.
Statistical Analysis
Statistical analyses were performed using SPSS Statistics 24.0. (IBM Corp., Armonk, NY, USA) Cross-sectional analyses were performed by one-way analysis of variance (ANOVA) to compare differences in age and cognitive screening scores at baseline (frontal assessment battery (FAB), Beck’s depression index (BDI), MMSE). Significant results were elaborated on by Tukey post-hoc analysis. Mann-Whitney has been used to compare level of education (classification by Verhage (1983)) between healthy controls and mutation carriers. Differences in gender were analysed by means of a chi-squared test. Interrater reliability coefficients are constructed by means of intraclass correlation analysis
For the first and second hypothesis, to compare decline in fluency scores between mutation carriers and healthy controls or converters and mutation carriers respectively, a multilevel regression model was used. Covariates were integrated (age, level of education, total number of words) in our analysis. Since not every participant was able to participate in all measures (baseline and two follow-up measures), there were significant missing data in our research. Multilevel regression analysis corrects for these selective missing values - without having to delete incomplete data points. Furthermore, this analysis is able to create a hierarchical structure, were clustered variables can be ‘nested’ within other variables. In this way we created two levels: the participants constituted the upper level; their repeated measures the lower level. Cognitive decline was analysed per mutation (1) and clinical status (2).
1. Mutation status (MAPT mutation carrier, GRN mutation carrier, healthy control), time (baseline, FU-1, FU-2), and first order interactions were included on our analysis. Age, gender and educational level were entered as covariates. Fluency subscores were treated as outcome measures. Thereafter, converters were excluded from our sample and the analysis was rerun to control for decline based on clinical status instead of mutation. 2. Clinical status (converter, non-converter), time and first order interactions were included
on our analysis. Age, gender and educational level were entered as covariates. Fluency subscores were treated as outcome measures. Thereafter, the analysis was rerun twice to investigate genotypical or phenotypical decline. Firstly, the group of converters was split by mutation (MAPT, GRN), and secondly by phenotype (bvFTD, nfvPPA)
For our third hypothesis, to show predictive capabilities of qualitative VF to discriminate between converters and non-converters, we measured the area under curve (AUC) by performing ROC-analyses. Delta values, longitudinal decline or improvement of subscores, have been calculated between 4 and two years before symptom onset (Delta 4-2), between two years before and at symptom onset (Delta 2-O), and between four years before and at symptom onset (Delta Total). Sensitivity and specificity optimal cut-off were
calculated using Youden’s index (Youden, 1950) for each fluency subscore, at every time-point. Thereafter, we included all significant delta’s in a logistic regression analysis. The model was constructed using a forward stepwise method according to the likelihood ratio test, with age, sex and education as covariates. Effect size is reported by using Nagelkerke R2.
Explanatory variables are fluency subtests, outcome measure is conversion status. Multilevel analysis and ROC-analysis are corrected for multiple comparisons (Bonferroni).
Results Demographics
Cross-sectional analyses of demographic variables (Table 2) show a significant
difference in age between healthy controls, MAPT carriers and GRN carriers, F(2, 96) = 6.44, p = .002). Tukey HSD post-hoc analysis revealed that MAPT mutation carriers are
significantly younger at baseline in comparison to GRN mutation carriers (M = 10.64, SD = 3.03, p = .002) and healthy controls (M = 8.20, SD = 2.86, p = .014). Level of education (Verhage), gender and cognitive screening scores did not differ between all three groups (p > .05).
Table 2 Demographics in healthy controls and mutation carriers
Healthy Controls (n=46) GRN Carriers (n=33) MAPT Carriers (n=33) p
Gender* 20 (43%) 11 (33%) 11 (55%) .296 Age 49.65 ± 12.59 52.09 ± 8.13 41.45 ± 9.51 .002 Education 5.17 ± 1.02 5.61 ± 0.93 5.05 ± 1.50 .344 MMSE 29.10 ± 1.30 29.13 ± 1.28 29.59 ± 0.88 .595 FAB 17.07 ± 1.10 16.77 ± 1.69 17.12 ± 0.89 .493 BDI 4.63 ± 5.75 3.93 ± 6.78 4.67 ± 2.29 .834 * percent males Interrater Reliability
Fluency subscores as rated by two independent raters (Table 3) show high levels of interrater reliability (range 0.84 – 0.99)
Table 3 Interrater reliability correlations of verbal fluency subscores Semantic Fluency
Number of words 0.99
Number of clusters 0.89
Number of switches 0.92
Total cluster size 0.95
Mean cluster size 0.93
Percentage words in clusters 0.84
Mutation Carriers and Healthy Controls
No longitudinal significant decline in fluency subscores were measured in MAPT or GRN mutation carriers compared to healthy controls (p > 0.05; Table 4). Also, no
longitudinal decline was found in the group of mutation carriers compared to healthy controls (Figure 2), nor were there any subscores in which GRN mutation carriers declined more or less than MAPT mutation carriers. Excluding the converters from the mutation carrier groups did not lead to any significant result. Healthy controls showed a significant increase in number of clusters compared to their baseline measurement (β = 0.443, p = .040), but when controlling for multiple comparisons or total number of words this effect did not remain significant.
Converters and Non-converters
The total group of converters showed fluency subscore decline in total correct words (β = -5.143, p < .001), number of clusters (β = -1.332, p = .008), total cluster size (β = -5.379, p < 0.001), and percent of words in clusters (β = 7.374, p = .037) compared to non-converters (Table 5A; Figure 3). Explorative analyses showed GRN converters declined in number of words (β = -4.626, p = .005), number of clusters (β = -2.020, p = .010), total cluster size (β = -6.247, p = .002), and percent of words in clusters (β = -11.562, p = .025) compared to
non-converters. Total correct words (β = -5.869, p < .001) and total cluster size (β = -5.400, p = .002) declined in MAPT converters compared to non-converters. No differences in decline were found between GRN and MAPT converters. When group comparison was based on phenotype, the bvFTD converters showed decline in total correct words (β = -5.641, p < .001) and total cluster size (β = -5.684, p =.001) compared to non-converters (Table 5B; Figure 3). Converters with nfvPPA showed similar decline, with reduced scores in total correct words (β = -4.464, p < .001) and total cluster size (β = -5.061, p =.001) compared to non-converters. No differences in decline were found between bvFTD and nfvPPA converters. When
controlling for total correct words, the variables number of clusters and total cluster size did not remain significant in any of the comparisons.
Table 4 Verbal fluency performance in healthy controls, GRN carriers, and MAPT carriers.
Healthy controls (n = 46) GRN MC (n = 33) MAPT MC (n = 20)
Baseline β p Baseline β p Baseline β p
Total Correct Words 23.79 ± 4.37 0.626 .230 25.54 ± 3.30 0.238 .747 23.40 ± 3.59 -1.790 .084 Number of Clusters 6.32 ± 1.43 0.443 .040 6.18 ± 1.06 -0.246 .435 7.04 ± 1.08 -0.019 .965 Number of Switches 10.93 ± 1.93 0.330 .083 10.03 ± 1.76 -0.577 .215 10.20 ± 1.53 -0.208 .751 Total Cluster Size 19.36 ± 4.68 0.774 .195 21.92 ± 3.50 0.410 .637 19.77 ± 3.76 -1.234 .300 Mean Cluster Size 3.19 ± 0.82 -0.187 .147 3.81 ± 0.57 0.309 .059 2.84 ± 0.61 -0.037 .868 Percent in Clusters 75.16 ± 8.48 -0.896 .527 81.71 ± 7.31 0.998 .466 84.56 ± 6.72 0.745 .786
Figure 2. Qualitative fluency trends in healthy controls and mutation carriers BL FU-1 FU-2 20 21 22 23 24 25 26 Moment of measurement To ta l n u m b er o f w o rd s Healthy controls (n=50) Mutation carriers (n=53) BL FU-1 FU-2 7 8 9 10 11 Moment of measurement Number of switches Healthy controls (n=50) Mutation carriers (n=53) BL FU-1 FU-2 2.8 3.0 3.2 3.4 3.6 3.8 4.0 Moment of measurement
Mean cluster size
Healthy controls (n=50) Mutation carriers (n=53) BL FU-1 FU-2 5.0 5.5 6.0 6.5 7.0 7.5 Moment of measurement Number of clusters Healthy controls (n=50) Mutation carriers (n=53) BL FU-1 FU-2 19 20 21 22 23 24 Moment of measurement To ta l c lu st er s iz e Healthy controls (n=50) Mutation carriers (n=53) BL FU-1 FU-2 80 85 90 95 Moment of measurement
Percentage of words in cluster
Healthy controls (n=50) Mutation carriers (n=53)
Table 5A Verbal fluency performance in converters, specified by gene mutation.
Total group of converters (n = 8) GRN converters (n = 3) MAPT converters (n = 5)
Baseline β p Baseline β p Baseline β p
Total Correct Words 23.60 ± 3.38 -5.143 .000* 23.98 ± 3.59 -4.626 .005 22.80 ± 3.41 -5.869 .000* Number of Clusters 7.73 ± 1.03 -1.332 .008 7.85 ± 1.03 -2.020 .010 7.41 ± 1.03 -0.93 .149 Number of Switches 10.19 ± 1.98 -1.024 .088 9.97 ± 2.04 -0.922 .386 10.27 ± 2.05 -1.072 .222 Total Cluster Size 21.32 ± 2.75 -5.379 .000* 21.90 ± 3.55 -6.247 .002* 20.24 ± 3.32 -5.400 .002* Mean Cluster Size 2.78 ± 0.46 -0.169 .469 2.84 ± 0.47 -0.116 .685 2.75 ± 0.48 -0.431 .092 Percent in Clusters 92.77 ± 7.39 -7.374 .037 94.21 ± 7.58 -11.562 .025 90.72 ± 7.65 -4.636 .272 * Survived Bonferroni correction for Multiple Comparisons
Table 5B Verbal fluency performance in bvFTD converters, nfvPPA converters and non-converters.
bvFTD converters (n = 6) nfvPPA converters (n= 2) Non-converters (n = 45)
Baseline β p Baseline β p Baseline β p
Total Correct Words 22.84 ± 3.31 -5.641 .000* 23.61 ± 2.72 -4.464 .017 22.94 ± 3.64 1.221 .009 Number of Clusters 7.55 ± 1.02 -1.098 .076 6.88 ± 0.85 -1.506 .068 7.49 ± 1.04 0.523 .015 Number of Switches 10.10 ± 2.01 -0.899 .288 10.04 ± 1.40 -1.309 .296 9.72 ± 2.13 0.296 .319 Total Cluster Size 20.62 ± 3.29 -5.684 .001* 20.26 ± 2.82 -5.061 .023 20.72 ± 3.57 1.721 .003* Mean Cluster Size 2.76 ± 0.47 -0.403 .101 3.07 ± 0.47 -0.394 .926 2.83 ± 0.49 0.051 .542 Percent in Clusters 91.83 ± 7.56 -6.743 .105 85.51 ± 5.67 -6.935 .207 92.56 ± 7.90 1.648 .250 * Survived Bonferroni correction for Multiple Comparison
Figure 3. Qualitative fluency trends in non-converters and converters, specified by genotype. BL FU-1 FU-2 15 20 25 30 Moment of measurement To ta l n u m b er o f w o rd s Non-converters (n=45) Converters (n=8) MAPT converters (n=5) GRN converters (n=3) BL FU-1 FU-2 7 8 9 10 11 12 13 Moment of measurement Number of switches Non-converters (n=45) Converters (n=8) MAPT converters (n=5) GRN converters (n=3) BL FU-1 FU-2 2.5 3.0 3.5 4.0 Moment of measurement
Mean cluster size
Non-converters (n=45) Converters (n=8) MAPT converters (n=5) GRN converters (n=3) BL FU-1 FU-2 4 5 6 7 8 9 Moment of measurement Number of clusters Non-converters (n=45) Converters (n=8) MAPT converters (n=5) GRN converters (n=3) BL FU-1 FU-2 10 15 20 25 30 Moment of measurement To ta l c lu st er s iz e Non-converters (n=45) Converters (n=8) MAPT converters (n=5) GRN converters (n=3) BL FU-1 FU-2 60 70 80 90 100 Moment of measurement
Percentage of words in cluster
Non-converters (n=45) Converters (n=8) MAPT converters (n=5) GRN converters (n=3)
Conversion Classification
The delta value between two years before and at symptom onset (Delta 2-O) showed that total correct words, number of clusters, and total cluster size proved to be predictive of symptom onset (Table 6). Between four and two years before symptom onset, the relevant delta value (Delta 4-2) failed to predict future conversion5. Qualitative fluency deltas failed to significantly distinguish between MAPT and GRN mutation carriers or between bvFTD and nfvPPA phenotypes.
With all significant predictors, as stated in the Delta 2-O ROC-analyses, a predictive model was constructed to increase the number of accurate cases of future conversion.
Performance of this model has been evaluated by stepwise logistic regression. Since eight of our 53 mutation carriers converted, 84.9% of all the cases would be correctly predicted by chance only. The constructed model would outperform chance by significantly improving the correctly predicted cases to 88.7% (χ2= 17.32, p = .001). In this model, the total number of words was the only variable that contributed significantly to our prediction (B = 0.407, p = .026, R2 = 0.487). Since the increased performance of our model is based solely on number of
words generated, the addition of qualitative fluency scores did not improve conversion prediction performance.
5 Three Delta 4-2 values could not be calculated, since baseline fluency scores were not available. Therefore, the Delta 4-2 and Delta Total values have been established by using the data of the other five converters. Delta 2-O has been calculated with all eight converters.
Table 6 ROC analyses on fluency decline (delta scores).
AUC p-value Optimal cut-off Sensitivity (%) Specificity (%)
Total Correct Words Delta 4-2 .627 .351 - - -
Delta 2-O .899 .000 2,5 87,5 77,8
Delta Total .947 .001 3,5 100,0 82,2
Number of Clusters Delta 4-2 .345 .253 - - -
Delta 2-O .744 .029 1,5 75,0 71,1
Delta Total .820 .020 0,5 80,0 77,8
Number of Switches Delta 4-2 .425 .581 - - -
Delta 2-O .636 .223 - - -
Delta Total .687 .174 - - -
Total Cluster Size Delta 4-2 .579 .563 - - -
Delta 2-O .846 .002 2,5 87,5 77,8
Delta Total .878 .006 2,5 80,0 82,2
Mean Cluster Size Delta 4-2 .643 .293 - - -
Delta 2-O .539 .729 - - -
Delta Total .698 .150 - - -
Percentage in Clusters Delta 4-2 .407 .494 - - -
Delta 2-O .664 .143 - - -
Discussion
This study examined qualitative performance in a semantic VF task in FTD mutation carriers and non-carriers. No differences were detected between healthy controls and
mutation carriers concerning qualitative fluency abilities and deterioration of these abilities. In our large (N = 120) cohort, eight mutation carriers converted during this study, and the qualitative fluency performance of these converters declined significantly. Converters produced fewer words, developed a smaller number of clusters, while these clusters contained fewer words. Furthermore, the percentage of produced words which were imbedded in a cluster was lower compared to non-converters. When investigating gene-specific patterns for our explorative analysis, we found deterioration was primarily influenced by the group of GRN-mutation carriers, since MAPT converters only showed impaired performance in the production of words and size of the clusters. Qualitative fluency scores failed to predict future conversion between two and four years before symptom onset. Number of generated words and clusters, as well as the total number of words within clusters, were predictive of conversion between two years before symptom onset and actual symptom onset.
Previous research produced inconsistent findings concerning gradual presymptomatic changes in FTD mutation carriers. Some papers show cognitive decline years before
symptom onset (Dopper et al., 2013; Hallam et al., 2014; Jiskoot et al., 2016; Rohrer et al., 2015), while other papers report no changes (Janssen et al., 2005, Rohrer et al., 2008). In contrast to semantic fluency scores in patients with AD, who produce fewer words
(Auriacombe et al., 2006) and fewer switches (Raoux et al., 2007) five years before symptom onset, no early decline was found in our study. These findings suggest that mutation carriers’ and healthy controls’ performances are equivalent until a few years before conversion. The
trend we observe seems to implicate a fast-pace, instead of a gradual pathological process. In contrast to our prediction, presymptomatic mutation carriers did not differ in qualitative fluency scores compared to healthy controls, and therefore our hypothesis that subtle changes would be measurable years before symptom onset has not been confirmed. Considering the fact that structural changes are observable years before symptom onset, brain plasticity and cognitive resilience possibly withstand early neurobiological changes – keeping cognition relatively intact.
Different processes contribute to clustering and switching. While clustering is
primarily a semantic process which depends on language and memory skills and is mediated by the temporal cortex, switching seems to focus most on executive functioning mediated by the frontal lobe (Troyer et al., 1997; Mayr et al., 2002). Results from our current study, which include intact switching and declining clustering abilities in converters, suggest that early dysfunction in mutation carriers start with a problem in semantic rather than executive functioning. The fact that these declined clustering abilities are only observed in GRN mutation carriers could be explained by early language dysfunction (Seelaar et al., 2011) or relatively increased rate of brain atrophy associated with GRN mutation carriers (Rohrer, 2010).
In this study, we used two groups who do not show any symptoms: the healthy control group and the presymptomatic mutation carriers (non-converters). When comparing these two asymptomatic groups (healthy controls and non-converters), we see similar fluency scores. However, although the healthy control group does show increased scores over time, it only produces significantly more clusters. Presymptomatic carriers show a learning effect on number of words, number of clusters, and total cluster size, resulting in increasing the
number of words produced with double the rate at each measurement compared to the healthy controls. The fact that mutation carriers show a stronger learning effect than healthy controls
suggests that this group puts more effort into the tests. Since the results of the
neuropsychological assessment can have serious consequences for this group, such as an indication of possible symptom onset, this may boost their motivation to perform well.
The cognitive profiles which are created by neuropsychological assessment are able to differentiate between different phenotypes in dementia – creating an essential role for
neuropsychological assessment in contemporary diagnostics (Libon et al., 2007; Salmon & Bondi, 2009). Although neuropsychological assessment is well suited to map current cognitive status, predicting future cognition in genetic FTD is complicated. Despite the presence of subtle differences in neuropsychological scores (Jiskoot et al., 2016; Rohrer et al., 2015), there are no known neuropsychological tests which can adequately predict future conversion. This paper found declining fluency scores between two years before and actual symptom onset were predictive of conversion. However, regarding four years before symptom onset, no fluency variables were predictive of future conversion, which suggests that FTD mutation carriers tend to show a rather sudden decline in VF performance. These results are consistent with previous findings in familial FTD, which showed structural differences in neuroimaging can be found years before cognitive decline in the
presymptomatic phase sets in (Rohrer, 2015). If cognition is relatively spared in the
presymptomatic phase, this may not be the most promising clinical marker to predict future conversion. Neuropsychological testing could be limited to differentiate between
symptomatic and presymptomatic phase. Future research should focus primarily on structural and physiological biomarkers. Furthermore, neuropsychological tests which have strong predictive values at the moment of conversion should be distinguished to benefit diagnostics at symptom onset.
The VF task is a multifactorial task in which many cognitive skills and brain regions are involved (Troyer, 1997; Mayr et al., 2002). Due to these features, many diagnostic
properties have been attributed to this task (Duff et al., 2004) and have tempted researchers to state the VF task is a ‘one-minute mental state examination’ (Cummings, 2004) more
sensitive to early changes in dementia than the MMSE (Duff et al., 2004). Although
(qualitative) fluency scores proved able to distinguish between AD and subtypes of FTD and PPA (Van den Berg et al., 2017; Marczinski & Kertesz, 2006) and predict conversion from MCI to AD (Clark et al., 2016), in our paper these expected results could unfortunately not be extended to the presymptomatic phase of FTD. As stated above, this study suggests FTD mutation carriers show a rather sudden decline in VF performance, creating a narrow window of time for neuropsychological assessment to predict future conversion. Based on our
research, the fluency task did not meet the high expectations of Duff et al. (2004) considering sensitivity to early changes in dementia. While qualitative fluency scores become
differentiated in a further stage of the disease (Van den berg et al., 2017), we did not find evidence that qualitative subscores have additional value in predicting future conversion to traditional fluency scores. However, although qualitative subscores did not show added value to the VF task, a reduction in total number of words was highly predictive of conversion. This paper shows the regular interpretation of fluency scores, and diminished performance compared to earlier measurement, is an important indicator of conversion at symptom onset. Our data shows that this five-minute, easy-to-interpret test, contains such richness in
information that it should be considered an essential component in every neuropsychological consult.
Strengths of the present study include the large sample of participants incorporated in this longitudinal study. Earlier research focused on sporadic FTD or used a cross-sectional design, which makes it difficult to state causality. The longitudinal character of this study made it possible to compare neuropsychological scores witch each participant’s earlier performance, thereby controlling for interindividual differences. The fact that so many participants with genetic FTD were followed during a substantial amount of time, with a low dropout rate, creates an insightful and balanced dataset. Moreover, since our cohort consisted of 50 percent at-risk participants, the non-mutation relatives are a matched control group, whom show great similarity to our experimental group.
A few limitations in this study should be mentioned. Firstly, our group of converters is relatively small, making incidental fluctuations in fluency scores influence our analysis heavily. This reduces our power and makes it more difficult to measure subtle deterioration. Furthermore, although Ledoux et al. (2014) elaborated on the scoring procedure with clear instructions and examples, qualitative scoring of the fluency task requires some
interpretation. Moreover, though interrater reliability was still high, we did not find correlations as strong as previous papers (e.g. Van den Berg, 2017). To increase interrater reliability, we suggest the creation of a more exhaustive set of category examples, including more non-American-oriented categories. Finally, though conversion was based on diagnostic criteria for PPA (Gorno-Tempini et al., 2011) or bvFTD (Rascovsky et al., 2011) not all diagnostic criteria were met at diagnosis. Participants were labelled as converters when they were known mutation carriers and symptoms were observed - even when neuropsychological assessment did not reveal scores on the level of disorder. As result, the mutation carriers examined in our study were diagnosed earlier than regular-diagnosed cases of FTD. Since the time span between starting cognitive decline and diagnosis is shorter, this makes is more difficult to predict future conversion years before symptom onset and could be a possible
explanation for the limited predictive value of the VF task four years before symptom onset. Three qualitative fluency subscores were predictive of conversion between two years before symptom onset and moment of conversion. Because the moment of conversion is diagnosed quite early, these subscores might show an increased predictive value (i.e. be predictive earlier than two years before symptom onset) when implemented in a regular patient population.
Since there was no conversion or follow-up data available for C9orf72 mutation carriers, they were not included in this study. However, the group which carries this newly discovered gene may prove to show their own neuropsychological idiosyncrasies. As this longitudinal study develops and more c9orf72 data is accumulated, this group should also be incorporated into our analyses.
In conclusion, this study showed a significant decline in total number of words, total cluster size, and clustering abilities in a traditional VF task between presymptomatic at-risk gene mutation carriers and converters to FTD. Results from the VF task can be used to
distinguish between healthy mutation carriers and converters, but qualitative subscores do not provide added clinical value to traditional fluency scores. Cognitive assessment by VF in the presymptomatic phase was not able to predict future conversion, rendering it an instrument which has greatest value at the moment of diagnosis.
References
Amieva, H. (2005). The 9 year cognitive decline before dementia of the Alzheimer type: a prospective population-based study, Brain, 128(5), 1093–1101.
Auriacombe, S., Lechevallier, N., Amieva, H., Harston, S., Raoux, N., & Dartigues, J. F. (2006). A longitudinal study of quantitative and qualitative features of category verbal fluency in incident Alzheimer’s disease subjects: Results from the PAQUID study, Dementia and Geriatric Cognitive Disorders. Karger Publishers, 21(4), 260–266. Baldo, J. V., Schwartz, S., Wilkins, D., & Dronkers, N. F.. (2006). Role of frontal versus
temporal cortex in verbal fluency as revealed by voxel-based lesion symptom mapping, Journal of the International Neuropsychological Society, 12(6), 896–900. Bathgate, D., Snowden, J. S., Varma, A., Blackshaw, A., & Neary, D. (2001). Behaviour in
frontotemporal dementia, Alzheimer’s disease and vascular dementia, Acta Neurologica Scandinavica, 103(6), 367–378.
Boxer, A. L., Mackenzie, I. R., Boeve, B. F., Baker, M., Seeley, W. W., Crook, R., & Rademakers, R (2011). Clinical, neuroimaging and neuropathological features of a new chromosome 9p-linked FTD-ALS family, Journal of Neurology, Neurosurgery & Psychiatry, 82(2), 196–203.
Clark, D. G., McLaughlin, P. M., Woo, E., Hwang, K., Hurtz, S., Ramirez, L., & Apostolova, L. G. (2016). Novel verbal fluency scores and structural brain imaging for
prediction of cognitive outcome in mild cognitive impairment, Alzheimer’s and
Dementia: Diagnosis, Assessment and Disease Monitoring. Elsevier Inc., 2, 113–122. Cummings, J. L. (2004). The one-minute mental status examination, Neurology, 62, 534–535. DeJesus-Hernandez, M. M., Mackenzie, I. R., Boeve, B. F., Boxer, A. L., Baker,
Hexanucleotide Repeat in Noncoding Region of C9ORF72 Causes Chromosome 9p-Linked FTD and ALS, Neuron. Elsevier Inc., 72(2), 245–256.
Dopper, E. G., Rombouts, S., Jiskoot, L. C., Heijer, T., de Graaf, J. R., Koning, I., & Van Swieten, J. C. (2013). Structural and functional brain connectivity in
presymptomatic familial frontotemporal dementia, Neurology, 80, 814–823. Dopper, E. G. P., & Van Swieten, J. C. (2012). Primair progressieve afasie: fronto-
temporale lobaire degeneratie of de ziekte van Alzheimer?, Tijdschrift voor Neurologie en Neurochirurgie, 113(3), 101–107.
Duff Canning, S. J. S. J., Leach, L., Stuss, D., Ngo, L., & Black, S. E. (2004). Diagnostic utility of abbreviated fluency measures in Alzheimer disease and vascular dementia., Neurology, 62(4), 556–62.
Ghetti, B. S. J., Leach, L., Stuss, D., Ngo, L., & Black, S. E. (2015). Invited review:
Frontotemporal dementia caused by microtubule-associated protein tau gene (MAPT) mutations: A chameleon for neuropathology and neuroimaging, Neuropathology and Applied Neurobiology, 41(1), 24–46.
Gorno-Tempini, M. S. J., Leach, L., Stuss, D., Ngo, L., & Black, S. E. (2011). Classification of primary progressive aphasia and its variants, Neurology, 76(11), 1006–1014. Hallam, B. J. S. J., Leach, L., Stuss, D., Ngo, L., & Black, S. E. (2014). Early
neuropsychological characteristics of progranulin mutation carriers, Journal of the International Neuropsychological Society, 20(7), 694–703.
Hsiung, G. Y. R., Dejesus-Hernandez, M., Feldman, H. H., Sengdy, P., Bouchard-Kerr, P., Dwosh, E., & Mackenzie, I. R. A. (2012). Clinical and pathological features of familial frontotemporal dementia caused by C9ORF72 mutation on chromosome 9p, Brain, 135(3), 709–722.
M., & Mackenzie, I. R. (2013). Anterior brain glucose hypometabolism predates dementia in progranulin mutation carriers, Neurology, 81(15), 1322–1331. Jiskoot, L. C., Dopper, E. G. P., Den Heijer, T., Van Minkelen, R., & Van Swieten, J. C.
(2016). Presymptomatic Cognitive Decline in Familial Frontotemporal Dementia: a Longitudinal Study, Neurology, (87), 385–391.
Josephs, K. A., Whitwell, J. L., Jack, C. R., Boeve, B. F., Senjem, M. L., Baker, M., & Petersen, R. C. (2009). Voxel-based morphometry patterns of atrophy in FTLD with mutations in MAPT or PGRN, Neurology, 72(9), 813–820.
Ledoux, K., Vannorsdall, T. D., Pickett, E. J., Bosley, L. V., Gordon, B., & Schretlen, D. J. (2014). Capturing additional information about the organization of entries in the lexicon from verbal fluency productions, Journal of Clinical and Experimental Neuropsychology. Taylor & Francis, 36(2), 205–220.
Libon, D. J., McMillan, C., Gunawardena, D., Powers, C., Massimo, L., Khan, A., & Grossman, M. (2009). Neurocognitive contributions to verbal fluency deficits in frontotemporal lobar degeneration, Neurology. Lippincott Williams & Wilkins, 73(7), 535–42.
Libon, D. J., Xie, S. X., Moore, P., Farmer, J., Antani, S., McCawley, G., & Grossman, M. (2007). Patterns of neuropsychological impairment in frontotemporal dementia, Neurology, 68(5), 369–375.
Marczinski, C. A., & Kertesz, A. (2006). Category and letter fluency in semantic dementia, primary progressive aphasia, and Alzheimer’s disease, Brain and Language, 97(3), 258–265.
Mayr, U. (2002). On the dissociation between clustering and switching in verbal fluency: Comment on Troyer, Moscovitch, Winocur, Alexander and Stuss, Neuropsychologia, 40(5), 562–566.
Onyike, C. U., & Diehl-Schmid, J. (2013). The epidemiology of frontotemporal dementia, International Review of Psychiatry, 25(2), 130–137.
Pickering-Brown, S. M., Rollinson, S., Du Plessis, D., Morrison, K. E., Varma, A., Richardson, A. M. T., & Mann, D. M. A. (2008). Frequency and clinical
characteristics of progranulin mutation carriers in the Manchester frontotemporal lobar degeneration cohort: Comparison with patients with MAPT and no known mutations, Brain, 131(3), 721–731.
Pottier, C., Ravenscroft, T. A., Sanchez-Contreras, M., & Rademakers, R. (2016). Genetics of FTLD: overview and what else we can expect from genetic studies, Journal of
Neurochemistry, 138, 32–53.
Premi, E., Gazzina, S., Bozzali, M., Archetti, S., Alberici, A., Cercignani, M., & Borroni, B. (2013). Cognitive Reserve in Granulin-Related Frontotemporal Dementia:From Preclinical to Clinical Stages, PLoS ONE, 8(9), 1–8.
Raoux, N., Amieva, H., Le Goff, M., Auriacombe, S., Carcaillon, L., Letenneur, L., & Dartigues, J. F. (2008). Clustering and switching processes in semantic verbal fluency in the course of Alzheimer’s disease subjects: Results from the PAQUID longitudinal study, Cortex, 44(9), 1188–1196.
Rascovsky, K., Hodges, J. R., Knopman, D., Mendez, M. F., Kramer, J. H., Neuhaus, J., & Freedman, M. (2011). Behavioural variant of frontotemporal dementia, (September), 2456–2477.
Rascovsky, K. , Salmon, D. P., Hansen, L. a, Thal, L. J., & Galasko, D. (2007). Disparate letter and semantic category fluency deficits in autopsy-confirmed frontotemporal dementia and Alzheimer’s disease, Neuropsychology, 21(1), 20–30.
Reiman, E. M. (2011). Alzheimer’s Prevention Initiative: A Plan to Accelerate the
321–329.
Rogers, T. T., Hodges, J. R., Knopman, D., Mendez, M. F., Kramer, J. H., Neuhaus, J., & Freedman, M. (2006). Semantic memory in Alzheimer’s disease and the
frontotemporal dementias: a longitudinal study of 236 patients, Neuropsychology, 20(3), 319–335.
Rohrer, J. D., Nicholas, J. M., Cash, D. M., Van Swieten, J., Dopper, E., Jiskoot, L., & Rossor, M. N. (2015). Presymptomatic cognitive and neuroanatomical changes in genetic frontotemporal dementia in the Genetic Frontotemporal dementia Initiative (GENFI) study: A cross-sectional analysis, The Lancet Neurology, 14(3), 253–262. Rohrer, J. D., Ridgway, G. R., Modat, M., Ourselin, S., Mead, S., Fox, N. C., & Warren, J. D.
(2010). Distinct profiles of brain atrophy in frontotemporal lobar degeneration caused by progranulin and tau mutations, NeuroImage. Elsevier Inc., 53(3), 1070–1076. Rohrer, J. D., Warren, J. D., Barnes, J., Mead, S., Beck, J., Pepple, T., & Fox, N. C. (2008).
Mapping the progression of progranulin-associated frontotemporal lobar degeneration, Nature Clinical Practice Neurology, 4(8), 455–460.
Rohrer, J. D., & Rosen, H. J. (2013). Neuroimaging in frontotemporal dementia, International Review of Psychiatry, 25(2), 221–229.
Rohrer, J. D., & Warren, J. D. (2011). Phenotypic signatures of genetic frontotemporal dementia, Current Opinion in Neurology, 24(6), 542–549.
Salmon, D. P., & Bondi, M. W. (2009). Neuropsychological Assessment of Dementia, Annual Review of Psychology, 60(1), 257–282.
Seelaar, H., Kamphorst, W., Rosso, S. M., Azmani, A., Masdjedi, R., De Koning, I., & Van Swieten, J. C. (2008). Distinct genetic forms of frontotemporal dementia, Neurology, 71(16), 1220–1226.
Clinical , genetic and pathological heterogeneity of frontotemporal dementia : a review.
Snowden, J. S., Adams, J., Harris, J., Thompson, J. C., Rollinson, S., Richardson, A., & Pickering-Brown, S. (2015). Distinct clinical and pathological phenotypes in frontotemporal dementia associated with MAPT, PGRN and C9orf72 mutations, Amyotrophic Lateral Sclerosis and Frontotemporal Degeneration, 16(7–8), 497–505.
Troyer, A. K., Moscovitch, M., & Winocur, G. (1997). Clustering and switching as two components of verbal fluency: evidence from younger and older healthy adults, Neuropsychology, 11(1), 138–146.
Van den Berg, E., Jiskoot, L. C., Grosveld, M. J. H., Van Swieten, J. C., & Papma, J. M. (2017). Qualitative assessment of verbal fluency performance in frontotemporal dementia, Dementia and Geriatric Cognitive Disorders, 35–44.
Van Langenhove, T., van der Zee, J., Gijselinck, I., Engelborghs, S., Vandenberghe, R., Vandenbulcke, M., & Van Broeckhoven, C. (2013). Distinct Clinical Characteristics of C9orf72 Expansion Carriers Compared With GRN, MAPT, and Nonmutation Carriers in a Flanders-Belgian FTLD Cohort, JAMA Neurology, 70(3), 365.
Van Swieten, J., & Spillantini, M. G. (2007). Hereditary frontotemporal dementia caused by Tau gene mutations, Brain Pathology, 17(1), 63–73.
Verhage, F. (1983). Het coderen van het opleidingsniveau voor researchdoeleinden
[Educational classification system for research purposes: revised version]. Internal publication Groningen, the Netherlands: Academic Hospital Groningen, State University Groningen.
Warren, J. D., Rohrer, J. D., & Rossor, M. N. (2013). Frontotemporal dementia, BMJ, 347(12), 4827.
Whitwell, J. L., Josephs, K. A., Rossor, M. N., Stevens, J. M., Revesz, T., Holton, J. L., & Warren, J. D.. (2005). Magnetic Resonance Imaging Signatures of Tissue Pathology in Frontotemporal Dementia, Archives of Neurology, 62(9), 1402.
Whitwell, J. L., Weigand, S. D., Boeve, B. F., Senjem, M. L., Gunter, J. L.,
DeJesus-Hernandez, M., & Josephs, K. A. (2012). Neuroimaging signatures of frontotemporal dementia genetics: C9ORF72, tau, progranulin and sporadics, Brain. Wiley-Interscience, New York, 135(3), 794–806.
