Alzheimer’s & Dementia 9 (2013) S84–S94
Sensitivity and specificity of ventromedial prefrontal cortex tests in behavioral variant frontotemporal dementia Maxime Bertouxa,b,c,d,*, Aurelie Funkiewiezc,d, Claire O’Callaghane,f, Bruno Duboisa,b,c,d, Michael Hornbergere,f a Sorbonne Universite - Paris 6, Paris, France Institut du Cerveau et de la Moelle Epiniere, UMRS 975, Paris, France c Institut de la Memoire et de la Maladie d’Alzheimer (IMMA), H^opital de la Pitie-Salp^etriere, Paris, France d Reference Centre on Rare Dementias, H^opital de la Pitie-Salp^etriere, Paris, France e Neuroscience Research Australia, Sydney, NSW, Australia f School of Medical Sciences, University of New South Wales, Sydney, NSW, Australia b
Abstract
Background: Behavioral variant frontotemporal dementia (bvFTD) is characterized by early and substantial ventromedial prefrontal cortex (VMPFC) dysfunction. To date, however, there is no consensus regarding which tests are most sensitive and specific to assess VMPFC dysfunction in this condition. Methods: In this study we compared the sensitivity and specificity of four common VMPFC specific tests (Mini-SEA, Go/No-Go Subtest of the Frontal Assessment Battery, Reversal-Learning Test, and Iowa Gambling Task) at first clinic presentation in two neurodegenerative cohorts (bvFTD, Alzheimer’s disease) and age-matched, healthy controls. Results: We found that the Mini-SEA, evaluating theory of mind and emotion processes, emerged as the most sensitive and specific of the VMPFC tests employed. The Mini-SEA alone successfully distinguished bvFTD and Alzheimer’s disease (AD) in .82% of subjects at first presentation. Similarly, the FAB Go/No-Go and Reversal-Learning Tests also showed very good discrimination power, but to a lesser degree. The Iowa Gambling Task, one of the most common measures of VMPFC function, was the least specific of these tests. Conclusion: Sensitivity to detect VMPFC dysfunction was high across all test employed, but specificity varied considerably. The Mini-SEA emerged as the most promising of the VMPFC-specific diagnostic tests. Clinicians should take into account the variable specificity of currently available VMPFC tests, which can complement current carer-based questionnaires and clinical evaluation to improve the diagnosis of behavioral dysfunctions due to VMPFC dysfunction. Ó 2013 The Alzheimer’s Association. All rights reserved.
Keywords:
Ventromedial prefrontal cortex; bvFTD; Social cognition; Alzheimer’s disease; Frontotemporal dementia
M.B. reports no conflicts of interest, no financial interests, and no disclosures. He was supported by the French Ministry of Defence and the National Centre for Scientific Research (CNRS) during his PhD. A.F. reports no conflicts of interest, no financial interests, and no disclosures. C.O’C. reports no conflicts of interest, no financial interests, and no disclosures. B.D. reports no conflicts of interest and no financial interests. He has consulted or served on advisory board for Bristol-Myers Squibb, Roche, Elan, Eli Lilly, Eisai, and Janssen. His institution has received grants from Novartis and Sanofi-Aventis. M.H. reports no conflicts
of interest and no financial interests. He is editorial board member of the Journal of Alzheimer’s Disease, Dementia & Cognitive Geriatrics, and the American Journal of Neurodegeneration. He receives grants and fellowships from the Australian government funded Australian Research Council (ARC) and National Health and Medical Research Council (NHMRC). *Corresponding author. Tel.: 133675508822; Fax: 133142167504. E-mail address:
[email protected]
1552-5260/$ - see front matter Ó 2013 The Alzheimer’s Association. All rights reserved. http://dx.doi.org/10.1016/j.jalz.2012.09.010
M. Bertoux et al. / Alzheimer’s & Dementia 9 (2013) S84–S94
1. Introduction Frontotemporal dementia (FTD) encompasses a spectrum of early-onset neurodegenerative conditions, of which the behavioral variant (bvFTD) is the most commonly occurring subtype [1]. BvFTD is clinically characterized by progressive changes in personality and impaired social interaction, with prominent symptoms, including disinhibition, apathy, loss of empathy, stereotyped or compulsive behaviors, dietary changes, and self-care decline [1]. Clinical symptoms in bvFTD have mostly been related to progressive hypometabolism and atrophy of frontal and polar temporal lobes [2,3], particularly in the medial prefrontal and orbitofrontal regions (collectively known as the ventromedial prefrontal cortex [VMPFC]), which are known to be affected in the very early disease stages of bvFTD [4]. Importantly, VMPFC dysfunction has been shown to covary with some of the most prevalent behavioral symptoms in bvFTD, such as antisocial behavior and disinhibition [5], as well as loss of empathy [6]. These anatomoclinical correlations in bvFTD are consistent with findings in human and animals with VMPFC lesions, which show deficits in emotion processing, learning, social functioning, and behavioral flexibility [7]. For example, macaques with lesions to the VMPFC are impaired at learning reward contingencies, particularly when the reward contingencies change, such as on reversal-learning tasks [7]. VMPFC-lesioned monkeys also show inhibitory deficits on motor-response tasks [8], as well as emotional changes (e.g., decreased aggression) [7]. Reversal-learning, inhibitory, and emotional deficits have also been found in humans after VMPFC damage, and features such as impulsiveness, impaired decisionmaking, reduced empathy, and lack of affect are commonly observed [9–12]. The lesion findings have been further corroborated by recent functional neuroimaging studies, showing that the VMPFC is involved in reversal-learning [13], theory of mind [14], inhibition [15], and emotion processing [16]. Diagnostic criteria for bvFTD have been recently revised [17], but despite the prevalent and early dysfunction of the VMPFC in bvFTD, such as social and emotional deficits [18,19], dysfunctions of this brain region in bvFTD have not been taken into account in the new diagnostic criteria. Yet, cognitive tests tapping into its functions could allow a more objective evaluation of observed symptoms and, likewise, a better characterization of bvFTD [20]. By contrast, current assessments of VMPFC function are mostly subjective as they rely on carer information and clinical evaluations. At present, only a few validated tests are available to tap into VMPFC function; however, as these are being increasingly utilized in bvFTD, it is becoming apparent that such measures are crucial to better characterize cognitive dysfunction in the disease and enable more accurate diagnosis. For example, the Hayling Test [21], an objective measure of inhibitory functioning, has shown good sensitivity and specificity to discriminate bvFTD from elderly controls and Alzheimer’s disease (AD) patients [22], and has been re-
S85
lated to the patients’ VMPFC atrophy [23]. Theory of mind assessments, such as the Faux-Pas Test [24], have also showed good sensitivity and discrimination power to identify bvFTD compared with AD [25,26] or major depressive disorder [27] and have been shown to correlate with VMPFC dysfunction in bvFTD [19]. Two classic measures of VMPFC integrity—the Reversal-Learning Test and Go/No-Go Subtest—have also been shown to be sensitive to bvFTD impairments [25–30]. Finally, the Iowa Gambling Task (IGT) [31], one of the most commonly used clinical tests to detect VMPFC dysfunction, has been used to evaluate decisionmaking processes in bvFTD, with a reportedly good ability to identify impairments in bvFTD patients and differentiate them from healthy controls [25]. To date, no study has compared the sensitivity and specificity across commonly used VMPFC tests in relation to bvFTD. Using a sample of bvFTD patients with comparison groups of AD patients and healthy controls the aim of the current study was to evaluate the sensitivity and specificity of four tests related to VMPFC functioning: the Go/No-Go Subtest of the Frontal Assessment Battery (FAB) [32]; a computerized version of the Reversal-Learning Test from Rolls et al [9]; the abbreviated version of the Social Cognition and Emotional Assessment (mini-SEA) [27]; and the IGT [31]. We hypothesized that all these tests should be sensitive to detect VMPFC dysfunction in bvFTD but would vary in their specificity compared with AD patients and healthy controls. 2. Methods 2.1. Subjects Twenty bvFTD patients; 20 AD patients, in the early or moderate stages of the disease; 30 age- and educationmatched control subjects; and 16 young control subjects were recruited for the study. All bvFTD and AD patients were seen and evaluated at the Memory and Alzheimer Institute of the Pitie-Salp^etriere Hospital, Paris, France. The final diagnosis was established by experts after multidisciplinary clinical synthesis based on neuropsychologic, neurologic, biologic, and neuroimaging evidence. bvFTD patients were enrolled according to the Lund and Manchester criteria for bvFTD diagnosis [33] and fulfilled the revised diagnostic criteria recently proposed [17]. All patients presented with prominent changes in personality and social behavior that were established from interviews with their caregivers. AD patients were enrolled according to the revised NINCDS-ADRDA criteria [34]. They presented with a prominent history of episodic memory impairment with temporal and spatial disorientation and had a Clinical Dementia Rating (CDR) score 0.5. All patients also underwent comprehensive neuropsychologic examination. Clinical magnetic resonance imaging (MRI) and/or single-photon emission computed tomography (SPECT) scans were performed for all patients and showed
S86
M. Bertoux et al. / Alzheimer’s & Dementia 9 (2013) S84–S94
frontotemporal atrophy or/and hypoperfusion for bvFTD patients and medial temporal lobe atrophy for AD patients. Patients were excluded from the study if they presented with any of the following: (1) substantial language production and comprehension deficits suggesting progressive nonfluent aphasia or semantic dementia; (2) comorbid medical conditions that could interfere with cognitive functioning; (3) vascular lesions validated by MRI or neurologic history suggesting vascular dementia; and (4) motor-neuron disease. To improve diagnostic accuracy, all patients had at least an 18-month follow-up at the National Reference Centre for Rare Dementias or in the Resource & Research Memory Centre of the Pitie-Salp^etriere Hospital, in order to validate the diagnosis according to their clinical evolution. Patients were compared with 30 age- and educationmatched healthy controls. They were either spouses of unrelated patients consulting in our center or recruited from hospital advertisements. Controls were included according to the following criteria: (a) Mini-Mental State Examination (MMSE) score 27/30 and FAB 16/18; (ii) no history of neurologic or psychiatric disorders; (iii) no memory complaints or cognitive impairments; and (iii) no history of drug abuse. Finally, 16 young healthy controls were also recruited for the study via advertisements in the hospital. The inclusion of this group was a post hoc decision to verify our IGT results in the elderly controls. The same inclusion criteria as for the older control group applied to the younger controls. For patients, all clinical data were obtained during routine clinical work-up in the neurology departments and were extracted solely for the purpose of this study. Only data from the first clinical presentation of the patients were included in all analyses. Thus, according to French legislation, explicit informed consent was waived. However, regulation concerning electronic filing was followed, and both patients and their relatives were informed that individual data may be used in retrospective clinical research studies. For healthy control subjects, the study was approved by the ethics committee for the protection of persons of the Pitie-Salp^etriere Hospital. All controls received oral and written information and we obtained a signed informed consent form before their participation. 2.2. VMPFC-sensitive tests We administered a battery of VMPFC-sensitive clinical tests. Older controls and patients were tested on all VMPFC-sensitive tests. The younger controls were only tested on the Go/No-Go Subtest and IGT. 2.3. Go/No-Go Subtest (FAB) The score from the Go/No-Go Subtest of the FAB [32] was extracted to obtain a measure of inhibitory functioning. The Go/No-Go followed the Conflicting Instructions Subtest in the FAB, in which participants were trained to give a particular response to a specific stimulus (e.g., tapping one time
when the examiner taps twice), followed by learning a new association (e.g., not tapping when the examiner taps twice) during three more training trials. Then, in the Go/No-Go, the subject had to tap once when the examiner tapped once and was to inhibit the response by not tapping when the examiner tapped twice, for 5 of 10 trials. A score of 3 was given if the participants committed no error. A score of 2 was given if the participants committed one or two errors. If the participants committed more than two errors, a score of 1 was given. A score of 0 was given if the participants imitated the examiner tapping on at least four occasions. 2.4. Mini-Social Cognition and Emotional Assessment (Mini-SEA) The Mini-SEA [27] taps into social cognition and emotion disturbances. It consists of two subtests and provides two weighted composite scores: (1) a facial emotion recognition test, scored from 0 to 15, in which participants must identify which emotion is being expressed (happiness, surprise, neutral, sadness, disgust, anger, and fear); (2) a shortened version of the Faux-Pas Recognition Test [24], scored from 0 to 15, which evaluates theory of mind, where participants must detect and explain social inconveniences through short stories. The overall Mini-SEA composite score is calculated by adding the two subscores, and scored from 0 to 30. For further details see Funkiewiez et al [26]. 2.5. Reversal-Learning Test The Reversal-Learning Test is a visual discrimination reversal/extinction task adapted from Rolls et al [9]. It measures participants’ decisionmaking when contingencies and rewards change. Briefly, participants learned through successive trials to choose one visually colored stimulus, which resulted in a point gain. At the same time, participants were required to withhold a response when a different stimulus appeared; otherwise, a point was lost. When participants reached a learning criterion of 9 correct responses within a sequence of 10 consecutive trials, the contingencies were unexpectedly reversed; that is, the previously correct and rewarded stimulus was now incorrect and resulted in point loss. In the third phase of the test, called “extinction,” points could only be won by refraining from choosing any of the stimuli (i.e., participants are not allowed to respond to the critical stimuli). During all test phases, only two different stimuli were presented. The rewards were presented in a visual (“you won” or “you lost”) and auditory manner by a pleasant or unpleasant noise. Scores for the Reversal-Learning task are: number of errors committed before successfully reversing the rule (“reversal errors”); the number of trials needed to reach the reversal phase (“reversal trials”); and the number of reversals (“rules reversed”) during the whole test. At the end of the test, a debriefing interview was performed to verify whether the participants understood how to
M. Bertoux et al. / Alzheimer’s & Dementia 9 (2013) S84–S94
successfully perform in this task. The participants were asked the following question: “Can you tell me what should be done in this game to maximize your winnings?” Participants were considered fully aware of the test rule when they indicated that when one stimulus must be chosen, the other one must be avoided until the rule was reversed. After rule reversal, the previously unrewarded stimulus needed to be chosen to gain points. When the participants gave only a partial explanation (e.g., “I win when I choose the blue square but lose when I choose the green one”), they were considered partially aware. Participants who were not able to indicate how to successfully solve the task were considered as not consciously aware of the rule of the Reversal-Learning Test. 2.6. Iowa Gambling Task (IGT) The IGT is a decisionmaking task. The computerized version of the IGT was administered according to the standard protocol described by Bechara et al [10]. Participants made 100 card selections from four decks of cards in an order of their choosing. They were free to choose from any deck at any time. Each time a card was selected, the participant received a monetary reward; however, at unpredictable times, some selections also resulted in monetary punishment. In the IGT, decks A and B are considered disadvantageous, as they have large gains, but also large punishments, resulting in a net loss of money; however, decks C and D are advantageous, as they have small gains, but also small punishments, resulting in a net gain of money. Participants were not informed about the specific differences between decks or that the task would be terminated after 100 card selections. At the end of the test, a debriefing session was performed to assess whether participants were (i) full, (ii) partial, or (iii) not consciously aware of the rule of the test. Awareness of the test rule was established by asking the participants the following question: “Can you tell me what should be done in this game to maximize your profit?” The participants were considered fully aware of the test rule when they indicated that the A and B decks must be avoided, whereas the C and D decks should be chosen. However, when the participants gave only a partial explanation (e.g., “the B decks were dangerous”), or did not answer the first question, they were prompted with a second question: “Did you perceive any differences between the four decks of cards?” If the participants answered with the full rule, they were considered fully aware again. By contrast, when only a partial explanation was given, they were considered partially aware. Finally, participants unable to indicate what the rule of the test was were considered as not consciously aware of the IGT test rule. The debriefing information for AD patients was not available, and only 15 of the 20 AD patients passed this test. Following convention, IGT performance was calculated by dividing the 100 trials into 5 blocks of 20 card selections each. For each block, a net score for the IGT was calculated (see Torralva et al [25]) by subtracting the number of card selections from the “disadvantageous” decks (decks A and
S87
B) from the number of card selections from the “advantageous” decks (decks C and D) and. Scores .0 indicate that more advantageous cards were selected than disadvantageous cards, resulting in a net gain of money, whereas scores ,0 indicate that more disadvantageous cards were selected, resulting in a net loss of money. Importantly, test order (Go/No-Go [FAB], IGT, MiniSEA, Reversal-Learning Test) was randomized across participants. For controls only, the Go/No-Go score was already established in pretest screening. 2.7. Statistical analysis Statistical analysis was performed using SPSS, version 20 (IBM, Armonk, NY). The normality of demographic and neuropsychologic data was assessed with the Shapiro-Wilk test, and they were not normal. Thus, we used a nonparametric statistical Kruskal-Wallis test to compare the three groups (except for IGT analysis, where we used classic univariate analysis of variance [ANOVA]) followed by the MannWhitney U-test to compare groups (2 ! 2). For the correlations, we used Spearman’s rank correlation coefficient and applied Bonferroni’s correction for multiple measures (a significance threshold of P 5 .008 was then retained). For the IGT, data were analyzed using a mixed 4 ! 5 ANOVA with one between-subjects factor (“group”: bvFTDs, ADs, old control subjects, young control subjects) and one within-subjects factor (“block”: trial blocks in the IGT, in sets of 20 trials/block). For each group, we also performed a 3 ! 5 ANOVA with one between-subjects factor (“rule knowledge”: full, partial, none) and one withinsubjects factor (“block”) and a 3 ! 1 ANOVA with two between-subjects factors (“level of conscious knowledge” and “IGT score”) in each group to assess the interaction between IGT score (across trials or for the total score) and conscious knowledge of the rule, except in the AD group, where no debriefing information was available. 3. Results 3.1. Demographics and neuropsychologic background data Demographics of all participants are shown in Table 1. The four groups were not significantly different with regard to educational level. Age was not significantly different between bvFTD, AD, and older controls (P . .1). MMSE and FAB scores were significantly higher in younger and older control groups compared with the bvFTD patients (K 5 36.82, P , .001 and K 5 45.38, P , .001, respectively) and AD patients (K 5 36.82, P , .001 and K 5 45.97, P , .001), with no difference between younger and older participants (P ..1). Importantly, the two patient groups did not differ on the general neuropsychology score (MMSE), but were significantly different on the FAB, with bvFTD patients showing significantly more frontal dysfunction than AD patients (Z 5 22.57, P , .01).
S88
M. Bertoux et al. / Alzheimer’s & Dementia 9 (2013) S84–S94
Table 1 Demographic characteristics of the three groups Demographics
bvFTD patients
AD patients
Old controls
Young controls
N Age at testing, years Education, years Gender ratio (F/M) MMSE (total score) FAB (total score) Disease duration, years
20 69.16 (10.59) 9.9 (5.12) 6/14 23.56 (4.5) 12.6 (2.7) 3.34 (2.55)
20 69.85 (11.3) 11.3 (4.9) 9/11 24 (1.8) 15.3 (1.7) 3.2 (1.3)
30 67.23 (8.68) 10.7 (3.71) 15/15 28.96 (0.83) 17.25 (0.93) —
16 26.18 (6.7) 10.8 (3.5) 8/8 29.21 (0.65) 17.4 (0.82) —
NOTE. Data expressed as number of subjects or mean (SD).
3.2. Go/No-Go Test (FAB) Inhibitory function as measured on the FAB Go/No-Go Subtest (Table 2 and Figure 1) revealed that older controls and young controls were significantly more successful in suppressing a prepotent motor response than bvFTD patients (Z 5 24.51, P ,.001 and Z 5 23.45, P ,.001, respectively) and AD patients (Z 5 23.45, P , .001 and Z 5 22.59, P , .01). There was no difference between old and young controls for the Go/No-Go score (P . .1), and no difference between bvFTD and AD patients (P . .1).
with bvFTD patients (Z 5 25.16, P , .001 and Z 5 24.53, P , .001, respectively) (Table 2 and Figure 1). More precisely, the two subscores of the Mini-SEA (Faux-Pas and Facial Emotion Recognition) were both significantly higher in the older control group (Z 5 24.17, P , .001 and Z 5 25.20, P , .001, respectively) and in the AD group (Z 5 24.01, P , .001 and Z 5 23.88, P , .001, respectively) compared with the bvFTD group. Despite old controls scoring significantly better than AD patients on Facial Emotion Recognition (Z 5 22.43, P , .05), there was no difference on the Faux-Pas or on the total score of the Mini-SEA between these two groups.
3.3. Mini-SEA Results on the Social Cognition and Emotion Test (MiniSEA) showed that older controls and AD patients scored significantly better on the Mini-SEA composite score compared
3.4. Reversal-Learning Test In the Reversal-Learning Test, old controls committed fewer overall errors than bvFTD (Z 5 22.89, P , .01)
Table 2 Scores of the IGT, Mini-SEA, Reversal-Learning Test, and Go/No-Go Subtest (FAB) for the three groups bvFTD patients Go/No-Go Subtest (FAB) (/3) Mini-SEA (/30) Short Faux-Pas Test (/15) Facial Emotion Recognition (/15) Reversal-Learning Test Reversal errors Rules reversed (/3) Iowa Gambling Task Total net score Total A and B selection Total C and D selection Block 1–20 net score Block 21–40 net score Block 41–60 net score Block 61–80 net score Block 81–100 net score Rule knowledge Full Partial None NOTE. Data expressed as mean (SD). *P , .05. y P , .001 (vs bvFTD). z P , .05. x P , .001 (vs AD). { P , .01 (vs old controls). # P , .05 (vs young controls).
AD patients
Old controls
Young controls
y,x
3 (0)y,z
0.55 (0.99) 18.1 (5.57) 9.67 (2.8) 9.35 (1.8)
2.55 (0.69) 25.15 (1.34)y 13.5 (1.37)y 11.64 (1.28)y
3 (0) 25.77 (1.9)y 13.21 (1.44)y 12.55 (1.15)y,z
5.6 (4.1) 1.37 (1.18)
4.43 (3.6) 2.36 (0.7)*
2.7 (2.9)*,z 2.16 (1.11)y
4.9 (23.24) 47.05 (11.7) 51.95 (12.7) 20.2 (6.4) 20.53 (4.7) 0.9 (6.7) 0.3 (7.8) 4.6 (8){
4.9 (23.48) 47.80 (11.8) 52 (11.9) 0.4 (6.8)*,# 0.13 (3.7) 1.47 (4.2) 2.26 (8.1) 0.67 (6.2)
23.2 (22.91) 51.8 (11.23) 48.27 (11.21) 22.4 (3.9) 20.13 (5.7) 0.8 (5.8) 20.2 (8) 21.33 (7.7)
7.5 (23.02) 44.37 (13.26) 51.55 (14.38) 23.6 (4.4) 21.4 (4.7) 2.1 (10.7) 5 (10.2) 3.9 (10.5)
2 5 13
—
5 14 11
6 5 5
— — — — — —
M. Bertoux et al. / Alzheimer’s & Dementia 9 (2013) S84–S94
S89
Fig. 1. Plot box of old controls, young controls, AD and bvFTD groups at the four VMPFC tests. R-L, Reversal-Learning Test. *P , .05; **P , .001; ***P , .000001 (vs bvFTD); &P , .05; &&P , .001 (vs AD). OC, old controls; YC, young controls.
and AD (Z 5 22.50, P , .05) patients, which presented comparable numbers of reversal errors (Table 2 and Figure 1). Therefore, older controls and AD patients showed significantly more successful reverse rule changes than bvFTD patients (Z 5 24.38, P , .001 and Z 5 22.73, P , .01, respectively), with no differences between them. 3.5. Iowa Gambling Task 3.5.1. bvFTD patients and older control performances As shown in Table 2 and Figure 1, older controls obtained a mean net score (total of C1D deck choices minus A1B choices) of 23.2, whereas AD and bvFTD patients similarly attained a mean net score of 4.9. Figure 2 shows the mean performance across the block for each group. ANOVA (3 [groups] ! 5 [test blocks]) showed no effect of group (D 5 2.39, P . .1), no effect of test block (D 5 0.97, P . .1), and no group ! block interaction (D 5 1.12, P . .1). Older controls, AD patients, and bvFTD patients showed no significant differences on the IGT total net score or the separate test block net scores, with the exception of the trials 1–20 block (Z 5 22.04, P , .05), where ADs had a better net score, and the trials 81–100 block net
score (Z 5 22.36, P , .01), where bvFTDs outperformed the older controls. The debriefing of the participants revealed that only 2 of our 20 bvFTD patients were fully consciously aware of how to maximize profit on the IGT. A further 5 bvFTD patients showed partial awareness, whereas the remaining 13 were unaware of the rule for successful performance of the task. Interestingly, only 5 of the older controls had full conscious knowledge of the IGT rule (Figure 2), whereas 14 were partially aware and 11 unaware. A 3 ! 1 ANOVA (“rule knowledge” ! “IGT total score”) showed an effect of the level of conscious knowledge by IGT total net score for older controls (D 5 7.65, P , .01), but bvFTD patients showed no significant difference for these factors. For old controls, there was a marginal significant correlation between level of knowledge and IGT total score (r 5 0.35, P 5 .05), which, however, became nonsignificant after correction for multiple comparison (P . .1). This correlation was not significant for bvFTD patients (r 5 0.23, P . .1). Unfortunately, debriefing scores for AD patients were not available. To investigate whether the variable performance of the old control participants was due to an aging effect, we tested additional younger controls to determine their performance and to compare it with that of bvFTD and old controls.
S90
M. Bertoux et al. / Alzheimer’s & Dementia 9 (2013) S84–S94
0–21 block (Z 5 22.04, P , .05), AD patients had a better net score than young controls. With regard to debriefing, in the younger controls, 6 participants were fully aware of the IGT rule at the end of the test, whereas 5 had partial and another 5 no conscious awareness (Table 2 and Figure 2). A 3 ! 1 ANOVA (“rule knowledge” ! “IGT total score”) showed an effect of the level of conscious knowledge by IGT total net score in this group (D 5 7.04, P , .01). The correlation between level of knowledge and IGT total score was significant for young controls (r 5 0.68, P , .005). We found no correlation between age and education level and IGT total score within each of the groups. 3.6. Specificity/sensitivity of VMPFC tests
Fig. 2. Plot of the IGT net score, (C1D) and (A1B), from the 1st to the 100th trial in block of 20 cards for bvFTD, old, and young controls groups across blocks (Figure 1A) and according to the level of conscious knowledge for old controls (Figure 1B) and young control (Figure 1C).
3.5.2. Performance of bvFTD patients, AD patients, old controls, and young controls Younger controls achieved an average net score of 7.33 (Table 2 and Figure 1). ANOVA (4 groups ! 5 test blocks) showed no effect of test block (D 5 2.65, P 5 .08), no group effect (D 5 1.41, P . .1), and no block ! group interaction (D 5 1.39, P . .1). Younger controls showed no significant differences on the IGT total and each block net score compared with bvFTD patients (Z 5 20.48, P . .1), AD patients (Z 5 21.00, P . .1), and older controls (Z 5 21.1, P . .1). In the trials
In a final analysis were explored the specificity/sensitivity of the VMPFC tests to differentiate bvFTD from AD patients and old controls via receiver-operating characteristic (ROC) curve analysis (Figure 3). The area under the curve (AUC) was used to measure the overall performance of each ROC curve (with 95% confidence interval), which revealed AUCs of 0.74 for the Go/No-Go Subtest (FAB) score, 0.93 for the Mini-SEA composite score, and 0.67 for the IGT total score. For the Reversal-Learning Test, the AUC was 0.78 for the number of rules reversed and 0.54 for the number of errors committed before to reverse the first rule. A logistic regression using the “enter” method and only employing bvFTD and AD patients, revealed that 62.5% of the participants could be correctly classified into the bvFTD or AD group on the Go/No-Go score. The distinction between bvFTD and AD was accurate in 82.5% of participants using the Mini-SEA, and in 69.7% using the number of reversals performed in the Reversal-Learning Test. For the IGT, logistic regression revealed that 57.1% of participants could be correctly classified as bvFTD or AD patients. An overall logistic regression employing all the aforementioned test scores showed that 85.7% of patients could be classified correctly. Importantly, this classification was driven mainly by the Mini-SEA and the number of rules reversed in the Reversal-Learning Test, which accounted for 84.8% of the variance in the overall analysis. In the analysis focused on old controls and bvFTD patients, logistic regression showed that the FAB Go/No-Go score classified bvFTD or healthy controls with 83.7% accuracy. This classification showed accuracy at 88% with the Mini-SEA, 86% with the Reversal-Learning Test, and 64% with the IGT. Results for each VMPFC test for the old control, AD, and bvFTD groups are summarized in Figure 4. 4. Discussion To our knowledge, this is the first study to compare the sensitivity and specificity across a range of tests addressing VMPFC dysfunction in bvFTD. The results show that all
M. Bertoux et al. / Alzheimer’s & Dementia 9 (2013) S84–S94
Fig. 3. Receiver-operating characteristic curves for the FAB Go/No-Go, the Mini-SEA, the number of reversal performed at the Reversal-Learning Test (R-L), and the Iowa Gambling Task for old controls compared with those for bvFTD patients.
measures employed were sensitive for detection of VMPFC dysfunction; however, specificity varied widely, with Reversal-Learning and Mini-SEA Tests being the most specific for detection of VMPFC-related impairment in a sample of bvFTD patients versus AD patients and healthy controls. The Go/No-Go Subtest of the FAB showed limited specificity with good discrimination of bvFTD patients from agematched controls, but similar performance in AD patients. The IGT emerged as the least specific VMPFC test, with poor discrimination of bvFTD patients and AD patients, as well as healthy older and even younger controls. More specifically, our results show that motor inhibition as measured by the FAB Go/No-Go Subtest is a sensitive measure to detect VMPFC dysfunction in bvFTD, but it does not fully discriminate bvFTD from AD. Recent findings have shown that inhibitory function in bvFTD is severely impaired [22,30] and this has been associated with atrophy in the VMPFC [23]. Nevertheless, deficits in
Fig. 4. Summary of the performance for all the VMPFC tests for bvFTD, AD, and old control groups. A red cross indicates that the test or measure showed impairment; a green mark indicates that the test or measure was well succeeded.
S91
controlled-response inhibition have also been documented in AD [35]; this could explain why a Go/No-Go measure might be not as specific at distinguishing patient groups [28]. By contrast, the Mini-SEA [27], which evaluates theory of mind and emotion processing, showed excellent discrimination between bvFTD and AD patients, as well as controls. In particular, the Faux-Pas Subtest has been shown to be very sensitive in detecting VMPFC damage in bvFTD as it taps more explicitly into theory of mind processes, which are usually impaired in bvFTD, even at the earliest disease stages, but not in AD [18,26]. The neural correlates of the Face Emotion Subtest of the Mini-SEA might be attributed to the amygdala and VMPFC regions [11,36]. These two areas have been known for a long time to be involved in emotion processing [37] and both are atrophic in bvFTD [2,4] and AD patients [38], which may explain why the Face Emotion Subtest was not as specific to bvFTD patients as the Faux-Pas Subtest. The ReversalLearning task also showed very good detection of VMPFC dysfunction in bvFTD patients, replicating previous VMPFC lesion findings [9,11]. It is striking that such a simple test— which only requires switching decisions between two rewarded and nonrewarded stimuli—would give rise to such a discrepancy between bvFTD and healthy controls. Importantly, patients could learn to associate each stimulus with its specific reward; however, when the contingencies were changed (i.e., the previously rewarded stimulus was now not rewarded and vice-versa), patients failed to adapt their decision accordingly. Not surprisingly, therefore, bvFTD patients committed significantly more reversal errors as well as fewer rule changes than healthy elderly controls. AD patients’ performance dissociated on the reversallearning task, where they committed a similar number of reversal errors as bvFTD patients, but they had as many rule changes as controls. It is not clear why the reversal error scores dissociated for AD patients, yet one could speculate that the AD patients had more subtle inhibitory deficits, leading to higher reversal errors over time; nevertheless, these errors were not as frequent so as to hinder them from having more rule reversals. More surprising was the performance of all participants on the IGT. The IGT is one of the most commonly used clinical tests to detect VMPFC impairment in patients [10,31], and has been shown to be very sensitive to VMPFC damage in bvFTD [25]. Our results corroborate these findings, while also showing that the specificity of this test appears quite low as AD patients, healthy elderly, and younger controls showed highly variable performance levels on the test, resulting in nonsignificant differences between patients and controls. The question arises as to why the IGT test was not a specific measure of VMPFC damage in our study. Numerous studies have shown that the IGT taps successfully into VMPFC functions in patients [10,31], and these findings have been supported by functional neuroimaging experiments [39,40]. Thus, it could be predicted that only the bvFTD patients in our sample with substantial VMPFC damage should have
S92
M. Bertoux et al. / Alzheimer’s & Dementia 9 (2013) S84–S94
been impaired on this test, whereas AD patients and healthy controls, with no or minor VMPFC dysfunction, should have significantly outperformed the bvFTD patients. One potential reason for the variable performance, particularly in controls, might be that some controls did not understand the goal of the task and were therefore performing more variably. This is corroborated by our debriefing scores, which showed that a large percentage of elderly and younger controls did not understand how to maximize their winnings in the IGT at the end of the task. Importantly, only the controls who understood the task rules performed well on the task, reinforcing the notion that explicit knowledge is fundamental to successful IGT performance (see Maia and McClelland [41]). Still, this notion is controversial as Bechara and colleagues originally asserted that only implicit task rule knowledge is required to perform successfully on the IGT. Nevertheless, there is mounting evidence, substantiated by the current results, to indicate that, although implicit knowledge may be involved, it is certainly not sufficient to perform well in this task. Studies with amnesic patients [42,43] have shown that patients with no explicit memory, but intact implicit memory, are largely impaired on the IGT. Fernie and Turney [44] showed that differences in instructions (explicit vs nonexplicit task rule explanation) can lead to substantial changes in IGT task performance. This clearly questions whether implicit knowledge is sufficient to perform well on this task and raises the question as to whether bvFTD, AD, and control subjects would have differed significantly from one another if they had been told explicitly the goal of the task beforehand. Another potential reason for the variable control performance could be that our elderly controls did have some subtle VMPFC dysfunction, which can occur with aging [45]. Indeed, several studies have shown that elderly participants may perform poorly on the task [46,47], which has been sometimes attributed to subtle VMPFC damage in those controls [48]. However, we believe this was unlikely in our sample, as the younger controls showed similar performance patterns to the older controls, and the elderly controls performed significantly better than the patients on the remaining VMPFC tasks. Finally, there is increasing evidence that the task complexity and implicit nature of the IGT can lure participants into more elaborate strategies than necessary to perform well on the test. This could be particularly true for highly educated participants, who may show a paradoxical effect on the IGT and, despite their higher education, perform poorly on the task [49]. This could be explained by the fact that such controls are influenced mainly by the frequency of positive outcomes rather than by the maximization of winnings in the longer term. Indeed, Caroselli et al [50] showed that young controls select the disadvantageous decks (A and B) more frequently. A similar result was obtained by Lin et al [51], who showed that control participants in particular chose the “B” deck of the IGT, which provides a higher frequency of rewards, whereas, at the same
time, offers few losses of larger amounts. The findings of these studies are consistent with animal reinforcement studies showing that animals are influenced more strongly by the frequency of reward than by amount of reward [52]. On a clinical level our findings have major implications. To date, there is no consensus on which tests are best to use to detect VMPFC damage, and establishment of behavioral VMPFC-specific symptoms is mostly done via clinical observation and carer questionnaires (e.g., the Neuropsychiatric Inventory [NPI] or Cambridge Behavioral Inventory [CBI]), which are subjective and highly dependent on the experience of the clinician and the perceptiveness of the carer. The current data, based on objective tests of VMPFC function, clearly indicate that some tests are better than others for determination of VMPFC dysfunction. In particular, the Mini-SEA showed excellent discrimination with 82.5% of patients correctly classified as bvFTD or AD. The Mini-SEA is therefore a major step into developing novel tests, which tap into the VMPFC-dependent social cognition processes, as many of the existing tasks that address these processes are experimental and not currently appropriate for clinical practice. bvFTD patients have known social cognition problems and simple clinical tests, such as the Mini-SEA, are therefore urgently needed to corroborate carer reports. The simple FAB Go/No-Go subscore showed very good discrimination of bvFTD and AD and can be easily employed for diagnostic purposes in a routine clinic setting and does not require complex setup. The ReversalLearning Test has so far been used infrequently in clinical settings, even though it distinguished bvFTD and AD in .67.5% of cases at clinical presentation. The employed version, based on work by Rolls et al [9], can be implemented easily in a clinical setting that routinely uses other computerized tests, such as the IGT. It can further corroborate the FAB findings in patients and generates results that are not influenced by the experience of the assessor. Moreover, both the FAB Go/No-Go and the Reversal-Learning tasks are particularly useful as they are less vulnerable to the confounding aspects of language degeneration in dementia syndromes. The IGT emerged as the least specific test employed to detect VMPFC damage, with the complexity of the task being one potential issue that drives the performance variability on this test. Based on our results, we recommend that clinicians complement the IGT with another VMPFC-specific test for corroboration of its results. Despite these promising findings, in our study we did not cross-correlate behavioral findings with the neuroimaging data, such as voxel-based morphometry or cortical thickness. Thus, we cannot know for sure whether the bvFTD patients’ VMPFC atrophy specifically contributed to the behavioral findings, even though previous studies have shown that all employed tests are VMPFC-specific. Taken together, our findings show that there are now several VMPFC-specific tests that can detect dysfunction in this region in bvFTD patients and can be easily employed
M. Bertoux et al. / Alzheimer’s & Dementia 9 (2013) S84–S94
in a clinical setting, which in turn will inform future disease-modifying therapies and patient interventions. Still, there was a discrepancy between tests that showed excellent (Mini-SEA), good (FAB Go/No-Go and Reversal-Learning), and poor (IGT) patient versus AD or control discrimination, which should be taken into account by clinicians. Finally, although future studies focusing on the characterization of VMPFC dysfunction in bvFTD are needed to improve comprehension of the early biologic, cognitive, and behavioral impairments of this disease, the results of the present study support the idea that VMPFC assessment could further increase the sensitivity and specificity of bvFTD diagnostic criteria [20] and should be taken into account in the next revision of the current criteria [17]. RESEARCH IN CONTEXT
1. Systematic review: The authors reviewed the literature on VMPFC tests in bvFTD using PubMed sources. VMPFC dysfunction is considered as an early feature of bvFTD and recent publications raised the importance of its assessment for diagnosis purpose. While Reversal-learning and Go/No-go tests have been designated as the classical evaluation of VMPFC dysfunctions, some groups have presented the IGT as the gold standard to evaluate this region in bvFTD. However, social-cognition tool like the mini-SEA has been recently published and described as promising in this purpose. 2. Interpretation: Here, we compared the sensitivity/ specificity of these tests in bvFTD, AD and controls. Sensitivity was high across all employed tests, however specificity varied considerably, supporting the usefulness of the mini-SEA and Reversal-Learning test but refuting the value of IGT in diagnosis purpose. 3. Future directions: Future cognitive studies should focus on the development of new VMPFC tests to evaluate and diagnose bvFTD independently of other cognitive functions such as language.
References [1] Piguet O, Hornberger M, Mioshi E, Hodges JR. Behavioural-variant frontotemporal dementia: diagnosis, clinical staging, and management. Lancet Neurol 2011;10:162–72. [2] Schroeter ML, Raczka K, Neumann J, von Cramon DY. Neural networks in frontotemporal dementia—a meta-analysis. Neurobiol Aging 2008;29:418–26. [3] Agosta F, Canu E, Sarro L, Comi G, Filippi M. Neuroimaging findings in frontotemporal lobar degeneration spectrum of disorders. Cortex 2012;48:389–413.
S93
[4] Seeley WW. Selective functional, regional, and neuronal vulnerability in frontotemporal dementia. Curr Opin Neurol 2008;21:701–7. [5] Nakano S, Asada T, Yamashita F, Kitamura N, Matsuda H, Hirai S, et al. Relationship between antisocial behavior and regional cerebral blood flow in frontotemporal dementia. Neuroimage 2006;32:301–6. [6] Eslinger PJ, Moore P, Anderson C, Grossman M. Social cognition, executive functioning, and neuroimaging correlates of empathic deficits in frontotemporal dementia. J Neuropsychiatry Clin Neurosci 2011; 23:74–82. [7] Rolls ET. The neurophysiology and functions of the orbitofrontal cortex. In: Zald DH, Rauch S, eds. The orbitofrontal cortex. New York: Oxford University Press; 2006. p. 95–125. [8] Iversen SD, Mishkin M. Perseverative interference in monkeys following selective lesions of the inferior prefrontal convexity. Exp Brain Res 1970;11:376–86. [9] Rolls ET, Hornak J, Wade D, McGrath J. Emotion-related learning in patients with social and emotional changes associated with frontal lobe damage. J Neurol Neurosurg Psychiatry 1994;57:1518–24. [10] Bechara A, Damasio H, Damasio AR. Emotion, decision making and the orbitofrontal cortex. Cereb Cortex 2000;10:295–307. [11] Berlin HA, Rolls ET, Kischka U. Impulsivity, time perception, emotion and reinforcement sensitivity in patients with orbitofrontal cortex lesions. Brain 2004;127:1108–26. [12] Shamay-Tsoory SG, Aharon-Peretz J, Perry D. Two systems for empathy: A double dissociation between emotional and cognitive empathy in inferior frontal gyrus versus ventromedial prefrontal lesions. Brain 2009;132:617–27. [13] O’Doherty J, Kringelbach ML, Rolls ET, Hornak J, Andrews C. Abstract reward and punishment representations in the human orbitofrontal cortex. Nat Neurosci 2001;4:95–102. [14] Carrington SJ, Bailey AJ. Are there theory of mind regions in the brain? A review of the neuroimaging literature. Hum Brain Mapp 2009;30:2313–35. [15] Horn NR, Dolan M, Elliott R, Deakin JF, Woodruff PW. Response inhibition and impulsivity: An fMRI study. Neuropsychologia 2003; 41:1959–66. [16] Moll J, de Oliveira-Souza R, Bramati IE, Grafman J. Functional networks in emotional moral and nonmoral social judgments. Neuroimage 2002;16:696–703. [17] Rascovsky K, Hodges JR, Knopman D, Mendez MF, Kramer JH, Neuhaus J, et al. Sensitivity of revised diagnostic criteria for the behavioural variant of frontotemporal dementia. Brain 2011;134:2456–77. [18] Gregory C, Lough S, Stone V, Erzinclioglu S, Martin L, BaronCohen S, et al. Theory of mind in patients with frontal variant frontotemporal dementia and Alzheimer’s disease: Theoretical and practical implications. Brain 2002;125:752–64. [19] Adenzato M, Cavallo M, Enrici I. Theory of mind ability in the behavioural variant of frontotemporal dementia: an analysis of the neural, cognitive, and social levels. Neuropsychologia 2010;48:2–12. [20] Schroeter ML. Considering the frontomedian cortex in revised criteria for behavioural variant frontotemporal dementia. Brain 2012; 135:e213. [21] Burgess PWS, Shallice T. The Hayling and Brixton Tests. Bury St. Edmunds, UK: Thames Valley Test Company; 1997. [22] Hornberger M, Piguet O, Kipps C, Hodges JR. Executive function in progressive and nonprogressive behavioral variant frontotemporal dementia. Neurology 2008;71:1481–8. [23] Hornberger M, Geng J, Hodges JR. Convergent grey and white matter evidence of orbitofrontal cortex changes related to disinhibition in behavioural variant frontotemporal dementia. Brain 2011;134:2502–12. [24] Stone VE, Baron-Cohen S, Knight RT. Frontal lobe contributions to theory of mind. J Cogn Neurosci 1998;10:640–56. [25] Torralva T, Kipps CM, Hodges JR, Clark L, Bekinschtein T, Roca M, et al. The relationship between affective decision-making and theory of mind in the frontal variant of fronto-temporal dementia. Neuropsychologia 2007;45:342–9.
S94
M. Bertoux et al. / Alzheimer’s & Dementia 9 (2013) S84–S94
[26] Funkiewiez A, Bertoux M, de Souza LC, Levy R, Dubois B. The SEA (Social Cognition and Emotional Assessment): A clinical neuropsychological tool for early diagnosis of frontal variant of frontotemporal lobar degeneration. Neuropsychology 2012;26:81–90. [27] Bertoux M, Delavest M, de Souza LC, Funkiewiez A, Lepine JP, Fossati P, et al. Social cognition and emotional assessment differentiates frontotemporal dementia from depression. J Neurol Neurosurg Psychiatry 2012;83:411–6. [28] Slachevsky A, Villalpando JM, Sarazin M, Hahn-Barma V, Pillon B, Dubois B. Frontal assessment battery and differential diagnosis of frontotemporal dementia and Alzheimer disease. Arch Neurol 2004; 61:1104–7. [29] Collette F, Amieva H, Adam S, Hogge M, van der Linden M, Fabrigoule C, et al. Comparison of inhibitory functioning in mild Alzheimer’s disease and frontotemporal dementia. Cortex 2007; 43:866–74. [30] Stopford CL, Thompson JC, Neary D, Richardson AM, Snowden JS. Working memory, attention, and executive function in Alzheimer’s disease and frontotemporal dementia. Cortex 2012;48:429–46. [31] Bechara A, Damasio H, Tranel D, Damasio AR. Deciding advantageously before knowing the advantageous strategy. Science 1997; 275:1293–5. [32] Dubois B, Slachevsky A, Litvan I, Pillon B. The FAB: A Frontal Assessment Battery at bedside. Neurology 2000;55:1621–6. [33] Neary D, Snowden JS, Gustafson L, Passant U, Stuss D, Black S, et al. Frontotemporal lobar degeneration: A consensus on clinical diagnostic criteria. Neurology 1998;51:1546–54. [34] Dubois B, Feldman HH, Jacova C, Dekosky ST, Barberger-Gateau P, Cummings J, et al. Research criteria for the diagnosis of Alzheimer’s disease: Revising the NINCDS-ADRDA criteria. Lancet Neurol 2007; 6:734–46. [35] Amieva H, Phillips LH, Della Sala S, Henry JD. Inhibitory functioning in Alzheimer’s disease. Brain 2004;127:949–64. [36] Fusar-Poli P, Placentino A, Carletti F, Landi P, Allen P, Surguladze S, et al. Functional atlas of emotional faces processing: A voxel-based meta-analysis of 105 functional magnetic resonance imaging studies. J Psychiatry Neurosci 2009;34:418–32. [37] Adolphs R. Neural systems for recognizing emotion. Curr Opin Neurobiol 2002;12:169–77. [38] Cavedo E, Boccardi M, Ganzola R, Canu E, Beltramello A, Caltagirone C, et al. Local amygdala structural differences with 3T MRI in patients with Alzheimer disease. Neurology 2011;76:727–33.
[39] Li X, Lu ZL, D’Argembeau A, Ng M, Bechara A. The Iowa Gambling Task in fMRI images. Hum Brain Mapp 2010;31:410–23. [40] Fukui H, Murai T, Fukuyama H, Hayashi T, Hanakawa T. Functional activity related to risk anticipation during performance of the Iowa Gambling Task. Neuroimage 2005;24:253–9. [41] Maia TV, McClelland JL. A reexamination of the evidence for the somatic marker hypothesis: What participants really know in the Iowa gambling task. Proc Natl Acad Sci USA 2004; 101:16075–80. [42] Gutbrod K, Krouzel C, Hofer H, Muri R, Perrig W, Ptak R. Decisionmaking in amnesia: do advantageous decisions require conscious knowledge of previous behavioural choices? Neuropsychologia 2006;44:1315–24. [43] Gupta R, Duff MC, Denburg NL, Cohen NJ, Bechara A, Tranel D. Declarative memory is critical for sustained advantageous complex decision-making. Neuropsychologia 2009;47:1686–93. [44] Fernie G, Tunney RJ. Some decks are better than others: The effect of reinforcer type and task instructions on learning in the Iowa Gambling Task. Brain Cogn 2006;60:94–102. [45] Kalpouzos G, Chetelat G, Baron JC, Landeau B, Mevel K, Godeau C, et al. Voxel-based mapping of brain gray matter volume and glucose metabolism profiles in normal aging. Neurobiol Aging 2009; 30:112–24. [46] MacPherson SE, Phillips LH, Della Sala S. Age, executive function, and social decision making: A dorsolateral prefrontal theory of cognitive aging. Psychol Aging 2002;17:598–609. [47] Isella V, Mapelli C, Morielli N, Pelati O, Franceschi M, Appollonio IM. Age-related quantitative and qualitative changes in decision making ability. Behav Neurol 2008;19:59–63. [48] Denburg NL, Tranel D, Bechara A. The ability to decide advantageously declines prematurely in some normal older persons. Neuropsychologia 2005;43:1099–106. [49] Evans CE, Kemish K, Turnbull OH. Paradoxical effects of education on the Iowa Gambling Task. Brain Cogn 2004;54:240–4. [50] Caroselli JS, Hiscock M, Scheibel RS, Ingram F. The simulated gambling paradigm applied to young adults: An examination of university students’ performance. Appl Neuropsychol 2006; 13:203–12. [51] Lin CH, Chiu YC, Lee PL, Hsieh JC. Is deck B a disadvantageous deck in the Iowa Gambling Task? Behav Brain Funct 2007;3:16. [52] Steinhauer GD. Preference for predictable small rewards over unpredictable larger rewards. Psychol Rep 1984;54:467–71.