Abstract
Loss of empathy is a core presenting feature of the behavioral-variant of frontotemporal dementia (bvFTD), resulting in socioemotional difficulties and behavioral transgressions. In contrast, interpersonal functioning remains relatively intact in Alzheimer’s disease (AD), despite marked cognitive decline. The neural substrates mediating these patterns of loss and sparing in social functioning remain unclear, yet are relevant for our understanding of the social brain. We investigated cognitive versus affective aspects of empathy using the Interpersonal Reactivity Index (IRI) in 25 AD and 24 bvFTD patients and contrasted their performance with 22 age- and education-matched controls. Cognitive empathy was comparably compromised in AD and bvFTD, whereas affective empathy was impaired exclusively in bvFTD. While controlling for overall cognitive dysfunction ameliorated perspective-taking deficits in AD, empathy loss persisted across cognitive and affective domains in bvFTD. Voxel-based morphometry analyses revealed divergent neural substrates of empathy loss in each patient group. Perspective-taking deficits correlated with predominantly left-sided temporoparietal atrophy in AD, whereas widespread bilateral frontoinsular, temporal, parietal, and occipital atrophy was implicated in bvFTD. Reduced empathic concern in bvFTD was associated with atrophy in the left orbitofrontal, inferior frontal, and insular cortices, and the bilateral mid-cingulate gyrus. Our findings suggest that social cognitive deficits in AD arise largely as a consequence of global cognitive dysfunction, rather than a loss of empathy per se. In contrast, loss of empathy in bvFTD reflects the deterioration of a distributed network of frontoinsular and temporal structures that appear crucial for monitoring and processing social information.
Keywords
INTRODUCTION
Empathy represents a multifaceted and adaptive process that provides the foundation for successful social functioning and prosocial behavior. Despite its fundamental contributions to social cognitive function, there remains no firm consensus on how to define or accurately measure empathy [reviewed by 1]. Definitions of empathy have tended to emphasize cognitive and affective dimensions, wherein the individual must understand and appreciate the other person’s experience (cognitive) and share the emotional experience of the other person (affective) [2, 3]. These separable dimensions are typically proposed to map onto underlying mentalizing and experience sharing processing streams, respectively, which in turn engage largely dissociable neural systems [reviewed by 4]. While the capacity for empathy has been proposed to rely upon a number of transversal cognitive processes including affective arousal, emotion recognition, executive function, and emotion regulation [1], the modulating role of a set of domain-specific mechanisms to the experience of empathy remains a source of debate in the literature [reviewed by 4]. As such, the utility of current models of empathy has been questioned along with the tendency to deconstruct such complex cognitive phenomena into a discrete set of sub-processes [4]. The complexity of empathy is further reflected on the neuroanatomical level with functional neuroimaging studies implicating a distributed neural network including dorsolateral and medial prefrontal, frontoinsular, temporoparietal, and limbic structures during experimental tasks assessing empathy in healthy individuals [reviewed by 2, 5].
Impaired capacity for empathy represents one of the core clinical features of the behavioral-variant of frontotemporal dementia (bvFTD) with an early loss of empathy emphasized as a primary symptom in the recently revised diagnostic criteria [6]. BvFTD patients display emotional blunting, decreased social interest, loss of affection and warmth, and diminished responsiveness to the feelings of others [7, 8] in the context of marked emotion recognition difficulties [9, 10] and an inability to understand and infer the beliefs and perspectives of others [11, 12]. These striking deficits in socioemotional processing are typically interpreted as reflecting the selective vulnerability of frontoinsular and paralimbic structures in this syndrome [13, 14], and have a profound impact on the patient-caregiver relationship [8].
A large body of evidence points to pervasive deficits in the ability to mentalize, via theory of mind, to consider the thoughts, beliefs and perspectives of others in bvFTD [12]. These mentalizing impairments manifest across a range of experimental tasks assessing false belief [11], detection of social faux pas [15, 16], attribution of intentions in humorous cartoons [17–20] and social scenarios [21, 22], and understanding and adopting the viewpoint of others [23]. Cognitive empathy, defined here as the ability to understand and appreciate the emotions of others, is proposed to rely upon a core mentalizing component which enables the individual to draw inferences about the emotional experience of others [4]. The well-documented deficits in cognitive empathy in bvFTD potentially reflect the disruption of a core mentalizing mechanism, as well as broader processes including executive function and emotion recognition [24]. Importantly, these impairments occur irrespective of the emotional valence of the scenario, with reduced cognitive empathy manifesting across positive and negative stimuli on the Multifaceted Empathy Test in bvFTD [25]. Accordingly, a fundamental deficit in the ability to understanding the beliefs [23] and affective experiences [20] of others is present. These observations are further corroborated by caregiver reports of disrupted cognitive empathy in bvFTD relative to Controls [7, 26–28] and patients with Alzheimer’s disease (AD) [8, 29] using the Perspective Taking subscale of the Interpersonal Reactivity Index, as a surrogate for cognitive empathy.
A similar picture emerges with respect to affective empathy, defined here as the ability to share the emotions or the emotional experience of others [4]. These impairments are perhaps not surprising when considered in relation to the profound socioemotional disturbances characteristic of the bvFTD syndrome [6, 30]. Marked deficits have been observed in bvFTD regarding the sharing of emotional experiences of others on story-based empathy tasks [31]. These affective empathy deficits appear most pronounced where negative experiences are concerned with patients displaying significantly reduced levels of shared emotional experiences and significantly attenuated emotional reactions to negative stimuli on the Multifaceted Empathy Test [25]. Caregiver ratings again confirm reduced feelings of concern, warmth, and compassion for less fortunate others in comparison with healthy Controls [7, 26] and patients with AD [8, 29] using the Empathic Concern subscale of the Interpersonal Reactivity Index (IRI) [3] as a surrogate for affective empathy.
Neuroimaging studies point overwhelmingly to the degeneration of key frontoinsular, limbic, and lateral temporal regions in the social brain in driving these social cognitive deficits in bvFTD. Notably, grey matter atrophy in medial prefrontal regions has been shown to correlate with a reduction in empathic concern on experimental tasks indexing intentional harm [32], and ratings of empathic concern on the IRI [26]. In addition, limbic structures including the amygdala, and the temporoparietal junction have been found to correlate with empathy impairments on a non-verbal emotion attribution empathy task [33], while right-lateral temporal structures have been implicated across a range of social cognitive impairments in bvFTD including mental state attribution from cartoon scenarios [17], emotion attribution on a nonverbal task [33], and ratings of total empathy in a mixed FTLD sample on the IRI [7]. Interestingly, bvFTD patients have been shown to overestimate their capacity for empathic concern on the IRI, attributable to atrophy in predominantly right-sided structures including the right anterior inferior temporal gyrus, and right posterior insula [28]. In addition, a recent study revealed that alterations in resting-state activity in medial prefrontal and limbic networks underpin deficits in the attribution of emotions on a cartoon task in this syndrome [31]. As such, the evidence to date converges to suggest a specific vulnerability in a distributed network of regions in bvFTD that support the capacity for cognitive and affective aspects of empathy. It remains unclear, however, whether this susceptibility produces a global disruption in empathy, or impacts differentially on cognitive versus affective dimensions. In addition, it remains unclear how the neuroanatomical signature of empathy deficits potentially differs according to dementia subtype.
Delineating the precise neural substrates of cognitive versus affective empathy disruption is directly relevant for understanding disease-specific profiles of impairment in younger-onset dementia syndromes. With increasing evidence of prominent episodic memory impairments in bvFTD [34, 35], it has been suggested that social cognitive dysfunction represents a more effective means of differentiating between bvFTD and patients with AD [36]. While socioemotional processing is generally held to remain relatively intact in AD with advancing disease severity, a recent longitudinal study uncovered a significant decline across a range of socioemotional processes in AD including emotion recognition and sarcasm detection [37], albeit in the context of global cognitive dysfunction. Similarly, studies of social cognition in AD have revealed significant deficits in mental state attribution; however, these deficits appear largely modulated by general cognitive decline [17, 20]. As such, the evidence to date reveals a mixed picture where social cognitive deficits in AD are concerned, and the neural substrates of these processes remain unclear.
The objective of the present study was to contrast directly the capacity for cognitive versus affective empathy in bvFTD in comparison with disease-matched cases of AD and to elucidate the neural substrates of these deficits. We predicted that empathy would be globally disrupted across cognitive and affective domains in bvFTD, however, only cognitive empathy deficits would be evident in AD. Moreover, we hypothesized that cognitive empathy disruption in AD would stem largely from global cognitive dysfunction in this syndrome rather than a primary deficit in empathy per se. Using voxel-based morphometry of structural MRI scans, we sought to determine the neural correlates of these deficits as an important step to identify the mechanisms which drive social cognitive impairment in neurodegenerative disorders and to elucidate the key structures which must be functional to support complex social cognitive processes.
METHODS AND MATERIALS
Participants
Empathy was investigated in a total of 71 participants who attended FRONTIER, the frontotemporal dementia clinic at Neuroscience Research Australia (NeuRA) in Sydney. Twenty-five individuals with a clinical diagnosis of AD, 24 individuals with a diagnosis of clinically probable bvFTD, and 22 healthy older Control participants were identified as suitable for inclusion in the study.
Patients were seen by a neurologist, neuro-psychologist, and occupational therapist and underwent comprehensive clinical investigation, cognitive assessment, carer interviews, and structural neuroimaging to establish diagnosis, which was made in line with current clinical diagnostic criteria for AD [38] or bvFTD [6]. AD patients presented with significant episodic memory loss in the context of relatively preserved socioemotional functioning and personality. BvFTD patients were principally characterized by an insidious decline in socioemotional competency (e.g., disinhibition, apathy), executive dysfunction, and functional decline. Amnestic presentations of bvFTD were also included, in recognition of mounting evidence pointing to marked episodic memory dysfunction in this syndrome [39, 40].
Healthy Control participants were recruited through the NeuRA research volunteer panel and local community groups. All controls scored 0 on the Clinical Dementia Rating scale (CDR) [41] and 88 or above on the Addenbrooke’s Cognitive Examination-Revised (ACE-R) [42]. Exclusion criteria for all participants included prior history of mental illness, significant head injury, movement disorders, cerebrovascular disease, alcohol and other drug abuse, and limited English proficiency. Control participants were included in this study for group comparisons with the patient cohort across cognitive and behavioral measures, and were selected to closely match patients with regard to age and years of education.
Ethical approval for this study was obtained from the South Eastern Sydney Local Area Health and University of New South Wales ethics committees. All participants provided informed consent, and in some cases dual consent was obtained from the carer, in accordance with the Declaration of Helsinki.
General cognitive assessment
Participants completed a standard battery of neuropsychological tests examining a range of cognitive domains. The battery included the ACE-R [42] to measure overall level of cognitive functioning, measures of attention and working memory (Digit Span) [43], language (SydBAT Naming and Comprehension subtests) [44], episodic memory (Rey Complex Figure and Rey Auditory Verbal Learning Task) [45, 46] verbal fluency (Letter fluency) [47], response inhibition (Hayling test) [48], and psychomotor speed and mental flexibility (Trail Making Test Parts A and B) [49]. Facial emotion recognition was measured using the Ekman 60 [50].
Carer rated behavioral measures
The Cambridge Behavioural Inventory-Revised (CBI-R) [51] was completed by each patient’s carer as an index of behavioral change. For the purposes of this study, we focused on the subscales assessing memory change and presence of abnormal behaviors. Disease severity was established using the Frontotemporal Dementia Rating Scale (FRS) [52] which is a dementia staging tool used to measure change in functional abilities (e.g., ability to use a telephone) and presence of behavioral symptomatology (e.g., impulsivity).
Interpersonal reactivity index
The IRI is a 28-item questionnaire used to measure levels of empathy [3]. It comprises four 7-item subscales, each of which represents a different aspect of empathy: Perspective Taking (PT), the capacity to imagine the perspective of another; Fantasy, the capacity to project oneself into the experiences of fictional characters; Empathic Concern (EC), the other-centered emotional response resulting from the perception of another’s emotional state; and Personal Distress, an indicator of one’s general anxiety and self-oriented emotional reactivity. Only the PT and EC subscales of the IRI were examined in this study as these subscales have been shown to be most relevant to patient care [53] and have previously been used to index cognitive and emotional aspects of empathy, respectively [26, 54].
Each item of the IRI comprises a statement to which a rating is given in accordance with how well it reflects the participant on a 5 point Likert scale, ranging from 0 (does not describe me/the patient well) to 4 (describes me/the patient very well). The maximum score for each subscale is 28. These scores were then transformed to a 1 to 5-point scale (i.e., where a rating of 0 is given a transformed score of 1, etc.). To ensure the reliability of scores and to avoid potential biases that occur when several items are missing from a particular subscale, at least 4 of the 7 questions for each subscale were required to have been completed for inclusion in the study [8]. To account for missing items when present, the sum of the transformed scores for each subscale was divided by the total possible subscale score (i.e., 35 where no items are missing) and multiplied by 100. These percentage scores were used as the measure of each participant’s level of empathy in the subsequent analyses.
Carer ratings of a patient’s level of empathy were obtained for two time points: (i) before the illness, and (ii) at the present time. The use of caregiver ratings has been demonstrated to be an effective and reliable method for assessing levels of empathy in dementia [7]. Self-ratings of empathy in Control participants were obtained for the present time only.
Statistical analyses
Cognitive data were analyzed using IBM SPSS Statistics (Version 22). Demographic and disease variables were analyzed using univariate ANOVAs. Chi-squared tests (χ2), based on the frequency patterns of categorical variables, were used where appropriate (e.g., sex). Neuropsychological test performance across the groups was analyzed using multivariate ANOVA, followed by Sidak corrected post hoc tests.
For the IRI ratings, a mixed model analysis of covariance (ANCOVA) was conducted with empathy score (PT and EC ratings) as the within subjects variable, diagnosis (AD, bvFTD, Control) as the between subject variable, and age included as a covariate. Post hoc tests with Sidak correction for multiple comparisons were conducted to examine the main and interaction effects. Paired sample t-tests were conducted to compare pre-illness and present PT and EC ratings within each of the patient groups. Finally, one-tailed Pearson correlations were conducted to examine the relationship between empathy ratings and neuropsychological test performance.
Image acquisition
Participants underwent whole-brain T1 weighted imaging using a 3T Philips MRI scanner with standard quadrature head coil (eight channels). Structural T1-weighted images were acquired using the following sequences: coronal orientation, matrix 256 × 256, 200 slices, 1 × 1 mm in-plane resolution, slice thickness 1 mm, echo time/repetition time = 2.6/5.8 ms, flip angle α= 19°. Structural MRI data were analyzed with FSL-VBM, a voxel-based morphometry (VBM) analysis [55, 56] using the FSL-VBM toolbox from the FMRIB software package (http://www.fmrib.ox.ac.uk/fsl/fslvbm) [57]. All scans were examined by a neuroradiologist for structural abnormalities; none were reported for control participants. Prior to analyses, all participant scans were visually inspected for significant head movements and artifacts.
Voxel-based morphometry analysis
Voxel-based morphometry (VBM) using structural MRI data was used to identify grey matter volume changes across groups on a voxel-by-voxel basis. Briefly, structural images were extracted using the FSL brain extraction tool (BET) [57]. Tissue segmentation was then carried out on the brain extracted images using FMRIB’s Automatic Segmentation Tool (FAST) [58]. The resulting grey matter partial volumes were aligned to the Montreal Neurological Institute standard space (MNI152) using the FMRIB non-linear registration approach (FNIRT) [59, 60] using a b-spline representation of the registration warp field. A study-specific template was created from the resulting images, combining AD, bvFTD and Controls images, to which the native grey matter images were re-registered non-linearly. The registered partial volume maps were then modulated by dividing by the Jacobian of the warp field, to correct for local expansion or contraction. Finally, the modulated segmented images were smoothed with an isotropic Gaussian kernel with a sigma of 3 mm.
A whole-brain voxel-wise general linear model was applied to investigate grey matter intensity differences via permutation-based non-parametric testing [61] with 5000 permutations per contrast. In the first step, differences in cortical grey matter intensities between patients (AD and bvFTD) and Controls were assessed. Clusters from the group atrophy analyses were extracted using the threshold-free cluster enhancement method (tfce) and corrected for Family-Wise Error (FWE) at p < 0.05.
Next, correlations between current ratings on the IRI and regions of grey matter intensity were investigated using a whole-brain approach. In a first step, we investigated associations between grey matter intensity and empathy ratings across the Perspective Taking and Empathic Concern subscales of the IRI for all participants combined (n = 71, See Supplementary Material). Then, we ran a set of subgroup analyses to determine the neural correlates of each IRI subscale specific to each patient group. Each patient group was combined with Controls to increase the statistical power to detect brain-behavior relationships across the entire brain by achieving greater variance in behavioral scores [62, 63]. For statistical power, a covariate only statistical model with a positive t-contrast was used, providing an index of association between grey matter intensity and IRI ratings. Age was included as a nuisance variable in the atrophy and covariate analyses. Anatomical locations of significant results were overlaid on the MNI standard brain, with maximum coordinates provided in MNI stereotaxic space. Anatomical labels were determined with reference to the Harvard-Oxford probabilistic cortical atlas. Clusters were extracted using a voxel-wise approach uncorrected at p < 0.001, with a cluster extent threshold of 100 contiguous voxels.
Finally, an overlap analysis was conducted using the results from the covariate analyses to identify common grey matter regions implicated in empathy disruption in bvFTD. The scaled contrasts of the statistical maps generated from the bvFTD and Control covariate analyses were multiplied to create an inclusive, or overlap, mask.
RESULTS
Demographics
Participants did not differ significantly in terms of age (p = 0.092), years of education (p = 0.193), and sex distribution (p = 0.907). The AD and bvFTD patient groups were matched for all demographic variables including age (t = 1.284, p = 0.205) and disease duration (months elapsed since onset of symptoms, p = 0.952) (see Table 1). Caregiver ratings of memory problems were reported more frequently in AD than bvFTD patients (F(1, 46) = 6.238, p = 0.016), whereas significantly more abnormal behaviors were reported by caregivers of bvFTD than AD patients (F(1, 46) = 6.552, p = 0.014). BvFTD patients demonstrated greater functional impairment than AD patients, as measured by the FRS (F(1, 44) = 5.210, p = 0.027).
Global cognitive function
On a general cognitive screening measure (ACE-R), a significant effect of diagnosis was observed (F(2, 25) = 26.995, p < 0.001), with patient groups (AD and bvFTD) performing significantly poorer than Control participants (all p values <0.0001). Overall cognitive impairment did not differ between the patient groups (ACE-R range: AD = 56–89; bvFTD = 50–95; p = 0.994).
Neuropsychological testing revealed deficits characteristic of each patient group (Table 1). In brief, compared with Controls, both patient groups demonstrated impairments across measures of attention (Digit Span forwards: p values <0.01), working memory (Digit Span backwards: p values <0.01), verbal fluency (Letter fluency: p values <0.05), episodic memory (RAVLT delayed score: p values <0.001; Rey Complex Figure percentage retained: p values = 0.001), and processing speed (Trails A: p values <0.005). Relative to Controls, both patient groups also demonstrated reduced mental flexibility (Trails B-A, p < 0.001). Additional impairments were observed for the bvFTD group, compared to Controls, in inhibition (Hayling, p = 0.025), language (Naming and Comprehension, p values <0.05), and emotion recognition (Ekman, p = 0.028).
Direct comparisons of the patient groups revealed no significant differences across any of the neuropsychological tests (all p values >0.065).
IRI Empathy behavioral results
Pre-morbid ratings of empathy
To ensure that patient groups were matched in terms of premorbid socioemotional processing, we conducted a multivariate ANOVA with Sidak post hoc tests comparing “before illness” PT and EC ratings between AD and bvFTD patients. Patient groups showed comparable ratings for both PT (F(1, 47) = 0.338, p = 0.564) and EC (F(1, 47) = 0.795, p = 0.377) pre-symptom onset.
Present ratings of empathy
As illustrated in Fig. 1, a significant main effect of diagnosis was observed (F(2, 67) = 29.108, p < 0.001), with Sidak post hoc analyses revealing that both AD (p = 0.042) and bvFTD (p < 0.0001) patient groups were rated significantly worse than Controls irrespective of empathy subscale. No main effect of empathy subscale was found (F(1, 67) = 1.645, p = 0.204).
A significant empathy subtype by diagnosis interaction effect was evident (F(2, 67) = 3.795, p = 0.027). This interaction reflected the fact that both patient groups had significantly poorer PT ratings relative to Controls (p values <0.0001) yet a dissociation was evident on the EC subscale. Whereas bvFTD patients were rated significantly lower than Controls in terms of EC (p < 0.0001), AD patients were rated in line with Controls (p = 0.940). Comparing the patient groups on the empathy measures, bvFTD patients were rated significantly poorer than AD patients on both the PT and EC subscales (p values <0.0001).
Within group comparisons revealed significant differences between pre-illness and present empathy ratings in the patient groups. On the PT subscale, both patient groups showed significantly poorer ratings at the present time relative to pre-illness (AD, p = 0.006; bvFTD, p < 0.0001). Similarly, present EC ratings for the bvFTD group were significantly lower than premorbid ratings (p < 0.0001). In contrast, EC ratings for AD patients remained stable across time (p = 0.111).
Correlations with disease severity/staging
Correlations between IRI current subscale ratings and cognitive and behavioral variables of interest are reported in Table 2. The results revealed distinct associations for each patient group.
In AD, no significant correlations between PT ratings and cognitive and behavioral variables were evident although an association on the threshold of significance was found between PT deficits and mental flexibility (Trails B-A, r = 0.459, p = 0.057). EC ratings in AD were significantly associated with behavioral dysfunction on the CBI (r = –0.388, p = 0.031). This association between EC and behavioral dysfunction on the CBI remained significant in AD even when controlling for level of functional impairment on the FRS (r = –0.482, p = 0.031). Finally, there was the suggestion that difficulties with emotion processing were associated with lower EC ratings in AD (r = 0.424, p = 0.058).
In bvFTD, PT ratings were significantly related to disease severity, as measured by the FRS (r = 0.388, p = 0.030), and behavioral dysfunction, as measured by the CBI (r = –0.349, p = 0.047). No significant correlations were found between EC ratings and cognitive or behavioral variables in bvFTD (all p values >0.1), although there was the suggestion that behavioral dysfunction was associated with poorer EC ratings in this group (CBI: r = –0.330, p = 0.058). Controlling for functional impairment on the FRS, rendered the associations between behavioral dysfunction and empathy ratings non-significant in the bvFTD group (PT: r = –0.130, p = 0.554; EC: r = –0.356, p = 0.095).
Relationship between empathy disruption and behavioral dysfunction
Finally, we explored the relationship between empathy performance and individual subscales of the CBI (corrected for multiple comparisons at p < 0.01). In AD, EC was significantly related to Stereotypical behaviors (r = –0.518, p = 0.009). In bvFTD, PT was found to correlate with Self-Care (r = –0.514, p = 0.010), whereas EC correlated with Loss of Motivation (r = –0.553, p = 0.005).
Analyses of covariance
To ensure that loss of empathy was not simply attributable to a decline in general cognitive functioning, we re-ran the main analyses controlling for cognitive function using the ACE-R total score. As before, a significant main effect of diagnosis was observed (F(2, 66) = 18.354, p < 0.0001), however this effect was now driven by an overall loss of empathy, irrespective of subscale, in the bvFTD group only (p < 0.0001). In contrast, AD patients were rated in line with Controls (p = 0.683). No main effect of empathy subscale was found (F(1, 66) = 1.821, p = 0.182).
Notably, the empathy subscale by diagnosis interaction effect observed in the original analysis was rendered not significant (F(2, 66) = 1.521, p = 0.226). Controlling for overall cognitive dysfunction served to ameliorate the PT deficit in the AD group, bringing their ratings in line with Controls (p = 0.193). In contrast, bvFTD patients remained impaired on both empathy subscales relative to Controls (p values <0.006) and AD patients (p values <0.0001).
In summary, loss of cognitive empathy in AD manifests largely as a product of general cognitive dysfunction, rather than a primary empathy impairment per se. For bvFTD patients, however, difficulties with cognitive and affective aspects of empathy cannot be explained in terms of global cognitive dysfunction and persist despite controlling for level of cognitive function.
Voxel-based morphometry analysis
Patterns of grey matter atrophy
Table 3 displays the patterns of grey matter intensity decrease in AD and bvFTD participants relative to Controls. AD patients showed widespread neural atrophy across medial temporal, frontal, parietal, and occipital regions of the brain compared to Controls. Significant atrophy was present in the frontopolar and frontoinsular cortices, bilaterally, medial temporal structures such as the bilateral hippocampus and thalamus, and extending posteriorly to include the bilateral supramarginal and angular gyri, lateral occipital cortices, and occipital poles.
BvFTD patients displayed characteristic grey matter intensity loss predominantly across frontoinsular and medial prefrontal regions, including the anterior cingulate cortex and orbitofrontal cortex, as well as lateral and medial temporal regions including the hippocampus, amygdala, and thalamus, bilaterally. Posterior regions were also significantly affected including the supramarginal and angular gyri, and the lateral occipital cortex,bilaterally.
Direct comparison of the patient groups revealed significantly greater atrophy in bvFTD relative to AD in the following regions; OFC, medial frontal cortex, temporal pole, caudate, and putamen, bilaterally. The reverse contrast failed to reveal any significant clusters at the p < 0.05 threshold. These patterns of atrophy are consistent with previous reports in AD [64] and bvFTD [65].
Neural correlates of cognitive empathy
Figure 2 and Table 4 display the significant regions to emerge from the covariate analysis investigating the neural correlates of Perspective Taking on the IRI in AD (A) and bvFTD (B). Reduced capacity for Perspective Taking in AD related to grey matter intensity decrease in predominantly left-sided regions including the left inferior, middle and superior temporal cortices, left angular gyrus, left parahippocampal gyrus, left cerebellum, and right middle temporal gyrus. In the bvFTD group, impairments in Perspective Taking were found to relate to grey matter intensity decrease in a largely bilateral set of frontoinsular regions including the inferior frontal gyrus, orbitofrontal and medial prefrontal cortices, anterior cingulate cortices, and insular cortices, bilaterally. Significant temporal lobe involvement was also noted including the temporal fusiform cortices and temporal poles, bilaterally as well as limbic structures including the right amygdala and bilateral hippocampi. Occipitoparietal brain regions were also significantly implicated in the bvFTD group, including the supramarginal gyrus, occipital pole, and cerebellum, bilaterally.
Neural correlates of Empathic Concern
Figure 3 and Table 5 display the significant regions to emerge from the covariate analysis investigating the neural correlates of Empathic Concern on the IRI in bvFTD. No significant clusters were evident for EC in the AD analyses, reflecting the fact that AD patients scored in line with Controls for this measure. Reduced capacity for Empathic Concern in bvFTD was found to relate to grey matter intensity decrease in predominantly left-sided regions including the left orbitofrontal cortex, left inferior frontal gyrus, left insular cortex, and the bilateral mid-cingulate gyrus. Subcortical structures including the left putamen and left thalamus were further implicated in empathic concern reduction in bvFTD. Lateral occipitoparietal regions including the left supramarginal gyrus left lateral occipital cortex were also involved.
Common regions underlying empathy disruption in bvFTD
To identify the regions commonly implicated across cognitive and affective empathy disturbance in bvFTD, we conducted an overlap analysis based on the statistical maps generated from the covariateanalyses. This overlap analysis revealed that the common structures implicated in reduced capacity for empathy, irrespective of subtype, in bvFTD were the left orbitofrontal and insular cortices (MNI coordinates: x = –38, y = 30, z = –10; Fig. 4).
DISCUSSION
The objective of this study was to investigate the capacity for empathy across cognitive and affective domains in bvFTD and AD, and to identify the neural substrates of empathy deficits in these syndromes. Our findings reveal important dissociations between bvFTD and AD patients in terms of their capacity to accurately understand and share the affective experience of others. Notably, a global disruption in the experience of empathy, irrespective of domain, was observed in bvFTD in keeping with their well-documented impairments in social cognition. In contrast, cognitive empathy was exclusively compromised in AD, in the context of a relative preservation of affective empathy. Controlling for overall cognitive functioning served to ameliorate cognitive empathy deficits in AD, yet impairments persisted in the bvFTD group across cognitive and affective subscales. This loss of empathy was found to relate to the vulnerability of a distributed network of regions, centered on the frontoinsular cortices, the integrity of which is crucial for successful social functioning.
A profound disruption in the capacity for empathy was found in bvFTD, in keeping with previous findings in the literature [7, 33] and reflecting the hallmark clinical features of behavioral egocentrism and interpersonal difficulties in this group [6]. These deficits emerged irrespective of empathy subdomain and persisted despite controlling for overall level of cognitive dysfunction. Voxel-based morphometry analyses permitted us to dissociate between current cognitive versus affective empathy disruption on the neuroanatomical level, revealing separable neural substrates of empathy loss in bvFTD. Impaired capacity for cognitive empathy in bvFTD was strongly associated with grey matter intensity decrease in a distributed set of regions including the lateral temporal cortices and temporal pole, OFC, medial PFC, and insular cortices bilaterally - key nodes of the “social brain” [66]. While this pattern of results was bilateral, right-sided involvement was particularly prominent, in keeping with previous reports of empathy disruption in neurodegenerative disorders using the IRI [7]. The right temporal pole has been suggested to play an integral role in the linking of sensory representations with emotional responses and social memory [67], facilitating the preferential processing of socially relevant concepts [68]. As such, damage to this region may disrupt knowledge of social concepts which are essential to accurately infer the perspective of another [69]. Our findings underscore the prominence of medial prefrontal regions in facilitating complex mental attribution processes [70] and highlight a specific vulnerability of medial prefrontal, frontoinsular, and anterior cingulate regions in the origin of cognitive empathy disruption in bvFTD [71]. Notably, these anterior brain regions largely recapitulate the Salience Network, which is proposed to play a unique role in processing emotionally salient internal and external stimuli, permitting the differentiation between self- and other [71, 72]. Importantly, our results add to a growing body of evidence emphasizing the importance of right-sided brain structures in mediating the characteristic interpersonal and socioemotional disturbances seen in this syndrome [17, 73–75].
Turning our attention to affective empathy disruption in bvFTD, our analyses revealed that grey matter intensity decrease in predominantly left-sided regions including the left OFC, left inferior frontal gyrus, left insular cortex, left thalamus, left putamen, and the bilateral mid-cingulate gyrus related to empathic concern deficits in this group. Our findings resonate with previous reports of significant left orbitofrontal, lateral temporal [30], and insula [33] involvement in socioemotional deficits in this syndrome, yet depart from a recent study demonstrating an association between right-sided anterior temporal and insular atrophy and the inaccurate estimation of affective empathy capacity in bvFTD [28]. As such, the neural substrates of the capacity for affective empathy diverge somewhat from those underpinning the accurate self-evaluation of this process. Interestingly, a recent study demonstrates that damage to left frontoinsular circuits in bvFTD disrupts prosocial reasoning with the suggestion that deviations from normative social behavior may, in part, reflect alterations in empathic concern [76]. Our finding of a prominent role for the mid-cingulate gyrus in empathic concern in bvFTD is in line with the view that this region occupies a central position in the neural circuits specialized for affective empathy [77]. Moreover, functional neuroimaging studies reveal robust activation of this region when healthy individuals immerse themselves in the affective context of another person, most prominently when that context requires empathy for personal distress or pain [5, 79]. As such, the mid-cingulate cortex is well placed to integrate affective with contextual information, and in coordination with the anterior insular, forms part of an affective-motivational network that is engaged in interoceptive awareness and the representation of emotional experiences [80]. Finally, our overlap analysis localized the left OFC and insular cortices as the common regions implicated in empathy disruption, irrespective of IRI subscale, in bvFTD in line with influential theories espousing a critical role for frontoinsular circuits in the origin of the hallmark socioemotional disturbances in this syndrome [71]. Damage to this frontoinsular network likely underlies much of the affective disturbance displayed by patients in their everyday lives, and potentially accounts for the consistent findings of grossly attenuated responses to the suffering of others in bvFTD [32].
In contrast, our findings suggest that perspective taking deficits in AD arise primarily as a consequence of cognitive dysfunction, rather than a primary impairment in empathy per se [15, 81]. Caregivers rated the capacity for cognitive empathy as being disrupted in AD, however, these impairments were ameliorated when overall cognitive functioning was controlled for in the analyses. The neural substrates of these deficits reflected atrophy primarily in left temporoparietal regions of the brain including the left temporal fusiform cortex, left inferior temporal gyrus, left angular gyrus, and bilateral middle and superior posterior temporal gyrus. These regions are consistently affected from early in the AD disease course with well documented alterations in volume [82] and glucose metabolism [83]. The temporoparietal junction is posited to serve a central role in representing the mental states of others [84] with the observation that damage to left temporoparietal structures impairs higher-level social reasoning [85]. Our findings resonate with a recent study which reported an impaired capacity to infer the beliefs of others in AD attributable to reduced metabolism in the left temporoparietal junction [23]. The authors proposed that with advancing disease severity, and increasing pathology in temporoparietal regions, social cognitive deficits likely become more pronounced in AD. This position is supported by longitudinal studies demonstrating significant decline in social cognitive functioning in AD with disease progression [37]. Our findings suggest, however, that the origin of these deficits are predominantly cognitively driven, in that controlling for global cognitive function ameliorated perspective-taking deficits in this group. This finding stands in contrast with the bvFTD group, in which the persistence of perspective-taking deficits likely reflects the disruption of core socioemotional processes, beyond that attributable to global cognitive dysfunction. Importantly, our findings underscore clinical and anecdotal reports of relatively spared capacity for affective empathy [8], with AD patients retaining a sense of warmth and compassion for others and a preserved sensitivity to detect salient socioemotional cues [71]. We suggest that it will be important for future studies to clarify the evolution of cognitive versus affective dimensions of empathy with disease progression in AD and to determine the neural substrates which support this spared capacity for empathic concern.
A number of methodological limitations warrant consideration in the current context. While our results converge well with previous studies using the IRI in bvFTD, the reliability and validity of this measure in neurodegenerative populations remains unknown. As we relied upon caregiver ratings of empathy in our patient groups, future studies using experimental tasks to investigate the expression of empathy in naturalistic settings are required to clarify how the deficits we report here manifest in the everyday environment of the individual. This is particularly relevant to determine the relationship between loss of empathy and the hallmark affective disturbances typically seen in FTD. Further, we compared caregiver ratings for patients with that of self-report in Controls. While we would not expect cognitively intact healthy older Controls to show a loss of empathy, it will be important to replicate our findings using informant ratings in the Control group. Notably, we did not find evidence of a significant relationship between facial emotion processing and affective empathy in FTD, a finding which may reflect some degree of separation between empathy and basic emotion processing. This proposal clearly warrants further exploration. Our voxel-based neuroimaging results did not survive conservative corrections for multiple comparisons (i.e., Family-Wise Error) and were therefore reported uncorrected at p < 0.001. We reduced the potential for false positive results, however, by employing a strict cluster extent threshold of 100 contiguous voxels [86]. Given our sample size and this application of strict thresholds, we are confident that our results do not represent false positives. In addition, co-atrophy effects represent a potential confound when dealing with neurodegenerative conditions displaying progressive brain atrophy across distributed regions, although our findings here dovetail well with previous reports in the literature. We suggest, however, that it will be important to replicate these findings in a larger patient cohort using corrected neuroimaging results and controlling for possible co-atrophy effects. Finally, given the recent shift in perspective from studying brain structures in isolation to considering the integrity of large-scale functional brain networks [71, 87], we suggest it will be important for future work to delineate how alterations in structural and functional connectivity particularly of frontoinsular regions of the Salience Network impinge upon the capacity for cognitive and affective empathy across dementia syndromes.
CONCLUSIONS
In summary, our findings corroborate the well-established observation of marked social cognitive deficits in bvFTD, manifesting across cognitive and affective forms of empathy. These socioemotional deficits reflect the degeneration of medial prefrontal, frontoinsular, and lateral temporal regions, which appear crucial for successful social cognitive functioning. Our results offer important insights into the neurocognitive mechanisms underpinning socioemotional changes in bvFTD and underscore the importance of frontoinsular regions in supporting successful social functioning.
Footnotes
ACKNOWLEDGMENTS
The authors are grateful to the patients and their families for their continued support of our research. This work was supported, in part, by the Australian Research Council Centre of Excellence in Cognition and its Disorders (CE110001021) and by ForeFront; a National Health and Medical Research Council of Australia Program Grant (1037746). MI is supported by an ARC Discovery Researcher Early Career Award (DE130100463). OP is supported by an NHMRC Senior Research Fellowship (APP1103258).
