INTRODUCTION
Social cognition refers to the set of cognitive processes involved in recognising, understanding, and responding to social cues (Beer and Ochsner, Reference Beer and Ochsner2006; Henry, von Hippel, Molenberghs, Lee, and Sachdev, Reference Henry, von Hippel, Molenberghs, Lee and Sachdev2016). It has been argued that this neurocognitive domain encompasses four broad components: (i) theory of mind (ToM), which refers to the ability to understand the mental states of others, (ii) empathy, which refers to the capacity to experience care or concern for others, (iii) social perception, which refers to the ability to recognise social cues, and (iv) social behaviour, the ability to appropriately present oneself in a social situation (Henry et al., Reference Henry, von Hippel, Molenberghs, Lee and Sachdev2016). Altogether, intact social cognitive function is pivotal in contributing to the development and maintenance of interpersonal relationships (Brodaty, Reference Brodaty1997; Carton, Kessler, and Pape, Reference Carton, Kessler and Pape1999; Sodian and Kristen, Reference Sodian, Kristen, Glatzeder, Goel and Müller2010) throughout the lifespan.
Consistently, research has shown that certain brain regions are particularly important for specific social cognitive domains. For instance, the amygdala appears to be crucial for processing positive and negative emotional valence stimuli, and the dorsal anterior cingulate cortex has been linked to empathy and concern (Allison, Puce, and McCarthy, Reference Allison, Puce and McCarthy2000; Berridge and Kringelbach, Reference Berridge and Kringelbach2013; Fan, Duncan, de Greck, and Northoff, Reference Fan, Duncan, de Greck and Northoff2011; Kanwisher and Yovel, Reference Kanwisher and Yovel2006; Roy, Shohamy, and Wager, Reference Roy, Shohamy and Wager2012; Van Overwalle, Reference Van Overwalle2009). More broadly, the limbic system, which encompasses core structures such as the amygdala and hippocampus, has been implicated in many aspects of social cognition (Adolphs, Reference Adolphs2010; Laurita & Nathan Spreng, Reference Laurita, Nathan Spreng, Hannula and Duff2017; LeDoux, Reference LeDoux2000). In addition to its role in processing emotionally salient stimuli, functional magnetic resonance imaging (fMRI) during face perception tasks has shown that the amygdala plays a key role in directing an individual’s gaze towards the eyes, most significantly in fearful faces, even if a cursor is initially fixated on the mouth region (Gamer & Buchel, Reference Gamer and Buchel2009). The hippocampus has been implicated in navigating social relationships and perception of interpersonal cues (Montagrin, Saiote, and Schiller, Reference Montagrin, Saiote and Schiller2018), although research has yet to elucidate its exact role in social cognition.
In Alzheimer’s disease (AD), prominent atrophy and tau deposition is observed in the hippocampus and amygdala and this may contribute to deficits in social cognition (Braak et al., Reference Braak, Alafuzoff, Arzberger, Kretzschmar and Del Tredici2006; Henry et al., Reference Henry, Ruffman, McDonald, O’Leary, Phillips, Brodaty and Rendell2008; Laisney et al., Reference Laisney, Bon, Guiziou, Daluzeau, Eustache and Desgranges2012; Martinez et al., Reference Martinez, Multani, Anor, Misquitta, Tang-Wai, Keren, Fox, Lang and Tartaglia2018). These neuropathological features are even evident in earlier prodromal phases, specifically in those with mild cognitive impairment (MCI) (Markesbery, Reference Markesbery2010). With evidence suggesting that these neuropathological features may be most pronounced in individuals with amnestic mild cognitive impairment (aMCI), a subgroup at greatest risk of progressing to AD (Fujie et al., Reference Fujie, Namiki, Nishi, Yamada, Miyata, Sakata, Sawamoto, Fukuyama, Hayashi and Murai2008; Kohler et al., Reference Kohler, Anselmo-Gallagher, Bilker, Karlawish, Gur and Clark2005; McCade et al., Reference McCade, Savage, Guastella, Hickie, Lewis and Naismith2013a; Petersen, Reference Petersen2004; Spoletini et al., Reference Spoletini, Marra, Di Iulio, Gianni, Sancesario, Giubilei, Trequattrini, Bria, Caltagirone and Spalletta2008).
Of clinical relevance, close informants (e.g., spouse/family member/friend/carer) of people with MCI also report changes in various aspects of behaviour such as everyday skills and motivation, which has been shown using measures such as the Revised Cambridge Behavioural Inventory (CBI-R) (Tsang, Diamond, Mowszowski, Lewis, and Naismith, Reference Tsang, Diamond, Mowszowski, Lewis and Naismith2012). These behavioural changes may in turn impact on their interpersonal relationships and perceived caregiver burden (McCade et al., Reference McCade, Savage, Guastella, Hickie, Lewis and Naismith2013b; Paradise et al., Reference Paradise, McCade, Hickie, Diamond, Lewis and Naismith2015). It is therefore important to gain a clearer understanding of how social cognitive deficits relate to behavioural changes in people with MCI, in order to improve awareness of early disease features, and to inform early interventions focused on optimising psychosocial and interpersonal functioning.
Thus far, the majority of studies that has examined social cognitive deficits in MCI has used tests of emotion recognition such as the Facial Expression of Emotion: Stimuli and Tests and ToM tasks such as the Reading the Mind in the Eyes Test (RMET) (Elferink, van Tilborg, and Kessels, Reference Elferink, van Tilborg and Kessels2015; McCade et al., Reference McCade, Savage, Guastella, Hickie, Lewis and Naismith2013a). Ever increasingly, however, false belief and referential communication tasks (see Moreau et al., Reference Moreau, Rauzy, Bonnefoi, Renié, Martinez-Almoyna, Viallet and Champagne-Lavau2015) have also been employed to assess ToM abilities. In fact, such tasks have been informative in highlighting that individuals with MCI reveal deficits in ToM abilities in natural conversation (Moreau et al., Reference Moreau, Rauzy, Bonnefoi, Renié, Martinez-Almoyna, Viallet and Champagne-Lavau2015). The RMET, in particular, is a well-validated social cognitive measure of relatively basic mental state decoding – requiring participants to make mentalistic inferences on the basis of observable features, specifically, via eye gaze cues (Baron-Cohen, Wheelwright, Hill, Raste, and Plumb, Reference Baron-Cohen, Wheelwright, Hill, Raste and Plumb2001; Immordino-Yang and Singh, Reference Immordino-Yang and Singh2013). There is now a large literature showing the RMET to be very sensitive to poorer ToM abilities in many different clinical groups (Chalah et al., Reference Chalah, Kauv, Lefaucheur, Hodel, Créange and Ayache2017; Heitz et al., Reference Heitz, Noblet, Phillipps, Cretin, Vogt, Philippi, Kemp, de Petigny, Bilger, Demuynck, Martin-Hunyadi, Armspach and Blanc2016). Furthermore, while only relatively few studies have explored this in MCI specifically, ToM abilities as indexed by the RMET have been shown to be clearly disrupted in those individuals with aMCI compared to healthy controls, even when RMET scores are adjusted for semantic fluency (Poletti & Bonuccelli, Reference Poletti and Bonuccelli2013). Altogether, research reveals that ToM abilities are clearly disrupted in MCI, with the greatest deficits observed in those with multi-domain aMCI (see Bora & Yener, Reference Bora and Yener2017 for meta-analysis).
Importantly, to date, there has been only limited empirical assessment of the neural networks that underlie poorer ToM abilities in MCI. One study of 16 people with aMCI used fMRI to examine neural networks associated with RMET performance (Baglio et al., Reference Baglio, Castelli, Alberoni, Blasi, Griffanti, Falini, Nemni and Marchetti2012). Possibly due to relatively small sample sizes, no neural network differences between aMCI and control participants during RMET performance were found. To the present authors’ knowledge, no study to date has examined how poorer ToM abilities may specifically relate to the integrity of the hippocampus and amygdala, despite these structures showing clear, yet subtle atrophic changes in MCI (Petersen et al., Reference Petersen, Parisi, Dickson, Johnson, Knopman, Boeve, Jicha, Ivnik, Smith, Tangalos, Braak and Kokmen2006), and despite their clear role within the limbic system.
Aims and Hypotheses
In this study, our first aim was to cross-validate previous research showing that social cognitive function, namely, ToM, as indexed via the RMET is impaired in people with MCI relative to controls. Based on our prior work (McCade et al., Reference McCade, Savage, Guastella, Hickie, Lewis and Naismith2013a), we hypothesised that social cognitive deficits would be evident in MCI and moreover would be most pronounced in the aMCI subgroup (compared to the naMCI subgroup). Our second aim was to determine whether poorer ToM abilities were volumetrically associated with key limbic structures. We predicted that poorer ToM abilities would be associated with smaller hippocampal and amygdalar volumes. A final exploratory aim was to determine whether ToM in MCI relates to functional and/or behavioural change, as rated by an informant.
METHODS
Sample
One hundred and fourteen individuals meeting criteria for single (n = 34) or multiple-domain (n = 80) MCI (Winblad et al., Reference Winblad, Palmer, Kivipelto, Jelic, Fratiglioni, Wahlund, Nordberg, Bäckman, Albert, Almkvist, Arai, Basun, Blennow, de Leon, DeCarli, Erkinjuntti, Giacobini, Graff, Hardy, Jack, Jorm, Ritchie, van Duijn, Visser and Petersen2004) were recruited from a specialist ‘Healthy Brain Ageing’ Clinic at the Brain and Mind Centre, The University of Sydney, Sydney, Australia. Refer to Clinical Ratings, below, for more information on the specific diagnostic criteria used in this study. This clinic receives referrals from neurologists, psychiatrists, geriatricians, and general practitioners and preferentially targets people over the age of 50 who have new onset cognitive and/or mood symptoms. In addition, 52 age- and education-matched healthy volunteers were recruited from the community as control participants. All participants gave a written informed consent, and all data were obtained in compliance with the Helsinki Declaration.
Inclusion criteria for all participants were as follows: aged between 50 and 75 years; English as a first language (with respect to validity of the standardised neuropsychological assessments); and a Mini-Mental State Examination Score (MMSE; Folstein, Folstein, and McHugh, Reference Folstein, Folstein and McHugh1975) ≥ 24. Exclusion criteria were nonaffective psychiatric disorder (e.g., schizophrenia) or neurological disorder (e.g., head injury, prior stroke or transient ischemic attack, epilepsy, and Parkinson’s disease); dementia (as determined by comprehensive clinical neuropsychological and psychiatric assessment); intellectual disability; current or past substance abuse; or impaired basic facial processing (as measured by a score of <41 on the Benton Facial Recognition Test; Benton et al., Reference Benton, Sivan, Hamsher, Varney and Spreen1994).
Measures
Clinical ratings
A neurologist or geriatrician conducted a structured clinical assessment for all participants to confirm inclusion/exclusion criteria. A semi-structured clinical interview was administered to determine lifetime and current major depression.
For clinical diagnosis of MCI, all participants needed to demonstrate at least a 1.5 standard deviation decline on standardised neuropsychological tests relative to premorbid estimates. MCI diagnosis was rated on consensus by two neuropsychologists and a specialist (e.g., geriatrician, psychiatrist, or neurologist). MCI was categorised further into amnestic (aMCI, n = 40) and nonamnestic (naMCI, n = 74) subgroups. Participants met aMCI criteria where they demonstrated clear evidence of impairment on either one or more delayed recall memory tasks (i.e., not just learning deficits). By contrast, naMCI criteria were defined by deficits on tests of other cognitive domains (e.g., processing speed, working memory, language, visuospatial, or executive function). Where participants demonstrated impairment on more than one cognitive domain, they were subsequently categorised as multiple-domain MCI. Overall, this study included participants with multiple-domain aMCI (n = 35), single-domain aMCI (n = 2), multiple-domain naMCI (n = 43), and single-domain naMCI (n = 32).
Neuropsychological assessment
A standardised neuropsychological test battery, described in detail elsewhere (Duffy et al., Reference Duffy, Lagopoulos, Hickie, Diamond, Graeber, Lewis and Naismith2014), was administered by a clinical neuropsychologist. For descriptive purposes, the MMSE was administered as a broad screening measure of cognitive functioning. Premorbid intellectual ability for each participant was estimated using the Wechsler Test of Adult Reading (Wechsler, Reference Wechsler2001). The following tests were included as outcomes of interest for this study:
(A) Rey Auditory Verbal Learning Test (RAVLT; Lezak, Reference Lezak1995): to measure unstructured verbal learning [i.e., total learning over five trials (RAVLT-1-5; maximum = 60)] and delayed recall [i.e., percent retention scores (RAVLT%), calculated as (Trial 7/Trial 5)*100].
(B) Wechsler Adult Intelligence Scale–III Digit Span subtest (Digit Span; Wechsler Reference Wechsler1997): to assess auditory working memory.
(C) Delis–Kaplan Executive Function System Color-Word Interference Test, Condition 3 (inhibition/switching) (DKEFS CWIT; Delis, Kaplan and Kramer, Reference Delis, Kaplan and Kramer2001): to measure response inhibition with concurrent cognitive flexibility, as aspects of higher-level executive functions.
(D) Controlled Oral Word Association Test (COWAT): to measure phonemic verbal fluency (letters ‘F’, ‘A’, and ‘S’) (COWAT-FAS; Ruff, Light, Parker and Levin, Reference Ruff, Light, Parker and Levin1996) and semantic fluency (animal names) (COWAT-Animals; Tombaugh, Kozak and Rees, Reference Tombaugh, Kozak and Rees1999).
(E) Boston Naming Test (Goodglass, Kaplan and Weintraub, Reference Goodglass, Kaplan and Weintraub1983): to assess naming to confrontation.
(F) Trail Making Test Part A and Part B (TMT-A and TMT-B; Reitan, Reference Reitan1979): to measure psychomotor speed and set-shifting/cognitive flexibility, respectively.
In this study, the RMET (Baron-Cohen et al., Reference Baron-Cohen, Wheelwright, Hill, Raste and Plumb2001) was administered to assess participants’ social cognitive function. The 36-item assessment requires the participant to infer mental/emotional states from a photograph showing the isolated eye region of an unknown person’s face. The participant must infer the mental/emotional state of that person from four options (e.g., terrified, upset, arrogant, or annoyed). The total raw score is calculated based on the number of total correct responses. In this study, age-adjusted z scores were calculated from normative data (Baron-Cohen et al., Reference Baron-Cohen, Wheelwright, Hill, Raste and Plumb2001) and used for subsequent analysis.
Self-report
a) Mood: To determine depression and anxiety symptom severity, participants completed the Hospital Anxiety and Depression Scale (Zigmond & Snaith, Reference Zigmond and Snaith1983). Scores for the entire scale range from 0 to 42, whereby individual scores for each subscale (anxiety and depression) range from 0 to 21. Higher scores indicate a greater severity of symptoms.
b) Behavioural changes: To determine any current neurobehavioural changes, close informants (spouse/family member/friend/carer) of both MCI and control participants completed the CBI-R (Wear et al., Reference Wear, Wedderburn, Mioshi, Williams-Gray, Mason, Barker and Hodges2008). This 45-item questionnaire measures observed changes in cognitive abilities, everyday function, and neuropsychiatric symptoms. Specifically, it assesses change in the domains of memory and orientation, everyday skills, self-care, abnormal behaviour, mood, delusions and hallucinations, eating habits, sleep, stereotypical and motor behaviour, and motivation. Close informants were asked to rate the frequency (0 – none; 1 – a few times per month; 2 – a few times per week; 3 – daily; and 4 – constantly) of potentially altered patterns of behaviour, over the last month.
Magnetic resonance imaging and volume segmentation analysis
A subsample of 84 participants (n = 17 aMCI multiple domain; n = 23 naMCI multiple domain; n = 1 aMCI single domain; n = 18 naMCI single domain; and n = 25 controls) underwent magnetic resonance imaging (MRI) scanning using a 3T GE Discovery MR750 scanner (GE Medical Systems, Milwaukee, WI, USA) within 2 weeks of their neuropsychological assessment at the Brain and Mind Centre imaging facility. As described previously (Elcombe et al., Reference Elcombe, Lagopoulos, Duffy, Lewis, Norrie, Hickie and Naismith2015), an eight-channel phased-array head coil using a T1-weighted magnetisation-prepared rapid gradient-echo sequence was used producing (196 sagittal slices, Repetition Time (TR) = 7.2 ms; Echo Time (TE) = 2.8 ms; flip angle = 10°; matrix 256 × 256; .9 mm isotropic voxels). For each participant, two T1-weighted scans were obtained, of which the sequence with the higher signal-to-noise ratio was used. Regions of interest were the amygdala and the hippocampus. Both left and right hemispherical regional volumes, as described previously (Elcombe et al., Reference Elcombe, Lagopoulos, Duffy, Lewis, Norrie, Hickie and Naismith2015), were extracted using a semi-automated segmentation routine based on the principles of the Active Shape and Appearance Models within a Bayesian framework as implemented by ‘FMRIB’s Integrated Registration and Segmentation Tool (FIRST)’ in FMRIB Software Library. Briefly, T1 data were reoriented, converted to Neuroimaging Informatics Technology Initiative (NIFTI) format, and skull-stripped, and the FIRST routine was applied to estimate regional volumes. As a part of the segmentation routine, data were registered to Montreal Neurological Institute (MNI) standard space and boundary correction was applied, with data visually inspected for errors. Total volume for each structure was calculated from the addition of boundary and intracranial volume corrected left and right regional volumes.
Statistical Analyses
All analyses were conducted using SPSS Version 24 (SPSS Inc., Chicago, IL, USA). To examine within- and between-group data, one-way analysis of variance with post hoc Tukey tests was used in addition to Levene’s test for equality of variance. Similarly, an analysis of covariance was performed to examine within- and between-group data, covarying for age. Where variance differed between groups or following a non-normal distribution, the Kruskal–Wallis nonparametric or the Mann–Whitney U test was used, with a Bonferroni correction for multiple comparisons. Categorical data (e.g., gender) were analysed using Pearson’s Chi-squared test. Continuous data were analysed using Spearman’s rank-order correlation (ρ) and nonparametric partial correlations, as appropriate for the data distribution. All correlations between ToM abilities and brain structure volumes were performed list-wise, and for all analyses an α level of .05 was applied.
For imaging analyses, one aMCI case was found to be a significant outlier (for both the amygdala and hippocampus) and was thus curtailed to the next highest score in the sample.
RESULTS
Demographic and Neuropsychological Characteristics
Table 1 outlines the clinical and demographic characteristics of the sample. There were no significant differences between the MCI and control groups in terms of mean age, years of education, or estimated premorbid intellectual ability. However, as expected, the MCI group had lower scores on the MMSE (p = .001) and higher symptoms of depression (p = .04). Within the control group, there was a higher proportion of females to males (p = .04). Additionally, post hoc analyses revealed significant differences on a number of clinical variables when comparing the MCI subgroups with the control group. Specifically, participants in the aMCI subgroup (n = 37) were significantly older than both controls (p = .03) and those with naMCI (n = 77, p = .01). Moreover, lower MMSE scores were observed in the aMCI subgroup compared to both controls (p < .001) and the naMCI subgroup (p = .002).
Demographic and clinical data for control and mild cognitive impairment groups (mean ± SD)
MCI = mild cognitive impairment; IQ = intelligence quotient; MMSE = Mini-Mental State Examination; HADS = Hospital Anxiety and Depression Scale.
* Comparative differences between the control and MCI groups only.
a MCI (n = 111); aMCI (n = 36); naMCI (n = 75).
b Control (n = 52); MCI (n = 100); aMCI (n = 35); naMCI (n = 65).
Table 2 outlines the neuropsychological characteristics of the sample. Compared to the control group, the MCI group revealed significantly poorer performance on all neuropsychological tests, with the greatest differences observed in the domains of verbal learning (RAVLT 1-5) and higher-level cognitive flexibility (TMT-B). As expected, the control and MCI groups did not differ on basic facial processing as measured by the Benton Facial Recognition Test (Benton et al., Reference Benton, Sivan, Hamsher, Varney and Spreen1994). Further post hoc analyses revealed that the aMCI subgroup performed significantly more poorly than the control group on all neuropsychological tests (p < .05 for all), while the naMCI subgroup was significantly (i.e., p < .05) worse than controls only on tasks measuring cognitive flexibility (TMT-B), working memory (Digit Span), verbal fluency (COWAT-FAS), and semantic fluency (COWAT-Animals). In comparison with each other, the two subgroups differed only in performance on the learning and memory tasks as well as semantic fluency, with aMCI demonstrating poorer performance than naMCI (RAVLT 1-5 and RAVLT%, p < .001; COWAT-Animals, p = .01)
Neuropsychological data for control and mild cognitive impairment groups (mean ± SD)
MCI = mild cognitive impairment; RMET = Reading the Mind in the Eyes Test; BFRT = Benton Facial Recognition Test; RAVLT 1-5 = Rey Auditory Verbal Learning Test-total learning over 5 trials; RAVLT % = Rey Auditory Verbal Learning Test-percent retention scores (i.e., (Trial 7/Trial 5) * 100); DKEFS CWIT = Delis–Kaplan Executive Function System Color-Word Interference Test, Condition 3 (inhibition/switching); COWAT-FAS = Controlled Oral Word Association Test with letters ‘F’, ‘A’, and ‘S’; COWAT-Animals = Controlled Oral Word Association Test with animal names; TMT-A = Trail Making Test Part A; TMT-B = Trail Making Test Part B.
a Control (n = 44).
b Control (n = 51).
c Control (n = 46) and MCI (n = 87).
d Control (n = 50) and MCI (n = 110).
e MCI (n = 113).
In addition, as outlined in Table 3, post hoc analyses revealed that the aMCI subgroup that underwent MRI scanning (n = 18) performed significantly more poorly than the control group (n = 25) on the learning and memory task (RAVLT 1-5 and RAVLT%, p < .05) and tasks of concurrent cognitive flexibility (DKEFS CWIT Condition 3 and TMT-B, p < .05). Similarly, the naMCI subgroup (n = 41) performed significantly more poorly than the control group on working memory (Digit Span, p < .05), one aspect of the learning and memory task (RAVLT 1-5, p < .05), and one task of concurrent cognitive flexibility (TMT-B, p < .05).
RMET, CBI-R, neuropsychological, demographic, and clinical variables for control and mild cognitive impairment subgroup participants who underwent MRI scanning (mean ± SD)
MCI = mild cognitive impairment; aMCI = amnestic mild cognitive impairment; naMCI = non-amnestic mild cognitive impairment; RMET = Reading the Mind in the Eyes Test; CBI-R = Cambridge Behavioural Inventory – Revised; BFRT = Benton Facial Recognition Test; RAVLT 1-5 = Rey Auditory Verbal Learning Test-total learning over 5 trials; RAVLT % = Rey Auditory Verbal Learning Test-percent retention scores (i.e., (Trial 7/Trial 5) * 100); DKEFS CWIT = Delis–Kaplan Executive Function System Color-Word Interference Test, Condition 3 (inhibition/switching); COWAT-FAS = Controlled Oral Word Association Test with letters ‘F’, ‘A’, and ‘S’; COWAT-Animals = Controlled Oral Word Association Test with animal names; TMT-A = Trail Making Test Part A; TMT-B = Trail Making Test Part B; IQ = intelligence quotient; MMSE = Mini-Mental State Examination; HADS = Hospital Anxiety and Depression Scale.
* Significant difference compared with the aMCI subgroup.
** Significant difference compared with the naMCI subgroup.
*** Significant difference when the aMCI subgroup is compared to the naMCI subgroup.
a Control (n = 20); aMCI (n = 14); naMCI (n = 34).
b Control (n = 22).
c Control (n = 24).
d Control (n = 24); aMCI (n = 17); naMCI (n = 37).
e Control (n = 24), aMCI (n = 17) and naMCI (n = 38).
f Control (n = 24) and naMCI (n = 40).
g aMCI (n = 17); naMCI (n = 39).
h aMCI (n = 16); naMCI (n = 32).
Social Cognitive Function: RMET
Table 2 shows RMET performance for the control and MCI groups. As a whole, participants with MCI demonstrated significantly poorer performance on the RMET compared to the control group, with the effect size approaching moderate in magnitude (d = .44). However, as shown in Table 4, when performance on this task is further compared among healthy controls and the MCI subgroups, it appears that this difference is driven by those with aMCI (p = .03) rather than naMCI, even when covarying for age (p = .01). There were no significant differences between aMCI and naMCI subgroups on the RMET.
RMET, CBI-R, and neuroimaging variables for control and mild cognitive impairment subgroups (mean ± SD)
MCI = mild cognitive impairment; aMCI = amnestic mild cognitive impairment; naMCI = non-amnestic mild cognitive impairment; RMET = Reading the Mind in the Eyes Test; CBI-R = Cambridge Behavioural Inventory – Revised.
* Significant difference compared with the aMCI subgroup.
** Significant difference compared with the naMCI subgroup.
*** Significant difference when the aMCI subgroup is compared to the naMCI subgroup.
a Control (n = 39); aMCI (n = 29); naMCI (n = 63).
b Control (n = 38); aMCI (n = 29); naMCI (n = 63).
c Control (n = 25); aMCI (n = 18); naMCI (n = 41).
Association Between RMET, Clinical Variables, and Regional Brain Volumes
Within the control group (n = 25), superior RMET performance was moderately associated with larger total hippocampal volume (Table 5), a finding which remained significant even after controlling for age (ρ = .52, p = .01). When controlling for both age and predicted IQ, this finding remained significant (ρ = .49, p = .02). Moreover, after controlling for age, semantic fluency, predicted IQ, and naming to confrontation, the results did not change (ρ = .49, p = .02). For the MCI group (n = 59) generally, and for the MCI subgroups (aMCI n = 18, naMCI n = 41), there were no significant correlations between RMET and hippocampal or amygdalar volume.
Correlations between RMET performance, neuroimaging variables, and neuropsychological data for control and MCI groups
MCI = mild cognitive impairment.
* p < .01.
To determine whether the relationship between RMET and hippocampal volume could be influenced by other variables, we then examined correlations between RMET performance with demographic, neuropsychological, and neuroimaging measures (Table 6). Within the MCI group, poorer RMET performance correlated with greater age (ρ = .29, p = .002), lower levels of education (ρ = .23, p = .01), and lower predicted IQ (ρ = .20, p = .04). Furthermore, within the MCI group, greater performance on higher-level skill tasks such as Digit Span (ρ = .21, p = .02) and TMT-B (ρ = .33, p = .0004), correlated with greater RMET performance.
Correlations between RMET performance and demographic, neuropsychological, and clinical variables for control and MCI groups
MCI = mild cognitive impairment; HADS = Hospital Anxiety and Depression Scale; CBI-R = Revised Cambridge Behavioural Inventory; DKEFS CWIT = Delis–Kaplan Executive Function System Color-Word Interference Test, Condition 3 (inhibition/switching); TMT-B = Trail Making Test Part B.
* p < .01;
** p < .05;
*** p < .001.
Upon further MCI subgroup analyses, greater RMET performance correlated with only greater TMT-B performance for both the aMCI (ρ = .34, p = .04) and the naMCI (ρ = .3, p = .01) subgroups. Further, MCI subgroup analysis showed that for the naMCI subgroup, poorer RMET performance correlated with lower levels of education (ρ = .24, p = .03) and younger age in the aMCI subgroup (ρ = −.40, p = .02). No other correlations were observed between the MCI subgroups’ performance on the RMET, neuropsychological measures, clinical variables, and regional brain volumes.
Association Between RMET and Behavioural Changes as Measured by a Close Informant
Finally, we examined the degree to which RMET performance was related to clinical outcomes. Table 6 displays correlations between RMET performance and CBI-R subscales. Within the MCI group, poorer RMET performance correlated with increased carer-reported difficulties in both the CBI-R memory and everyday skills subscales. When controlling for age within the MCI group, the correlation remained significant between both poorer RMET performance and increased difficulties on the CBI-R everyday skills (ρ = −.25, p = .02) and the memory and orientation subscale (ρ = −.23, p = .03).
DISCUSSION
This is the largest known study to examine ToM in individuals with MCI using the RMET, and to the authors’ knowledge, the first to determine how ToM in MCI relates to volumetric changes in key limbic structures, namely, the amygdala and hippocampus. Broadly, these results align with prior work which has shown that deficits in areas of social cognitive function are evident in MCI involving an array of social cognitive tasks and may be more prominent in the aMCI subtype (McCade et al., Reference McCade, Savage, Guastella, Hickie, Lewis and Naismith2013a, Reference McCade, Savage, Guastella, Hickie, Lewis and Naismith2013b). However, the results also make an important novel contribution by showing that while the hippocampus is linked to social cognition in healthy controls, this relationship is absent in those with MCI. Of clinical significance, we further found that for individuals with MCI, poorer ToM abilities are associated with a decline in everyday skills such as difficulties in writing, handling money, or paying bills.
Our finding that ToM difficulties are pronounced in the aMCI subgroup compared to healthy controls is consistent with prior work focused on a different area of social cognition (facial affect recognition) (McCade et al., Reference McCade, Savage, Guastella, Hickie, Lewis and Naismith2013a), as well as other reports employing the RMET (Poletti & Bonuccelli, Reference Poletti and Bonuccelli2013) and other false belief and referential communication tasks (Moreau et al., Reference Moreau, Rauzy, Bonnefoi, Renié, Martinez-Almoyna, Viallet and Champagne-Lavau2015). Together, these studies therefore support the importance of early assessment of social cognitive change in this subgroup that is known to be of greatest risk of progressing to dementia (Gauthier et al., Reference Gauthier, Reisberg, Zaudig, Petersen, Ritchie, Broich, Belleville, Brodaty, Bennett, Chertkow and Cummings2006; Petersen, Reference Petersen2004). This is because the same social cognitive deficits seen in individuals with aMCI (e.g., problems recognising facial emotions and engaging in ToM inferences) have also been identified using similar methodological approaches in studies focused on mild AD (Henry et al., Reference Henry, Ruffman, McDonald, O’Leary, Phillips, Brodaty and Rendell2008; Kohler et al., Reference Kohler, Anselmo-Gallagher, Bilker, Karlawish, Gur and Clark2005; Laisney et al., Reference Laisney, Bon, Guiziou, Daluzeau, Eustache and Desgranges2012; Spoletini et al., Reference Spoletini, Marra, Di Iulio, Gianni, Sancesario, Giubilei, Trequattrini, Bria, Caltagirone and Spalletta2008; Weiss et al., Reference Weiss, Kohler, Vonbank, Stadelmann, Kemmler, Hinterhuber and Marksteiner2008).
While some studies have identified intact social cognitive function (specifically emotion recognition) in aMCI subgroups, it is important to note that these studies included both single- and multiple-domain aMCI (Bediou et al., Reference Bediou, Ryff, Mercier, Milliery, Henaff, D’Amato, Bonnefoy, Vighetto and Krolak-Salmon2009; Henry et al., Reference Henry, Ruffman, McDonald, O’Leary, Phillips, Brodaty and Rendell2008). Because the single-domain subgroup by definition has more circumscribed impairments in memory alone, this may explain the different pattern of results identified in these studies. In the present study, we focussed on multiple-domain MCI because this subgroup has been shown to convert to dementia more frequently (Serrano, Dillon, Leis, Taragano, and Allegri, Reference Serrano, Dillon, Leis, Taragano and Allegri2013). In particular, the multiple-domain aMCI subgroup appears to exhibit greater neuropathological changes, poorer patient outcomes, and greater burden on their respective carers (Alexopoulos, Grimmer, Perneczky, Domes, and Kurz, Reference Alexopoulos, Grimmer, Perneczky, Domes and Kurz2006; Hunderfund et al., Reference Hunderfund, Roberts, Slusser, Leibson, Geda, Ivnik, Tangalos and Petersen2006; McCade et al., Reference McCade, Savage, Guastella, Hickie, Lewis and Naismith2013b; Paradise et al., Reference Paradise, McCade, Hickie, Diamond, Lewis and Naismith2015; Petersen and Negash, Reference Petersen and Negash2008).
One important finding from this study was the identification of a robust relationship between ToM and hippocampal volume in the healthy control group, even after controlling for age, predicted IQ, semantic fluency, and naming to confrontation. This finding is not only consistent with other work that has shown that the hippocampus plays a critical role in social cognition (Heitz et al., Reference Heitz, Noblet, Phillipps, Cretin, Vogt, Philippi, Kemp, de Petigny, Bilger, Demuynck, Martin-Hunyadi, Armspach and Blanc2016; Perry, Hendler, and Shamay-Tsoory, Reference Perry, Hendler and Shamay-Tsoory2011; Montagrin et al., Reference Montagrin, Saiote and Schiller2018) but also extends this literature in an interesting way. Specifically, the current work suggests that, beyond the hippocampus’ active role in the navigation of social relationships and perception of interpersonal cues, hippocampal integrity, at least in part, may serve a more specific aspect of social cognition: ToM function in older adults who show no clinical signs of cognitive decline (Calder et al., Reference Calder, Keane, Manly, Sprengelmeyer, Scott, Nimmo-Smith and Young2003; Isaacowitz et al., Reference Isaacowitz, Lockenhoff, Lane, Wright, Sechrest, Riedel and Costa2007; Laurita & Nathan Spreng, Reference Laurita, Nathan Spreng, Hannula and Duff2017). Nevertheless, although a robust relationship between ToM and hippocampal volume in the healthy control group was identified, no relationship was observed with amygdalar volume. This finding, which is contrary to our initial hypotheses given our understanding that the amygdala is associated with the processing of positive and negative emotional valence stimuli, requires further investigation (Allison et al., Reference Allison, Puce and McCarthy2000; Berridge and Kringelbach, Reference Berridge and Kringelbach2013; Fan et al., Reference Fan, Duncan, de Greck and Northoff2011; Singer et al., Reference Singer, Seymour, O’Doherty, Kaube, Dolan and Frith2004; Van Overwalle Reference Van Overwalle2009).
The other major finding to emerge was that this relationship between ToM and hippocampal volume was not observed within the MCI group. This may suggest that the social cognitive difficulties seen in individuals with MCI (and perhaps also early dementia) may reflect a breakdown in a more distributed neural network rather than a reduction in the structural integrity of specific limbic structures implicated in social cognition such as the amygdala and hippocampus (Bell-McGinty, Lopez, Meltzer, and et al., Reference Bell-McGinty, Lopez, Meltzer, Scanlon, Whyte, Dekosky and Becker2005; Petersen et al., Reference Petersen, Parisi, Dickson, Johnson, Knopman, Boeve, Jicha, Ivnik, Smith, Tangalos, Braak and Kokmen2006; Schott, Kennedy, and Fox, Reference Schott, Kennedy and Fox2006). For example, previous studies have demonstrated that individuals with aMCI have diminished functional connectivity within the default mode network (DMN), when compared to individuals with naMCI (Dunn et al., Reference Dunn, Duffy, Hickie, Lagopoulos, Lewis, Naismith and Shine2014). In fact, in other age groups with healthy adults, fMRI studies have demonstrated that significant overlap exists between the activation of the DMN and social brain networks underpinning social cognitive tasks such as the RMET (Mars et al., Reference Mars, Neubert, Noonan, Sallet, Toni and Rushworth2012; Schilbach, Eickhoff, Rotarska-Jagiela, Fink, and Vogeley, Reference Schilbach, Eickhoff, Rotarska-Jagiela, Fink and Vogeley2008). Additionally, growing evidence now suggests that at least some aspects of ToM also recruit the inferior–posterior sector of the posteromedial cortex, encompassing portions of the posterior cingulate and precuneus (Immordino-Yang and Singh, Reference Immordino-Yang and Singh2013; Sestieri, Corbetta, Romani, and Shulman, Reference Sestieri, Corbetta, Romani and Shulman2011). Consequently, further investigations are warranted to understand the extent of broader neural networks on poorer ToM abilities in individuals with MCI.
The present study has potentially important implications for broader clinical and psychosocial functioning in aMCI. Firstly, early changes in ToM are evident in aMCI and may even be one of the earliest clinical signs (McCade et al., Reference McCade, Savage, Guastella, Hickie, Lewis and Naismith2013a, Reference McCade, Savage, Guastella, Hickie, Lewis and Naismith2013b; (McCade et al., Reference McCade, Savage, Guastella, Hickie, Lewis and Naismith2013a, Reference McCade, Savage, Guastella, Hickie, Lewis and Naismith2013b; Phillips, Scott, Henry, Mowat, and Bell, Reference Phillips, Scott, Henry, Mowat and Bell2010). While future research studies should continue to examine social cognition in MCI and preclinical AD to elucidate early changes in this complex cognitive process, neuropsychologists and clinicians should also more routinely assess social cognition in clinical settings early in the disease course (Henry et al., Reference Henry, von Hippel, Molenberghs, Lee and Sachdev2016). Importantly, in the present study, we found that poorer ToM abilities even in this very early stage of cognitive decline are associated with informant-observed behavioural changes. Specifically, poorer ToM abilities were associated with poorer everyday skills in social contexts (CBI-R), a result that has previously been reported in individuals with MCI (Tsang et al., Reference Tsang, Diamond, Mowszowski, Lewis and Naismith2012), although not linked to ToM deficits. However, our study adds to this knowledge by showing (for the first time) a relationship between poorer ToM abilities as well as greater difficulties with everyday skills such as writing, handling money, or paying bills in those individuals with MCI. In terms of educating carers and family members, it will therefore be important for health professionals to communicate the nature and extent of these social cognitive deficits, alongside adaptive or compensatory strategies to facilitate communication. Since prior studies have shown that broader social cognitive deficits are associated with increased carer burden, programs specifically targeting these symptoms should be developed and evaluated (McCade et al., Reference McCade, Savage, Guastella, Hickie, Lewis and Naismith2013b). In psychiatric disorders, social cognitive training is a robust method of intervention and may even be coupled with cognitive training for targeting a more holistic range of cognitive deficits (Bogdanova, Yee, Ho, and Cicerone, Reference Bogdanova, Yee, Ho and Cicerone2016; Cacciotti-Saija et al., Reference Cacciotti-Saija, Langdon, Ward, Hickie, Scott, Naismith, Moore, Alvares, Redoblado-Hodge and Guastella2015).
In addition to behavioural interventions, there is now a considerable literature for disorders that are characterised by a core deficit in social functioning such as autism spectrum disorder and schizophrenia, suggesting that intranasal administration of the neuropeptide-hormone oxytocin (OT) may enhance social cognitive function, including performance on the RMET (Guastella et al., Reference Guastella, Einfeld, Gray, Rinehart, Tonge, Lambert and Hickie2010; Guastella & MacLeod, Reference Guastella and MacLeod2012; Guastella et al., Reference Guastella, Mitchell and Mathews2008; Guastella et al., Reference Guastella, Ward, Hickie, Shahrestani, Hodge, Scott and Langdon2015). Although poorer social cognitive function is not traditionally recognised as a core deficit in MCI (perhaps because it is not routinely assessed), a few studies have examined OT’s potential in ageing, with one feasibility study in frontotemporal dementia (FTD) showing tolerance of high doses of intranasal oxytocin (INOT; up to 72 IU) even revealing improvements in apathy, empathy, and overall patient–caregiver interactions (Finger et al., Reference Finger, MacKinley, Blair, Oliver, Jesso, Tartaglia, Borrie, Wells, Dziobek, Pasternak and Mitchell2015), when compared to FTD participants receiving placebo. Extending on this, a study investigating the role of INOT administration on all four domains of social cognitive function and carer–patient interactions to ultimately reduce caregiver burden in older people with AD is currently underway within our research team (ACTRN12617001531303). If found to be effective, and given the present results, it may be worthwhile for future research to explore the use of INOT at earlier stages of the neurodegenerative pathway, that is, MCI.
There were some limitations to this study. Firstly, the current study was cross-sectional in nature. In order to better understand the importance of the hippocampus and amygdala for social cognition in ageing and MCI, longitudinal investigation would be invaluable. In addition, the MCI sample in the present study were relatively well educated (average of 13.9 years), as were the control group, and this may limit the generalisability of our results to the broader population. In particular, since both education and IQ have been considered to be neuroprotective factors in the decline of mental function and social cognition (Brayne et al., Reference Brayne, Ince, Keage, McKeith, Matthews, Polvikoski and Sulkava2010), it is possible that those with lower rates of education may demonstrate even greater social cognitive impairment than those observed in this sample. Further research in those living with MCI could perhaps investigate whether higher rates of education translate to a greater cognitive reserve or ‘buffer’ against more deleterious effects on a significant other (spouse/family members/friends) in relation to social cognition. Moreover, while functional skills were associated with ToM abilities, the authors note that this study did not include an extensive assessment of executive function. Specifically, key higher-level skills for everyday functioning, such as planning, organisational skills, judgment/reasoning, abstract thinking, and problem-solving, were not assessed here and may therefore be of further interest. In addition, the MRI subsample was not representative of the broader sample as they had poorer performance on tasks of cognitive flexibility compared to healthy controls, a factor that is relevant given that cognitive flexibility was associated with greater ToM ability.
Lastly, it is worth noting that while the RMET is a well-validated measure of ToM (Baron-Cohen et al., Reference Baron-Cohen, Wheelwright, Hill, Raste and Plumb2001; Immordino-Yang & Singh, Reference Immordino-Yang and Singh2013), it focuses on only one of the four core domains of social cognitive function. Thus, to further understand the extent to which specific social cognitive deficits relate to underlying brain changes in limbic structures such as the hippocampus and amygdala, additional tasks that tap into other domains of social cognition would be important to explore (McDonald et al., Reference McDonald, Bornhofen, Shum, Long, Saunders and Neulinger2006). Furthermore, given that other brain structures such as the ventromedial prefrontal cortex are also heavily relied on for ToM, future studies should ideally include this region to further understand how it is implicated for ToM abilities in MCI.
In summary, the present study builds and extends on a growing body of evidence demonstrating that social cognitive deficits, in this case ToM, are evident in individuals with MCI, and are particularly pronounced in the aMCI subtype. The novel neuroimaging results of this study showing that ToM abilities are significantly correlated with hippocampal volume in healthy controls support the role of the hippocampus in social cognition. With the absence of this correlation in the MCI group, a breakdown in a more distributed network rather than a reduction in the structural integrity of isolated limbic structures may be implicated. These findings therefore add to a growing literature that suggests it is clinically important to include measures of social cognitive function in the routine clinical workup of individuals with suspected MCI. In addition, it may prove useful to develop and test targeted treatments directed at communication and relationship quality among patients and families, for this critical component of psychosocial interaction.
Acknowledgements
Dr L. Mowszowski is supported by a National Health and Medical Research Council (NHMRC) – Australian Research Council (ARC) Dementia Research Development Fellowship (ID 1109618). Prof S. Naismith is supported by an NHMRC Dementia Leadership Fellowship (ID 1135639). Prof J. D. Henry is supported by an Australian Research Council Future Fellowship (FT170100096).
CONFLICTS OF INTEREST
The authors have nothing to disclose.
