Assessment of Patient Self-awareness and Related Neural Correlates in Frontotemporal Dementia and Corticobasal Syndrome

Abstract

Objective

We compared two different methods of assessing self-awareness (clinician-rated vs. self- and caregiver report) in participants with neurodegenerative conditions. Additionally, we examined the contribution of memory dysfunction to assessment of self-awareness.

Method

Sixty-seven participants with various neurodegenerative disorders participated in this study. Data were collected on brain volume, neurocognitive function, demographic characteristics, and two measures of patient self-awareness, defined as (1) the discrepancy between patient and caregiver ratings of dysexecutive syndrome and (2) clinician-observed rating of patient insight. Penalized regression with best subset variable selection and 10-fold cross-validation was used to evaluate three neurocognitive frameworks: self-regulation, language, and perspective-taking, each predicting the results from the two methods of self-awareness measurement.

Results

The self-regulation framework was more robustly predictive for both the clinician rating and discrepancy method than language or perspective-taking. Frameworks in which the clinician rating was the criterion were more robust than those with the discrepancy method as criterion. When a measure of memory functioning was added to the framework, there was no appreciable improvement in the prediction of self-awareness.

Conclusions

A self-regulation neurocognitive framework, consisting of regions of interest and neuropsychological test scores, was more effective in understanding patient self-awareness than perspective-taking or language frameworks. Compared to the discrepancy method, a clinician rating of self-awareness was more robustly associated with relevant clinical variables of regional brain volume and neuropsychological performance, suggesting it may be a useful measure to aid clinical diagnosis.

Introduction

Self-awareness, or insight, involves the accurate appraisal and understanding of different aspects of the self, including one’s competence in performing different acts (Sherer & Fleming, 2014). In clinical settings, a deficit in self-awareness resulting from damage to the brain is also termed anosognosia (“without disease knowledge”), and is characterized by an inability to recognize one’s own neurologic or psychiatric impairment (i.e., cognitive, physical, behavioral, or emotional) (Zamboni et al., 2013). Level of self-awareness is of considerable importance in clinical settings and is a factor in determining to what extent an individual can benefit from therapy (Prigatano & Fordyce, 1986). Consequences of reduced self-awareness can include unrealistic expectations and goals, poor compliance with treatment, and a reduced ability to utilize therapy strategies (Richardson, McKay, & Ponsford, 2015). Poor self-awareness can also contribute to functional dependence, relationship problems, and emotional distress in family members (Richardson et al., 2015).

Disorders of self-awareness occur following damage to the frontal, temporal, and parietal cortices, are reported to be particularly likely following right-hemisphere damage (Sollberger et al., 2014; Stuss, Picton, & Alexander, 2001), and consequently have been related to a variety of cognitive deficits. In particular, right-lateralized brain regions including the precuneus, temporoparietal junction, posterior cingulate cortex, and superior temporal gyrus, have been implicated in deficits of self-awareness related to perspective-taking, theory of mind, and attention (Zamboni, Grafman, Krueger, Knutson, & Huey, 2010; Vago & Silbersweig, 2012). It follows that perspective-taking, the cognitive process of inferring other people’s mental states, beliefs, thoughts, and feelings, is relevant for self-awareness (Zamboni et al., 2010). Brain regions involved in perspective-taking overlap for those with adequate self-awareness (Northoff et al., 2006; Tusche, Böckler, Kanske, Trautwein, & Singer, 2016) and for adults with cognitive impairment (Zamboni, 2013).

Also relevant to self-awareness are anterior brain regions important for self-regulation. This aspect of self-awareness involves the ability to modulate one’s behavior and emotion processes (Vago & Silbersweig, 2012). Deficits of self-regulation are particularly prominent in the behavioral variant of frontotemporal dementia (bvFTD), a type of dementia that is characterized by behavioral disturbances and damage to anterior brain regions (Banks & Weintraub, 2008; Prigatano, 1991; Stuss, 1991). Several studies clarify the neural architecture of socioemotional regulation. Sollberger and colleagues (2014) found that predominately right hemispheric frontotemporal and insular regions were associated with one’s capacity for self-awareness and socioemotional functioning during an emotion-regulation task. Additionally, Beer and colleagues (2006) found that patients with damage to the orbitofrontal cortex required attentional cuing to identify faux pas they had committed in a prior social interaction task (Beer, John, Scabini, & Knight, 2006).

Although the right hemisphere is thought to play a special role in self-awareness, there is also compelling evidence for left hemispheric involvement through disruption of inner speech or our “internal dialogue” (Morin & Michaud, 2007). Simply put, inner speech may be a prerequisite for introspective processes inherent in completing self-report measures (Morin, 2005). Extending this further, Vygotsky’s theory of cognitive development suggests that the development of self-talk is related to aspects of cognition and behavior, such as memory, problem-solving, and self-regulation (Vygotsky, 1934/1987). Neuroimaging studies have also supported the role of speech and self-awareness with associations between language processing regions (i.e., inferior frontal and middle temporal gyri) and self-awareness (Chavoix & Insausti, 2017; Morin & Michaud, 2007). Given the multiple mechanisms and neural pathways postulated for self-awareness, investigation that contrasts cortical and cognitive mechanisms is called for.

There are two primary methods for measuring the deficits in self-awareness arising from these disparate cognitive and neural underpinnings: (1) a rating by a clinical observer and (2) a discrepancy score on a scale between the patient and another person (e.g., caregiver, clinician). The clinician-observed method requires an observer, such as a clinician, to rate the patient’s level of awareness (Sherer et al., 1998). The clinical observation of self-awareness assessment can be based on an unstructured interview, structured interview, and/or incorporation of patient records (Prigatano, 2010). An example of this type of assessment is the Neurobehavioral Rating Scale (NRS), a 28-item clinician-observed instrument that measures cognitive- and mood-based symptoms and includes an item for poor insight (Levin et al., 1987). The clinician-observed method of self-awareness is especially useful when the nature of a patient’s cognitive dysfunction may be interfering with the reliability of patient self-report. The discrepancy method, on the other hand, directly uses a patient’s self-report scores of a given neurobehavioral presentation, such as the dysexecutive syndrome, and contrasts it with a caregiver/significant other or clinician rating to provide a comparison with the patient report (i.e., self minus other) (Fleming, Strong, & Ashton, 1996; Zamboni et al., 2010). The discrepancy method is often used to evaluate self-awareness in patients with frontal systems injuries/disorders, including traumatic brain injury (Hart et al., 2009) and Alzheimer’s disease (Zamboni et al., 2013). A self- and other-report test that has been adapted for the discrepancy method is the Frontal System Behavior Scale (FrSBe)—a sensitive measure of behavioral change in people with frontal system dysfunction (Grace & Malloy, 2001). Although clinician-observed and discrepancy methods of measuring self-awareness tend to be strongly, positively correlated, assessment of patient self-awareness will vary (sometimes widely) (McGuire et al., 2014; Leicht, Berwig, & Gertz, 2010). Additionally, while both methods have been reported on, there has been a lack of head-to-head comparison in the literature. Given the disparate cognitive and neural architecture, and varied methods of assessment, further study is needed to better understand the results yielded by measures of self-awareness.

Hypotheses and Study Aims

This paper has three main goals: (1) comparing three neurocognitive frameworks for self-awareness, (2) evaluating the contribution of memory dysfunction to self-awareness and its assessment, and (3) comparing the psychometric approaches for assessing patient self-awareness. Our first hypothesis was that, of three frameworks (self-regulation, language, and perspective-taking), the self-regulation framework would best predict self-awareness, consistent with the association between the frontal lobes, executive function, and self-awareness. Our second hypothesis was that memory functioning would not significantly contribute to the self-awareness framework, consistent with the relative independence of self-knowledge and episodic memory networks. An additional aim of this paper was to explore the extent to which predictors (i.e., neuropsychological and brain variables) were associated with two different measures of assessing self-awareness (e.g., clinician-observed vs. discrepancy method).

Methods

Participants

Participants were derived from a cohort of 243 participants, who were referred by outside neurologists to the Cognitive Neuroscience Section of the National Institute of Neurological Disorders and Stroke (NINDS) from 2001 to 2009. Participants were required to have a clinical diagnosis of either frontotemporal dementia (FTD) or corticobasal syndrome (CBS) from a neurologist experienced with both disorders. Diagnoses were either confirmed, or the appropriate diagnosis was made, at NINDS by an experienced behavioral neurologist (EMW) and neuropsychologist (JHG) via a consensus based upon standardized neuroimaging, genetic testing, and core clinical neuropsychological and neurologic examinations.

Case definitions of subtypes of FTD included behavioral variant frontotemporal dementia (bvFTD) and primary progressive aphasia (PPA) based on Lund-Manchester Group consensus criteria (Neary et al., 1998). Although more recent criteria have since been established (Rascovsky et al., 2011) the previous criteria were more stringent, with relatively lower sensitivity (52%) and higher specificity (estimated 90%–100%). In bvFTD, lead symptoms were progressive dysexecutive and behavioral abnormalities with early loss of insight noted by caregivers. Participants with PPA had at least 2 years of prominent impairment of language or speech without other relevant symptoms, and although the dia1 gnoses were conferred prior to 2011 they were consistent with the more recent criterion (Gorno-Tempini et al., 2011). Case definition for CBS was progressive, predominantly lateralized ideomotor apraxia and/or non-DOPA-responsive extrapyramidal motor dysfunction (limb dystonia or rigidity) with cortical sensory loss (astereognosis, agraphesthesia) and characteristic patterns of atrophy on MRI (Boeve, 2004). While diagnostic criteria emphasize motor difficulties in CBS and language impairment in PPA, these are degenerative conditions and therefore patients with these disorders may show diminished abilities, such as insight, over time (Banks & Weintraub, 2009; O’Keeffe et al., 2007). Dementia severity for all participants was assessed by the Mattis Dementia Rating Scale-2 (DRS-2) (Jurica, Leitten, & Mattis, 2001).

Data Availability and Imaging Subsample

Participants were excluded if they could not tolerate scanning, if excessive motion degraded image quality, if gray and white matter segmentation was unsuccessful, they did not have self-report or informant data from the Frontal Systems Behavioral Scale (FrSBe), were diagnosed with another neurodegenerative disorder, or their behavior precluded neuropsychological testing. Out of the total 243 participants, our final sample included 67 participants (34 women) with diagnoses of bvFTD (n = 26), CBS (n = 29), and PPA (n = 12).

Demographics

Relevant demographics and clinical characteristics of the 67 participants are shown in Table 1. On average, participants’ self-ratings (standardized to T-scores) on the FrSBe were T = 66.9 (±18.2), indicating participants rated themselves slightly more than 1.5 SD above the mean degree of neurobehavioral dysfunction. Caregiver-reported ratings on the FrSBe were 80.5 (±25.9) indicating that, on average, participants’ frontal system impairment was 3 SD above the mean. The average score of cognitive impairment as measured by the DRS-2 (121.9±14.8) was just below the cutoff for dementia (>123) (Springate, Tremont, Papandonatos, & Ott, 2014). This would, on average, collectively suggest a substantial degree of dysexecutive syndrome and cognitive impairment in the sample. Independent samples t-tests and chi-square analyses confirmed demographics from this subsample did not significantly differ from the larger cohort by age, education, race, handedness, or disease duration.

Table 1.

Participant demographics and clinical characteristics

	Overall
Participant sample	67 (100%)
bvFTD	26 (38.8%)
CBS	29 (43.2%)
PPA	12 (17.9%)
Gender
Male	33 (49.3%)
Female	34 (50.7%)
Race
White	62 (92.5%)
Black	2 (3.0%)
Asian	1 (1.5%)
Pacific Islander	1 (1.5%)
Other	1 (1.5%)
Ethnicity
Non-Hispanic	67 (100%)
Handedness
Right-handed	59 (88.1%)
Left-handed	7 (10.4%)
Ambidextrous	1 (1.5%)
Age	62.2 (±8.3)
Education (years)	15.6 (±2.8)
Age of onset	57.5 (±8.9)
Years since disease onset	4.6 (±3.2)
Token Test (% correct)	89.12 (±13.6)
DRS-2 Initiation/Perseveration (raw)	28 (±6.7)
DRS-2 Construction (raw)	4.2 (±2.1)
DRS-2 Memory (raw)	21.2 (±3.7)
DRS-2 Total (raw)	121.9 (±14.8)
NRS-Inaccurate Appraisal	2.4 (±1.6)
FrSBe Caregiver (T-score)	80.5 (±25.9)
FrSBe Self (T-score)	66.9 (±18.2)

Standard Protocol Approvals, Registrations, and Participant Consents

Participants gave assent for the study. Study and consent procedures were approved by the NINDS Institutional Review Board.

MRI Protocol

Brain scans were obtained on three scanners of two field strengths over the course of the study; 42 on a General Electric (GE) 1.5 T with standard quadrature head coil, 24 on a Philips 3 T, and 1 on a Philips 1.5 T magnetic resonance scanner. The GE scanner (GE Medical Systems, Milwaukee, WI) T1-weighted spoiled gradient-echo sequence was used to generate 124 contiguous 1.5 mm-thick axial sections (repetition time = 6.1 ms; echo time=minimum full; flip angle=20°; field of view = 240 mm; matrix size = 256 × 256 × 124). Philips scan parameters were TE/TR = 3/6.5 ms, flip angle=8°, inplane resolution=256 × 256 voxels, 1 mm slice thickness, 140–180 slices, field of view = 240 mm.

Volumetric Analysis

Automated cortical reconstruction, volumetric segmentation, and visualization were performed with FreeSurfer software. Full documentation and download is available online (http://surfer.nmr.mgh.harvard.edu/). Following automated cortical reconstruction, all brains were visually inspected for quality control. Abnormalities in intensity normalization, resulting in poor white matter segmentation, were fixed by manual placement of control points (Gansler, Huey, Pan, Wassermann, & Grafman, 2016). Automated parcellation of the cerebral cortex into 68 total (right and left hemispheric) gyral- and sulcal-based regions of interest (ROIs) was performed (Desikan et al., 2006).

Predictors

Neuroimaging variables

To contrast different frameworks of self-awareness, several a priori regions of brain volume were selected based on literature suggesting their potential to contribute to one of the frameworks. Specifically, there were three different groupings of neural regions and neuropsychological scores to analyze self-awareness: (1) self-regulation, (2) language, and (3) perspective-taking. Self-regulation brain regions included the right and left: medial orbitofrontal, insular, and rostral anterior cingulate cortices (Goldin, McRae, Ramel, & Gross, 2008; Kühn, Gallinat, & Brass, 2011; Petrovic et al., 2016; Petrovic & Castellanos, 2016). Language areas were lateralized to the left hemisphere and included pars opercularis and triangularis, and middle and superior temporal gyri (Binder et al., 1997). Perspective-taking regions of interest included the right superior temporal gyrus, right posterior cingulate cortex, right precuneus, right inferior parietal lobule, and bilateral caudal anterior cingulate cortices (Northoff et al., 2006; Tusche et al., 2016; Zamboni et al., 2010). Each brain region volume was divided by the participant’s total intracranial volume to account for differences in head size and reduce inter-individual variation (Whitwell, Crum, Watt, & Fox, 2001).

Cognitive variables

A research neuropsychological battery was administered to every participant in this sample, and included tests of intelligence, executive function, memory, language, visuospatial processing, and motor speed. The tests were administered by research assistants trained in manual-based administration of each test. Test scores used for current analyses were selected based on the functional neuroanatomical relevance to our three frameworks of self-awareness. The Token Test (De Renzi & Vignolo, 1962) total correct percentage was included in the language framework due to its relevance for receptive language and comprehension of verbal commands of increasing complexity (Strauss, Sherman, & Spreen, 2006). The Mattis Dementia Rating Scale-2 (DRS-2) Total score was used as a global measure of cognitive functioning, while relevant subtests were entered into one of the three frameworks for self-awareness (Jurica, Leitten, & Mattis, 2001). The self-regulation framework incorporated the DRS-2 Initiation/Perseveration subtest, an executive function task that involves initiating and maintaining goal-directed behavior. The DRS-2 Construction subtest involves copying simple visual designs and was included in the perspective-taking framework given the involvement of right posterior brain regions for visual processing, attention, and reasoning (Lezak Howieson, Bigler, & Tranel, 2012). Memory was measured with the DRS-2 and included measures of orientation (to time, date, and situation) and verbal recall. The memory subtest score was added as a predictor in all three frameworks to test for the contribution of memory in self-awareness. Scores for the DRS-2 subscales and Token Test underwent the same transformative process: (1) raw scores were converted to z-scores, (2) DRS-2 total scores were regressed on the z-scores, and (3) the standardized statistical residual from the regression was used for subsequent analyses. The transformed scores each represented an index of the specific function (initiation/perseveration, language, memory, and construction) and relatively free of the contribution of general neuropsychological ability.

Assessment of self-awareness

Extent of patient self-awareness was measured by two approaches: the discrepancy (i.e., the difference between self and caregiver report) and clinician-observed method. The Frontal Systems Behavioral Scale (FrSBe) measured patient dysexecutive behavior. Self-report and caregiver-report FrSBe scores were obtained for all 67 participants. Caregiver FrSBe scores were based on patient status within the past month (4 weeks prior to assessment date) (Grace & Malloy, 2001). A final self-awareness score was obtained by subtracting the “patient total” self-report score from their “caregiver total” report score. Positive scores indicate that the caregivers identified more dysexecutive behavior than the patients, whereas negative scores indicate the inverse. The FrSBe age- and education-adjusted T-scores were used. The Neurobehavioral Rating Scale (NRS) is an example of a clinician-rated measure, and has 27 items that assess behavioral sequelae of neurologic insult (Levin et al., 1987). The NRS yielded the observation-based rating of patient self-awareness from the “Inaccurate insight and self-appraisal” item (NRS-Inaccurate Appraisal). This item’s description reads: “Poor insight, exaggerated self-opinion, overrates level of ability and underrates personality change in comparison with evaluation by clinicians and family” (Levin et al., 1987). The item was rated using a 7-point scale ranging from 1 “Not Present” to 7 “Extremely Severe,” meaning that higher scores indicated greater deficits of self-awareness. The psychometrist research assistant, who also administered the neuropsychological testing, completed the NRS after 1 week based on up to 30 hr of contact with the participant.

Pre-Analysis

Data were analyzed using SPSS version 21.0 (IBM Corp, 2012) and RStudio (R version 3.3.1) (RStudio, 2015). There were no missing data. Initial analyses in SPSS were conducted to ensure the suitability of variables for ANOVA and regression. Stem-and-leaf plots indicated three outlying scores from the FrSBe discrepancy method for self-awareness (>2.5 SD above the mean), which were treated with outlier replacement with a data point 2.5 SD above the mean. Following outlier treatment, FrSBe discrepancy scores (caregiver minus self-report) ranged from −37 to 71 (M = 13.07, SD = 27.22) and were normally distributed (skewness = 0.37 and kurtosis = −0.51). NRS-Inaccurate Appraisal scores ranged from 1 to 6 (M = 2.43, SD = 1.57) and were very mildly positively skewed (skewness = 0.75 and kurtosis = −0.49) but still within an acceptable range for normality. Plots of brain volume data indicated a total of eight outliers (>2.5 SD above the mean) in the selected regions of interest. These data points were modified to 2.5 SD above or below the mean (based on where their original value fell relative to the mean), rather than removing the outliers from the analysis in order to conserve power and adhere to the assumption of the general linear model. For coefficient comparisons, the brain volume variables were standardized using Fisher’s z-transformation. The discrepancy and clinician-rated methods of self-awareness showed a moderate, positive relationship, r = 0.54, p > .000.

Hypothesis Testing

For evaluating Hypothesis 1, penalized regression with best subset variable selection and 10-fold cross-validation method was used (Hastie, Tibshirani, & Wainwright, 2015). The best subset regression was chosen because it tolerates a lower predictor to participant ratio, penalizes non-robust predictors by setting their coefficients to zero, and increases the likelihood of replication (Hastie et al., 2015). For Hypothesis 1, best subset regressions were performed with either of the self-awareness scores as the criterion and the constituents of the three variable clusters as predictors for each framework (i.e., self-regulation, language, and perspective-taking). The adjusted R-square was used to compare frameworks, and the standardized coefficients were used to compare predictors. For Hypothesis 2, we included the DRS-2 Memory score in the same regression models as Hypothesis 1. Appreciable gain of variance predicted by the frameworks with the addition of the DRS-2 memory score was used to evaluate the contribution of memory.

Results

Hypothesis 1

Framework comparisons

Best subset regressions were performed across disparate sets of variables to evaluate the three frameworks of self-awareness. The strongest framework of each of the two self-awareness methods was self-regulation (adjusted R² = 0.45 and 0.17 for the NRS-Inaccurate Appraisal and FrSBe discrepancy rating, respectively) (Tables 2 and 3). The strongest predictors within the self-regulation frameworks were the left orbitofrontal cortex and right rostral anterior cingulate cortex. Notably, in three of six instances, neuropsychological variables contributed to the overall framework.

Table 2.

Best subset regression with FrSBe discrepancy score as criterion

Framework	Best subset	β	Adj R2
Self-regulation	Left Medial Orbitofrontal	−0.27	0.17
	Right Rostral Anterior Cingulate	−0.23
	Right Medial Orbitofrontal
	Left Rostral Anterior Cingulate
	Right Insula
	Left Insula
	DRS Initiation/Perseveration
Language	Left Middle Temporal Gyrus	−0.33	0.05
	Left Superior Temporal Gyrus	0.32
	Left Pars Opercularis
	Left Pars Triangularis
	Token Test
Perspective-taking	DRS Construction	0.22	0.04
	Right Precuneus	−0.19
	Right Posterior Cingulate	−0.19
	Right Inferior Parietal	−0.17
	Right Caudal Anterior Cingulate	−0.13
	Right Superior Temporal

Variables selected in the best subset are bolded.

Table 3.

Best subset regression with clinician-observed NRS-Inaccurate Appraisal as criterion

Framework	Best subset	β	Adj R2
Self-regulation	Right Rostral Anterior Cingulate	−0.41	0.45
	Left Medial Orbitofrontal	−0.3
	Right Insula	0.27
	DRS Initiation/Perseveration	0.21
	Right Medial Orbitofrontal	−0.19
	Left Insula
	Left Rostral Anterior Cingulate
Language	Left Pars Opercularis	−0.27	0.06
	Left Middle Temporal
	Left Superior Temporal
	Left Pars Triangularis
	Token Test
Perspective-taking	Right Caudal Anterior Cingulate	−0.27	0.13
	Right Posterior Cingulate	−0.21
	DRS Construction	0.19
	Right Superior Temporal	0.15
	Right Inferior Parietal	0.12
	Right Precuneus	0.03

Hypothesis 2

Contribution of memory

Memory was selected as a predictor in the best subset for each of the three frameworks with NRS-Inaccurate Appraisal as the criterion. There was a small gain in the strength of the relationship when memory was added to the language framework (Δadj R² = 0.09) and the perspective-taking framework (Δadj R² = 0.06) (Tables 3 and 5). As expected, decreases in memory functioning were associated with greater deficits in patient self-awareness. Although memory was selected as a predictor in the best subset for the self-regulation framework, there was no appreciable increase in adjusted R² (Tables 3 and 5). Across the three frameworks with the FrSBe discrepancy score as the criterion, only the perspective-taking framework selected memory as a significant predictor of self-awareness (Table 4).

Table 5.

Best subset regression with clinician-observed NRS-Inaccurate Appraisal as criterion, inclusive of DRS memory as predictor

Framework	Best subset	β	Adj R2	ΔAdj R2a
Self-regulation	Right Rostral Anterior Cingulate	−0.42	0.45	0.00
	Right Insula	0.27
	Left Medial Orbitofrontal	−0.26
	Right Medial Orbitofrontal	−0.2
	DRS Memory	−0.14
	DRS Initiation/Perseveration	0.14
	Left Insula
	Left Rostral Anterior Cingulate
Language	DRS Memory	−0.33	0.15	−0.09
	Left Pars Opercularis	−0.28
	Left Middle Temporal
	Left Superior Temporal
	Left Pars Triangularis
	Token Test
Perspective-taking	Right Caudal Anterior Cingulate	−0.34	0.19	0.06
	DRS Memory	−0.32
	Right Precuneus
	Right Inferior Parietal
	Right Posterior Cingulate
	Right Superior Temporal
	DRS Construction

^aDenotes change in adjusted R-square with the addition of DRS Memory to the model.

Table 4.

Best subset regression with FrSBe discrepancy score as criterion, inclusive of DRS memory as predictor

Framework	Best subset	β	Adj R2	ΔAdj R2a
Self-regulation	Left Medial Orbitofrontal	−0.27	0.17	0.00
	Right Rostral Anterior Cingulate	−0.23
	Right Medial Orbitofrontal
	Left Rostral Anterior Cingulate
	Right Insula
	Left Insula
	DRS Memory
	DRS Initiation/Perseveration
Language	Left Middle Temporal Gyrus	−0.33	0.05	0.00
	Left Superior Temporal Gyrus	0.32
	Left Pars Opercularis
	Left Pars Triangularis
	DRS Memory
	Token Test
Perspective-taking	Right Posterior Cingulate	−0.18	0.05	0.01
	Right Precuneus	0.18
	DRS Construction	−0.17
	DRS Memory	−0.16
	Right Inferior Parietal	−0.14
	Right Caudal Anterior Cingulate	−0.12
	Right Superior Temporal

^aDenotes change in adjusted r-square with the addition of DRS Memory to the model.

Even when memory was selected for the perspective-taking framework, there was no appreciable gain in variance explained (Δadj R² = 0.01) (Tables 2 and 4). Additionally, the standardized coefficient of memory (β = −0.16) indicated a small degree of association.

Comparison of modalities for measuring self-awareness

Across the frameworks, the clinician-observed method was more strongly associated with selected variables of self-awareness compared to the discrepancy method. The self-regulation framework accounted for 45% of the variance of the clinician-rated score compared to 17% of the variance with the discrepancy score as the criterion (Tables 2 and 3). The perspective-taking framework showed a similar pattern with a stronger association with the clinician-rated score (adj R² = 0.13) compared to the discrepancy score (adj R² = 0.04). Surprisingly, memory function contributed to the clinician-rated method of self-awareness, but not to the discrepancy assessment method, even though the discrepancy method was partly based on patient self-report.

Discussion

Increasing study of self-awareness (Suchy, 2016; Zimmermann, Mograbi, Hermes-Pereira, Fonseca, & Prigatano, 2017), points to the potential for a level of systematic study similar to that achieved in memory and language disorders (Butters, 1984; Geschwind, 1970; Goodglass & Kaplan, 1983). The three principal findings of the current study were (1) that a neurocognitive framework of self-regulation accounted for more of the variance than neurocognitive frameworks of perspective-taking and language towards an understanding of self-awareness, (2) that while both clinical observation and discrepancy methods of measuring self-awareness were validated, the clinical observation method was more broadly and robustly associated with the neuroimaging and neuropsychological predictors, and lastly (3) that memory function did not contribute appreciably to the assessed measurement of self-awareness.

Consistent with our first hypothesis neither the perspective-taking or language framework approached the strength of the self-regulation framework in explaining either modality of self-awareness assessment. The self-regulation framework contained ventral and anteromedial regions of interest and a measure of executive function. This finding was consistent with previous research showing that individuals who have impaired self-awareness also tend to show a pattern of deficits in the domain of executive function (Lezak, 1982). Research into self-awareness tends to involve disorders known for ventral and anterior dysfunction (Hart, Whyte, Kim, & Vaccaro, 2005; Suchy, 2016; Zamboni et al., 2010). In neurodegenerative disorders with other pathological foci, brain-behavior relationships could differ.

Within our self-regulation network, analyses consistently found robust negative associations between the left OFC, right rostral anterior cingulate cortex (ACC), and self-awareness scores, such that lower volume was related to poorer insight. These results may suggest a bilateral OFC–ACC network of self-awareness. The role of the left OFC is counterintuitive given the wealth of research implicating the right hemisphere in the loss of awareness (Blake, Duffy, Myers, & Tompkins, 2002). However, the lack of findings of self-awareness and the left hemisphere in the literature may be partially attributed to a study selection bias. For instance, Cocchini and colleagues (2009) suggested that findings of the right hemisphere and self-awareness may be inflated due to the exclusion of patients with left hemisphere brain damage (due to test administration difficulty). It may also be that each hemisphere makes unique contributions to a self-awareness network. For example, a functional MRI study of healthy individuals found that agency tasks activated the right hemisphere whereas thinking about the past and future, one’s emotions, and personality recruited the left side (Morin, 2011). The current findings of left OFC involvement in unrealistically positive self-perception are consistent with recent literature of OFC dysfunction in mild traumatic brain injury (Beer & Hughes, 2010). The specific role of the OFC is less understood, and could either be inhibitory or facilitative regarding the flow of social-cognitive information. For example, OFC atrophy might lead to an inability to properly attenuate false information, or atrophy to the OFC might weaken the suppression of negative information, allowing for mostly positive self-appraisals (Beer, 2007).

The ACC has a critical role in self-regulation/control and self-referential processes (Bush, Luu, & Posner, 2000; Tang & Tang, 2013). Lesion studies have linked damage to the rostral ACC with deficits in error monitoring (Maier, Di Gregorio, Muricchio, & Di Pellegrino, 2015). Disruption between or damage to the OFC and ACC may result in a diminished ability to detect inaccurate self-appraisal. In the context of decision-making, the medial OFC processes stimulus reward value (Rushworth, Noonan, Boorman, Walton, & Behrens, 2011). If valuation becomes inaccurate and there is not detection of conflict between performance and its value, a person could have impaired self-awareness.

Our second hypothesis was supported in that memory was not a robust predictor of self-awareness. The relationship between memory and self-awareness is complex and depends on the type of memory (e.g., episodic, semantic) and the components of the self being studied. Our finding that memory was not predictive of self-awareness was consistent with previous research showing a dissociation between episodic and semantic memory and self-schema (Klein, Cosmides, & Costabile, 2003). The results may indicate that at lower to moderate levels of memory dysfunction patient report is useful in assessing self-awareness. Our participant sample had diagnoses in which memory was relatively spared, and memory could have played a stronger role in our frameworks were amnestic participants included (e.g., Alzheimer’s disease). Although the orbitofrontal cortex (OFC) has been implicated in memory, this relationship is contextualized by the role of the OFC for representing emotional salience and its top-down control of the amygdala. While memory and self-awareness deficits may co-occur, they may not share a substrate.

Comparison of the moderately associated self-awareness assessment modalities showed that the clinician-observed method related more robustly to brain volume variables and executive dysfunction, that are in theory linked to self-awareness, than the FrSBe discrepancy method. Furthermore, the clinician-observed rating of self-awareness seemed to more strongly relate to neuropsychological test scores than the FrSBe discrepancy method. This may be because the discrepancy method is based on self and other ratings in relation to the participant’s pre-illness functional level, and therefore less associated with current neurobehavioral status. Given that most evaluations are multi-modal and use the method of multiple dissociations, the clinician-observed rating might be more useful. However, the discrepancy method offers relevant information in understanding patients in the context of their support network, which is influenced by extent of decline, distress, and resource availability.

Several study limitations are worthy of consideration when interpreting the data. The sample consisted of three patient groups of differing size with neurodegenerative disorders, and imbalance in group membership could be influencing outcome. There are differences in the number of items of the two assessment methods, with 43 FrSBe items constituting the discrepancy method, and a single item comprising the NRS clinician-observed method. Differences in item extent could have influenced outcome. Also, the NRS is usually collected on the basis of a 45-min interview, rather than the 30 hr of observation in this study, a difference that could affect generalization to clinical settings in which geriatric evaluations typically occur in a morning or afternoon. In order to preserve statistical power left-handed individuals were not excluded, and anomalous language dominance could have influenced results, particularly those related to the left lateralized language network.

Our findings add to an emergent literature of the neural correlates of self-awareness. These results indicate that brain regions important for self-regulation are associated with self-awareness in persons with neurologic disease involving varying regions of atrophy. Future studies of self-awareness in patient groups should pay attention to the possible association between self-regulation and level of awareness.

Funding

This work was supported by the National Institute of Neurological Disorders and Stroke intramural funding to Dr. Jordan Grafman. Dr. Grafman’s work on this manuscript was also partly supported by the Therapeutic Cognitive Neuroscience Fund and the Smart Family Foundation.

Conflict of interest

None declared.

Acknowledgements

The authors gratefully acknowledge the brain image analysis processing contributions of Ms. Jessica Jie Pan and Mr. Ryan Andrew Mace, and the statistical contributions of Mr. Rob Nyenhuis. We also thank the patients and their family members for participating in our studies.