Structural MRI Correlates of Anosognosia in Huntington’s Disease

INTRODUCTION

Huntington’s disease (HD) is a neurodegenerative disease caused by a CAG trinucleotide repeat expansion in the huntingtin (Htt) gene. 1 Anosognosia, or unawareness of one’s symptoms, is a common symptom of HD. 2 Persons with HD can have anosognosia for symptoms across motor, behavioral, and cognitive domains. It is difficult to measure and often requires clinicians to compare patient history and exam with caregiver reports. Informant-based measurements, such as the Frontal Systems Behavior Scale (FrSBe), compare assessment results from patients and their informants (typically a caregiver).

Prior work has generally found that HD anosognosia correlates with the severity of other symptoms attributable to HD.3–9 Anosognosia can delay initiation of therapy and is associated with a higher caregiver burden, higher CAG burden, 10 and poorer executive function in HD. 11 Few studies have assessed HD-specific neuroimaging correlates of anosognosia. One prior report found that in the PREDICT-HD study, pre-symptomatic patients reported more executive dysfunction, apathy, and disinhibition than informants early on, but that this difference reversed later in the study, suggesting anosognosia develops slowly. 12 In the current study, we leveraged a large dataset of aggregated structural MRI and clinical assessments from the PREDICT-HD and TRACK-HD cohort studies to uncover clinical-imaging correlates of HD anosognosia.

METHODS

RESULTS

Characteristics of the TRACK-HD and PREDICT-HD participants who were included in the present analysis are described in Table 1, including the total FrSBe score and subscales. Companion-participant score difference (‘Anosognosia Index”) means and standard deviations for the total FrSBe score (–3.5±23.7) and subscales of executive dysfunction (–0.9±11.6), apathy (–0.5±7.8), and disinhibition (–2.1±7.2) indicated significant variation that was approximately normal in distribution (Supplementary Figure 1).

To test whether we could identify structural MRI correlates of anosognosia, we performed a correlation analysis between anosognosia scores and 44 brain ROI volumes. As presented in Table 2, this analysis revealed significant inverse relationships between anosognosia and brain volume in key basal ganglia regions affected in HD (globus pallidus, putamen, caudate). Additionally, anosognosia correlated with reduced volume in basal forebrain, substantia nigra, and angular gyrus, as well as relatively preserved volume (the relative “increase” indicating absence of volume loss) in cingulate cortex (Table 1). A full list of correlation results is presented in Supplementary Table 1. We also estimated the relative changes in volume present for patients with a relatively high degree (top 25th percentile) of anosognosia (Supplementary Table 2). Consistent with prior studies, anosognosia correlated with motor impairment, though with a correlation coefficient of ρ= 0.17 (p < 0.0001).

To identify brain regions that may be most useful collectively as volumetric markers of anosognosia, we performed linear regression with stepwise ROI variable addition. This procedure generated a list of 6 regions (Table 3) that contributed to a final model of anosognosia with the lowest loss of information, as determined by the AIC criterion. The final model has an adjusted R² of 7.4%, which is notably higher than a model that includes only CAP score (2.7%).

We also considered that depression may influence self-reported symptom severity, i.e., that more depressed patients could have more negative self-assessments and reduced score differences with participants. To test this possibility, we compared companion-participant FrSBe score differences among patients with or without depression, as determined using validated cutoffs for the HADS and BDI-II scales used in TRACK-HD and PREDICT-HD respectively. In the subset of participants (n = 342) with depression scores collected, we found that 73 (21%) met cutoffs for depression. As depicted in Supplementary Figure 2, depressed participants were more likely to rate their impairment more highly relative to companions than were non-depressed participants (mean companion-participant difference = –9.56, t = 3.09, df = 340, p = 0.002). To determine whether participants reporting much worse symptoms than their companions could be biasing results, we repeated the correlation analysis summarized in Table 2, excluding participants in the bottom quartile of companion-participant score discordance (i.e., symptom “over-estimators”). The result (Supplementary Table 3) demonstrates that, generally, exclusion of these participants did not reduce the strength of correlations; in most cases, correlation strength increased.

DISCUSSION

In this study we identified several volumetric MRI correlates of HD anosognosia using two well-characterized datasets, PREDICT-HD and TRACK-HD. Specifically, volumes of globus pallidus (GP), putamen, caudate, basal forebrain, substantia nigra, angular gyrus, and cingulate cortex significantly correlated with anosognosia. Further, we provide evidence through multivariable modeling that the variance in anosognosia severity can be explained better by structural MRI data than CAP score or motor symptomatology alone.

Clinically, anosognosia poses a significant challenge to patients with HD and their family. Differences in understanding about symptom severity can create dangerous scenarios; for example, a patient may need to stop engaging in previously normal tasks, such as driving, that are no longer safe for them. These barriers to mutual understanding can also compound with co-existing neuropsychiatric challenges, such as irritability and depression. Thus, a better understanding of anosognosia in HD is not only of scientific interest, but urgent clinical need. We would suggest, based on our data, that it would be helpful for families to understand that anosognosia is not simply a “denial” of one’s symptoms, but rather an intrinsic aspect of HD caused by neurodegeneration.

HD neuropathology studies have generally focused more on the striatum than GP. 19 Our univariate correlation data suggest that GP degeneration is at least as important to HD anosognosia as striatum. In the context of multivariate regression, GP volume was included first by the stepwise regression algorithm, instead of other basal ganglia structures, suggesting that the GP captures the most overall variance in anosognosia severity. Various other regions that did not correlate very strongly with anosognosia at a univariate level, such as the superior occipital gyrus, were selected by the stepwise algorithm as well. This may indicate that these regions contribute some additional explanatory value for the anosognosia measure that is derived from extrastriatal degeneration and impacts on circuit-level function.

Existing knowledge of anosognosia mechanisms largely derives from lesion studies (i.e., parietal cortex), investigation of the neurological basis of consciousness, and research on Alzheimer disease.20,21, 20,21 Of particular current interest is the default mode network (DMN), which may hold relevance for anosognosia within and beyond the context of HD. DMN function is associated with self-referential thinking and internally generated thought, i.e., cognitive content that appears to arise independently of external stimuli (e.g., “mind-wandering”). Interestingly, structures identified by our study as associated with anosognosia, namely the angular gyrus and cingulate cortex, are key nodes of the DMN. 22 It has been suggested that angular gyrus, posterior cingulate cortex, and medial prefrontal cortex are the three core regions that link subcircuits of the DMN and are most regularly engaged as the self-referential components of the DMN. 22 The GP and putamen have also been proposed to be key subcortical sites enabling DMN function. 23 However, our results can only draw attention to relative differences in volume in these regions among HD participants with anosognosia and cannot be interpreted as mechanistic evidence for the presence of dysfunctional DMN nodes.

We acknowledge some limitations of our analysis. Volumetric estimates of small regions that were significant in our results, such as basal forebrain, are less reliable than those of larger brain areas. 24 Similarly, cingulate cortex volume estimates are vulnerable to segmentation errors that over-estimate volume due to its complex mid-sagittal anatomy. 25 The participant-companion discrepancy method of anosognosia measurement employed herein may entail over- or under-reporting. For example, the companion is often a caregiver, a role that can be felt as emotionally straining and stressful. Their reporting may be impacted by their own perceptions about how well-controlled the HD symptoms are, even if the person with HD has an objectively accurate notion of their symptomatology. However, caregiver scores are generally more closely aligned with objective ratings than patient-reported measures, suggesting this discrepancy is likely to be generally reliable. 9 Future work may employ targeted scales of anosognosia that are designed to minimize error related to reporting biases and control for the impact of co-existing affective disorders or non-specific attentional impairment. Such measures, particularly by incorporating assessment of cognition, could operationalize anosognosia more specifically than a linear measure of caregiver-participant incongruence as employed here. It is also important to note that the GP is a prominent structure associated with anosognosia due to its high degree of atrophy correlation with the striatum in HD. 14 The multivariate analyses therefore identified brain regions that make best model when combined with other covariates.

In conclusion, the present work identifies key structural MRI correlates of anosognosia in HD. In particular, our finding that the GP is a key volumetric correlate of anosognosia may motivate further investigation of this region specifically. We anticipate that future studies employing other imaging modalities (e.g., functional MRI) and more specific measures of anosognosia may provide further information about the circuit-level dysfunction related to HD anosognosia.