Assessment of autonomic symptoms may assist with early identification of mild cognitive impairment with Lewy bodies

Abstract Objectives Autonomic symptoms are a common feature of the synucleinopathies, and may be a distinguishing feature of prodromal Lewy body disease. We aimed to assess whether the cognitive prodrome of dementia with Lewy bodies, mild cognitive impairment (MCI) with Lewy bodies (MCI‐LB), would have more severe reported autonomic symptoms than cognitively healthy older adults, with MCI due to Alzheimer's disease (MCI‐AD) also included for comparison. We also aimed to assess the utility of an autonomic symptom scale in differentiating MCI‐LB from MCI‐AD. Methods Ninety‐three individuals with MCI and 33 healthy controls were assessed with the Composite Autonomic Symptom Score 31‐item scale (COMPASS). Mild cognitive impairment patients also underwent detailed clinical assessment and differential classification of MCI‐AD or MCI‐LB according to current consensus criteria. Differences in overall COMPASS score and individual symptom sub‐scales were assessed, controlling for age. Results Age‐adjusted severity of overall autonomic symptomatology was greater in MCI‐LB (Ratio = 2.01, 95% CI: 1.37–2.96), with higher orthostatic intolerance and urinary symptom severity than controls, and greater risk of gastrointestinal and secretomotor symptoms. MCI‐AD did not have significantly higher autonomic symptom severity than controls overall. A cut‐off of 4/5 on the COMPASS was sensitive to MCI‐LB (92%) but not specific to this (42% specificity vs. MCI‐AD and 52% vs. healthy controls). Conclusions Mild cognitive impairment with Lewy bodies had greater autonomic symptom severity than normal ageing and MCI‐AD, but such autonomic symptoms are not a specific finding. The COMPASS‐31 may therefore have value as a sensitive screening test for early‐stage Lewy body disease.


| INTRODUCTION
The synucleinopathies, including Parkinson's disease (PD), dementia with Lewy bodies (DLB), and multiple system atrophy, are associated with pronounced dysfunction of the autonomic nervous system. 1 Autonomic symptoms are therefore common in DLB and Parkinson's disease dementia (PDD). 2 These may manifest in DLB prior to onset of dementia 3 and so could be an early marker of a "bottom-up" staged prodromal synucleinopathy. 4 However, identifying autonomic symptoms is complicated by the range of causes of these in older adults, in particular these symptoms may occur in other forms of dementia such as Alzheimer's disease (AD), 5 and so are not specific to DLB. 6,7 Recent efforts have led to the characterisation of the cognitive prodrome of DLB -mild cognitive impairment (MCI) with Lewy bodies (MCI-LB) -in research settings, with autonomic symptoms included as supportive clinical features. 8 While previous research has suggested that a retrospective history of autonomic dysfunction is common in DLB, 9 that individuals with pure autonomic failure are at high risk of later conversion to DLB, 10 MCI-LB may feature greater heart rate variability than in AD, 11 and that MCI-LB may be characterised by abnormal cardiac metaiodobenzylguanidine (MIBG) imaging of the sympathetic nervous system, 12 it is yet unclear whether MCI-LB features greater severity of autonomic symptoms when prospectively assessed, and whether these may differentiate MCI-LB from normal ageing or AD.
We therefore aimed to assess whether autonomic symptoms would be more severe in MCI-LB than in normal ageing, and whether these might help to distinguish MCI-LB from MCI due to Alzheimer's disease (MCI-AD). We hypothesised that probable MCI-LB would have more common and severe autonomic symptoms, assessed using the COMPASS (the Composite Autonomic Symptom Score scale), than cognitively healthy controls, and that the presence of autonomic symptoms would help differentiate MCI-LB from MCI-AD. We also hypothesised that cardiac MIBG abnormalities would be associated with greater autonomic symptom severity.

| Participants
Mild cognitive impairment patients and cognitively healthy controls were recruited into a longitudinal study, as described previously. 13

| Mild cognitive impairment patients
Patients were recruited from local older persons' memory, psychiatric and neurology services in North East England given the presence of a health service diagnosis of MCI, and either the reported presence of any core clinical feature of DLB (complex visual hallucinations, rapid eye movement sleep behaviour disorder (RBD), cognitive fluctuations, or parkinsonism not preceding cognitive impairment by more than 12 months), or any supportive clinical feature of DLB (e.g. mood change or sleep disturbance). Exclusion criteria were the presence of dementia, no objective cognitive impairment, or possible clinical stroke, vascular cognitive impairment or frontotemporal aetiology.

| Healthy controls
Cognitively healthy participants were recruited through the Join Dementia Research platform, and from friends or families of the patient group. They were required to be cognitively healthy with no MCI, dementia, or other suspected brain pathology, and normal structural MRI.

| Design
All participants underwent detailed medical review and cognitive assessment, 13 and imaging 12,14,15

| Imaging
Dopaminergic 123I-N-fluoropropyl-2β-carbomethoxy-3β-(4-iodophenyl) single-photon emission computed tomography (FP-CIT) and cardiac MIBG imaging were carried out, as previously described. 12,14 FP-CIT images were visually rated as normal or abnormal by a fiveperson panel of experienced image analysts, blind to clinical information. 14 MIBG images were classified as abnormal given a heart: mediastinum uptake ratio (HMR) of <1.86 based on data from locally-recruited healthy controls. 17 FP-CIT and MIBG imaging results were then incorporated into differential classifications as reported below. These diagnoses were updated at each annual follow-up by the consensus panel. Following a clinical diagnosis of dementia and dementia subtype, participants did not receive any further follow-up.

| Analysis
All analyses were conducted in R software. Group differences in the presence and severity of total and specific autonomic symptoms were assessed with zero-adjusted gamma models to address two key issues likely to be introduced by the COMPASS scoring system: (1) the likely abundance of zero scores when symptoms are absent and (2) scores being additive and positive when symptoms are present leading to right-skewness. This procedure allowed us to account for these two separate processes, symptom presence and symptom severity, in a single model by estimating two parameters: odds of non-zero scores (nu parameter with logit link) and relative severity of non-zero scores (mu parameter with log link). All models included age as a mean-centred, scaled continuous covariate, and were estimated with the gamlss package. Simple associations were assessed with rank based Spearman's correlation tests.
As a small group with uncertain diagnosis, possible MCI-LB were not considered in the primary hypothesis, but statistics are reported for additional context.
Receiver operating characteristic curves were plotted to assess the discriminatory utility of COMPASS total in identifying probable MCI-LB. Diagnostic cut-offs for discriminating probable MCI-LB from MCI-AD and healthy controls were identified by Youden's index.

| Data availability
Data supporting this analysis are available upon request through the Dementias platform UK (study reference: 'SUPErB').

| Group differences
Clinical characteristics have been described previously in depth and are summarised in Table 1 Of this cohort, 126 completed the COMPASS at baseline and seven did not; for this group, autonomic symptoms reported as present over the previous year are described in Table 2  The presence, severity, frequency, and any worsening of these symptoms contributed to symptom group sub-scores, and a total autonomic symptom score. Medians and ranges for these scores are summarised in Table 3, alongside analysis of the mu component of the zero-adjusted model: the relative severity of each symptom when present, and of overall autonomic symptom burden (compared to healthy controls in all cases). These indicated that overall age-adjusted autonomic symptom burden was approximately twice as severe in MCI-LB as in cognitively health comparators; an effect not found to be significant in MCI-AD.
While orthostatic intolerance was not significantly more common in MCI-LB or MCI-AD than in healthy controls (see Table 2), in both groups symptoms were rated as more severe than in controls. In addition to bladder symptoms being more common in MCI-LB, they were also rated as significantly more severe when present.
Across both MCI groups, total COMPASS score was not signifi- When controlling for cholinesterase inhibitor use, which was less common in MCI-AD (see Table 1 T A B L E 2 Autonomic symptoms reported by those who completed the COMPASS-31 questionnaire, and odds/age-adjusted odds ratios of these being present in comparison to cognitively healthy controls (inverted nu parameter from zero-adjusted gamma model, exponentiated)

| Secondary analysis: Metaiodobenzylguanidine imaging and autonomic dysfunction
An additional analysis was undertaken including both MCI-LB groups only, to assess if MCI-LB with abnormal MIBG imaging had more severe autonomic symptoms than those with normal MIBG, given that reduced cardiac uptake of MIBG may reflect a 'body first' rather than 'brain first' synucleinopathy, 20 correlating with plasma alphasynuclein, 21 and may therefore be associated with synuclein accumulation in the peripheral nervous system with resulting dysautonomia. There was no significant difference in overall COMPASS severity

| Discriminating mild cognitive impairment with Lewy bodies from MCI-AD and controls
Receiver operating characteristic analyses found that total COM-PASS score was a good discriminator of MCI-LB from healthy HAMILTON ET AL.
-5 controls, though less so in distinguishing probable MCI-LB from MCI-AD (see Figure 1). Youden's index identified the best cut-off as >4.5 for the latter; this provided a high sensitivity of 92%, though specificity was low at 42%. In discriminating probable MCI-LB from healthy controls, a cut-off of >9.5 was identified, providing sensitivity of 68% and specificity of 79%; the previously identified cut-off of >4.5 discriminated MCI-LB from controls with low specificity of 52%.

| Exploratory analysis with reduced COMPASS
An exploratory post hoc analysis was undertaken using a reduced COMPASS including orthostatic intolerance, bladder, and secretomotor symptom severity scores, based on broad group differences identified in the primary analysis. Receiver operating characteristic curves were plotted, and cut-offs chosen which maximised Youden's index. In discriminating probable MCI-LB from both MCI-AD and healthy controls a >4.5 cut-off was identified; this was much less sensitive to MCI-LB than the full COMPASS-31 (55% sensitivity), though specificity was substantially better, differentiating MCI-LB from MCI-AD with 78% specificity, and from controls with 94% specificity.

| DISCUSSION
We hypothesised that, consistent with other clinical syndromes due to synucleinopathies, probable MCI-LB would have greater autonomic symptom severity than cognitively healthy controls. The results supported this, identifying higher overall dysautonomia in probable MCI-LB, but not in MCI-AD who were intermediate to controls and MCI-LB. MCI-AD were more likely to experience T A B L E 3 Autonomic symptom severity rated by the COMPASS-31, and ratio coefficients of symptom severity when present (mu parameter from zero-adjusted gamma model, exponentiated), in comparison to cognitively healthy controls

Median [Range] of symptom scores Control (N = 33) MCI-AD (N = 36) Possible MCI-LB (N = 19) Probable MCI-LB (N = 38)
Composite Autonomic Symptom Score - gastrointestinal symptoms than controls, as were MCI-LB, who were also more likely to experience secretomotor and bladder symptoms.
When present, symptoms of orthostatic intolerance were more severe in MCI-AD than in controls; this was also the case in MCI-LB, in addition to more severe bladder symptoms. Composite autonomic symptom scores were highly sensitive to MCI-LB, but specificity was low, consistent with relatively high rates of autonomic symptoms in other neurodegenerative aetiologies, as well as in cognitively healthy older adults.
The presence and severity of autonomic symptoms in probable MCI-LB is consistent with evidence from other early clinical presentations of the synucleinopathies. 1 These findings provide support for the inclusion of autonomic symptoms as a supportive clinical feature of MCI-LB as in DLB 8 ; autonomic symptoms appear to be common in the cognitive prodrome of DLB.
In a sensitivity analysis, we found some observational evidence This study benefits from the inclusion of a thoroughly-assessed, prospectively-identified clinical cohort with comprehensive imaging and longitudinal medical review. However, due to the screening process seeking to identify potential Lewy body disease then autonomic symptoms could be more common in this MCI-AD group than would be found in the wider population, which would mean that specificity might be higher than we found in this study. Both MCI-AD and MCI-LB groups are limited by the absence of specific AD biomarkers which might verify the presence of AD in the former, or identify mixed aetiologies in the latter.
These findings are limited by reliance upon self-report of autonomic symptom presence and severity, which may be more sensitive to some symptoms than others; quantified measures of autonomic symptoms such as objective measures of orthostatic hypotension 28 may provide clearer information on some symptoms into which patients may not have insight at the early stages of onset, as well as a method of quantifying any change over time. Relatedly, the COM-PASS is likely not to have uniform sensitivity/specificity for each subscale, as some symptoms may be more or less common in the general older population. This is evidenced by the variable presence and severities across different symptom scales in controls. Differences in autonomic symptom expression within MCI-LB may therefore reflect differences in sub-scale specificity, rather than any fundamental differences in the pathophysiology of different symptoms.
Cardiac scintigraphy with MIBG is a biomarker for DLB 29 and in this cohort we recently reported that MIBG also had diagnostic accuracy in distinguishing MCI-LB from MCI-AD, 12 providing further evidence that autonomic nervous system involvement is present at the MCI stage. However, we did not find any clear association between cardiac denervation assessed with MIBG imaging and autonomic symptoms within MCI-LB. The number of subjects for this exploratory analysis was modest and any link between objective denervation and subjectively-reported symptoms may be unclear at this early stage, particularly when autonomic symptoms may be mild and heterogeneous. Future research should incorporate longitudinal data to assess whether cardiac denervation anticipates later onset of subjectively-reported or objectively-measured dysautonomia, as discussed previously.
In conclusion, autonomic symptoms are more common and severe overall in MCI-LB. High severity of autonomic symptoms is sensitive to MCI-LB, but not specific to this condition.