Clinical symptoms in mild cognitive impairment with Lewy bodies: Frequency, time of onset, and discriminant ability

Abstract Background and purpose Mild cognitive impairment with Lewy bodies (MCI‐LB) is associated with a range of cognitive, motor, neuropsychiatric, sleep, autonomic, and visual symptoms. We investigated the cumulative frequency of symptoms in a longitudinal cohort of MCI‐LB compared with MCI due to Alzheimer disease (MCI‐AD) and analysed the ability of a previously described 10‐point symptom scale to differentiate MCI‐LB and MCI‐AD, in an independent cohort. Methods Participants with probable MCI‐LB (n = 70), MCI‐AD (n = 51), and controls (n = 34) had a detailed clinical assessment and annual follow‐up (mean duration = 1.7 years). The presence of a range of symptoms was ascertained using a modified version of the Lewy Body Disease Association Comprehensive LBD Symptom Checklist at baseline assessment and then annually. Results MCI‐LB participants experienced a greater mean number of symptoms (24.2, SD = 7.6) compared with MCI‐AD (11.3, SD = 7.4) and controls (4.2, SD = 3.1; p < 0.001 for all comparisons). A range of cognitive, parkinsonian, neuropsychiatric, sleep, and autonomic symptoms were significantly more common in MCI‐LB than MCI‐AD, although when present, the time of onset was similar between the two groups. A previously defined 10‐point symptom scale demonstrated very good discrimination between MCI‐LB and MCI‐AD (area under the receiver operating characteristic curve = 0.91, 95% confidence interval = 0.84–0.98), replicating our previous finding in a new cohort. Conclusions MCI‐LB is associated with the frequent presence of a particular profile of symptoms compared to MCI‐AD. Clinicians should look for evidence of these symptoms in MCI and be aware of the potential for treatment. The presence of these symptoms may help to discriminate MCI‐LB from MCI‐AD.

10-Symptom Discrimination" section of the results reports data only from participants who were not part of the LewyPro study, to replicate the discriminant ability of the 10-point symptom scale that was developed in the LewyPro cohort.
Potential participants were approached if they experienced symptoms that might be related to prodromal DLB, such as autonomic symptoms, visual disturbances, olfactory impairment, and mood changes as well as any indication of the presence of core and supportive features of DLB. Participants were excluded if they had a diagnosis of dementia, an Mini-Mental State Examination score < 20, a Clinical Dementia Rating score > 0.5, parkinsonism that developed >1 year prior to cognitive impairment, or evidence of clinical stroke or a serious neurological or medical condition that would affect their performance in study assessments. Participants with a current episode of major depression or a history of bipolar disorder or schizophrenia were also excluded. In the SUPErB study, participants with symptomatic heart failure (New York Heart Association Class II or greater) were excluded to avoid false-positive cardiac metaiodobenzylguanidine (MIBG) results [12].
All participants gave their written informed consent to take part in the study. The study received ethical approval from the National

Assessment
Participants had a comprehensive clinical assessment as detailed previously [13,14].
During their clinical assessment, participants (with a relative, friend, or carer present where possible) were asked whether they experienced a range of symptoms, adapted from a previous version of the Lewy Body Disease Association Comprehensive LBD Symptoms Checklist (Table S1). An updated version of this is freely available online at https://lbda.org/wp-conte nt/uploa ds/2020/09/ compr ehens ive_lbd_sympt om_check list_2019.pdf. The questionnaire was administered at baseline and follow-up assessments. The mean duration of follow-up was 1.7 years (SD = 1.4).
The interviewer asked whether each symptom was present.
When participants reported a symptom, they were asked how long it had been present.
Ten symptoms were identified in our previous paper as being common in DLB (>50%) and uncommon in AD (<20%) and were found to accurately discriminate between MCI-LB and MCI-AD in our previous report [11]. The number of these 10 symptoms that were present was recorded for each participant.

Diagnosis
An expert consensus clinical panel (A.J.T., P.C.D., J.-P.T.) reviewed all the clinical assessment data to confirm subjects met NIA-AA all-cause MCI criteria [15] without considering aetiology. Where the first two raters did not agree, the third made a final decision. The consensus panel also rated the presence or absence of each of the four core symptoms of DLB (cognitive fluctuations, complex visual hallucinations, motor parkinsonism, and clinical rapid eye movement [REM] sleep behaviour disorder [RBD]). These symptoms were evaluated with specific scales during the clinical assessment (the Dementia Cognitive Fluctuations Scale [16], the Clinician Assessment of Fluctuation Scale [17], the North East Visual Hallucinations Interview [18], the Unified Parkinson's Disease Rating Scale motor scale [19], and the Mayo Sleep Questionnaire [20]). Parkinsonism was defined as the presence of bradykinesia, a parkinsonian rest tremor, or rigidity, as set out in the MCI-LB criteria [3]. The symptom ratings were combined with imaging biomarker results where available The "1-year rule" was applied so that no subjects had had evidence of parkinsonism for >1 year before the onset of their cognitive decline. Cerebrospinal fluid (CSF) and imaging biomarkers were not used in the diagnosis of MCI-AD; therefore, the MCI-AD cases fulfilled the NIA-AA "Core Clinical Criteria" for MCI-AD.
Assignment to these diagnostic categories was based on information from both baseline and follow-up clinical evaluations where available. Participants were included if their final diagnosis was probable MCI-LB, probable DLB [2], MCI-AD, or AD [21]. Participants with a final diagnosis of probable MCI-LB or probable DLB will be referred to as "MCI-LB" in this article. Participants with a final diagnosis of MCI-AD or AD will be referred to as "MCI-AD."

Statistics
Statistical comparisons were performed using SPSS and SAS software. MCI-AD and MCI-LB groups were compared using chisquared test, Fisher exact test, Mann-Whitney U-test, and t-tests where appropriate. As there were more males in the MCI-LB group than the MCI-AD group, and longer time of follow-up, comparison of cumulative frequency of symptoms was tested using logistic regression with sex and years of follow-up as covariates. As 43 symptoms were tested, correction for multiple comparisons was carried out using the method of Benjamini and Yekutieli [22]. In the cross-sectional replication comparison, area under the receiver operating characteristic curve (AUROC) was plotted to determine the ability of the 10-point symptom score to discriminate between MCI-LB and MCI-AD [11].

Longitudinal cumulative prevalence
The group demographics and cumulative prevalence of each symptom are displayed in Table 1  Although many symptoms were more common in MCI-LB than MCI-AD, when present, the time of onset of symptoms was similar in the two groups (Table S2).

Cross-sectional replication of 10-symptom discrimination
The demographics of the cohort and the prevalence of each symptom in the diagnostic groups are displayed in Table S3.
Ten symptoms were identified in our previous publication [11] as being relatively sensitive (>50%) and specific (>80%) to DLB in comparison to AD (these are listed in Table 2 Figure S2). Good discrimination between MCI-LB and MCI-AD was demonstrated for threshold scores of >1 (sensitivity = 83%, specificity = 83%), >2 (sensitivity = 76%, specificity = 90%), and >3 (sensitivity = 59%, specificity = 97%).  There were no significant differences between MCI-LB and MCI-AD in timing of symptom onset. It is important to note that many noncognitive symptoms in MCI-LB have a time of onset soon after the development of first cognitive symptoms. Therefore, the initial memory assessment presents an excellent opportunity to screen for these symptoms.

Clinical relevance: Discriminant ability
Core clinical features are less common in MCI-LB than DLB, and diagnostic biomarkers also have lower sensitivity at the MCI stage [11,12,24]. Consideration of supportive clinical features has the potential to aid clinical diagnosis.
As expected, symptoms associated with the core clinical features of MCI-LB were more common in MCI-LB than MCI-AD. The following features were also significantly more common in MCI-LB: difficulties with planning and problem-solving, drooling, hearing things not present, vivid dreams, nightmares, excessive sleepiness, sensitivity to heat/cold, and misjudging objects. These symptoms could be considered during clinical assessment in addition to core clinical features, to help improve the detection of MCI-LB.
Our findings are in keeping with previous reports, which have identified high rates of neuropsychiatric, motor, sleep, and autonomic symptoms in MCI-LB or prodromal DLB [6][7][8][9][10]. The control group in our study had higher rates of some symptoms in comparison to other cohorts, such as constipation [6] and drooling/hypersalivation [9]. We recruited MCI participants with possible symptoms related to prodromal DLB; therefore, the rates of the symptoms reported in MCI-AD here may be higher than in the general clinical population. Some symptoms have high variation in prevalence globally, for example, constipation [25]. This highlights the importance of understanding the rates of symptoms within local populations when applying clinical scales.
Importantly, constipation and loss of smell were relatively common in MCI-AD (>30%) and should not be considered specific to MCI-LB. A significant difference between MCI-LB and MCI-AD in constipation has been reported in another cohort. However, constipation was still relatively common in MCI-AD (62% vs. 21%) [6]. This study also found obstipation (severe constipation) to be more common in MCI-LB than MCI-AD (43% vs. 15%). Differences have been noted between MCI-LB and MCI-AD in direct testing of olfactory function, which may be more effective than patient/carer report of hyposmia [26,27].

Research relevance: Symptom burden
The scale of the symptom burden experienced in MCI-LB is highly

Research relevance: Discriminant ability
We replicated a previous finding that a brief list of 10 symptoms can accurately differentiate between MCI-AD and MCI-LB. We do not currently recommend the use of such a questionnaire as a diagnostic tool in the clinic, where clinical expertise and application of current criteria are a more appropriate approach. However, in research settings, it may be valuable to enrich cohorts with MCI-LB cases using a simple questionnaire that has the potential to be used remotely, by nonexpert diagnosticians and in a large number of potential participants. The appropriate threshold would depend on the sensitivity and specificity required for the particular research application.

Strengths and limitations
This article presents a clinically well-characterized MCI cohort with biomarker support for MCI-LB diagnosis from the LewyPro [11] and SUPErB [14] studies. This paper reports baseline cross-sectional symptom frequency data from the SUPErB study and combined longitudinal cumulative frequency from both the SUPErB and LewyPro studies. The LewyPro study cross-sectional data have previously been reported [11], and in the current paper we have replicated the discriminant ability of a 10-point symptom scale developed using data from that cohort. The cross-sectional data reported now only include participants who were not part of the LewyPro study; therefore, there is no overlap in the cross-sectional data presented here and that of our previous paper. The cognitive profile of the cohorts has been published previously [13,14,28].  [12,24].
Control of the false discovery rate (FDR) was carried out using the method of Benjamini and Yekutieli [22]. This method is more conservative than other methods of FDR correction (e.g., Benjamini and Hochberg [31]). Importantly, this method remains valid when there is dependency in p-values (as would be expected in this analysis). The MCI-LB and MCI-AD groups differed in proportion of males and duration of follow-up, but these variables were included as covariates in the statistical analysis. Despite the substantial sample size, some symptoms appeared approach statistical significance, and it is likely they would be significant in a larger sample. Nevertheless, the sample size is likely sufficient to detect differences that will be clinically relevant.
There is a possibility that the presence of core features such as visual hallucinations may have made interviewers more likely to endorse other symptoms thought to be related to DLB. However, interviewers were instructed to accept the participant's answer without interpretation, and the relatively low prevalence of symptoms previously thought to be strongly associated with Lewy body disease (e.g., constipation) suggests that this was successful. Diagnostic raters were not blind to the symptom questionnaire; however, more detailed, symptom-specific scales were used to identify the presence or absence of core clinical features.

CON CLUS IONS
MCI-LB is associated with a high prevalence of a range of cognitive, motor, neuropsychiatric, sleep, autonomic, and visual symptoms. A range of symptoms in MCI-LB have their onset soon after the first cognitive symptoms. Therefore, the first presentation to a memory assessment service is a good opportunity to identify these symptoms, explain their cause, and where appropriate, offer treatment or onward referral.
Clinicians should be aware of the prevalence of these symptoms, the potential for treatment, and their discriminant ability to identify