Abstract
INTRODUCTION
Dementia with Lewy bodies (DLB) accounts for 10–25% of dementia cases in clinical populations [1] and is characterized by the presence of cognitive, psychiatric, and motor symptoms. Early and accurate diagnosis of DLB is important since DLB patients have a more severe disease course with more rapid cognitive decline [2], earlier nursing home placement [3], higher costs [4], and different treatment response compared to AD [5]. However, clinical diagnosis of DLB can be difficult because of the variability and the overlap of pathological and clinical features between DLB and other related dementias, notably Alzheimer’s disease (AD). In a large cohort of patients (n = 2,861), sensitivity of the clinical diagnosis of DLB against autopsy was 32%, with a specificity of 98% [6]. Thus, while the clinical criteria rarely lead to false positive DLB diagnoses, clinicians need better tools to identify patients with DLB and avoid false negatives.
One possible such tool is standardized neurocognitive tests. Knowledge of the early cognitive changes typical for DLB may aid in the differential diagnosis between DLB and AD and provide insight into the nature of the underlying impairments. Identifying similarities and differences between DLB and AD is the first step in developing tests that are differentially sensitive to the different conditions. Understanding the phenomenology of cognitive impairments also provide valuable information for patients and carers, helping them to develop better coping and care strategies. The diagnostic consensus criteria [7] specify that the cognitive profile in DLB may differ from AD in that confrontation naming is relatively preserved as well as better short and medium term memory recall and recognition, with more severe impairment on verbal fluency, visual perception, and “performance tasks”, but without citing strong supporting evidence. Further, the claim that verbal fluency is relatively more impaired in DLB as compared to AD has not been corroborated in recent studies [8]. Also, preliminary evidence suggests that visuospatial deficits are more pronounced in DLB on visuoconstructional tasks than on tests involving visual perception [9]. Significant variations between studies in subject groups, methodology, and cognitive measures have limited the opportunity to conclude. In addition, most studies are based on small cohorts, and few studies have focused on the early cognitive changes [10].
In the consensus diagnostic criteria [7] for diagnosing DLB, parkinsonism, visual hallucinations, and cognitive fluctuations are core diagnostic criteria for DLB. Identifying relations between neurocognitive functions and the core DLB criteria would help increase our understanding of the relations between pathophysiology, core symptoms, and cognition in DLB. It has been proposed, partly based on findings from studies on dementia associated with Parkinson’s disease, that visual hallucinations are the result of combined deficits in visual processing with executive, attentional deficits [11]. However, a recent study [9] on cognitive correlates of visual hallucinations in DLB did not find a systematic relation to perceptual visuospatial cognitive deficits as assessed by the VOSP battery [12], only to visuoconstruction, assessed by the clock-drawing test and with psychomotor speed and digit span. A similar pattern was also recently reported by Hu and colleagues [13], finding only an association between visual hallucinations and executive deficits, but not with visuospatial deficits. Cognitive fluctuations should by definition correlate with attentional measures, as has been shown [14]. Finally, we have not found any studies which have investigated the relationship between symptomatic level of parkinsonism with neurocognitive measures in DLB, although Molloy and colleagues reported findings regarding the impact of Levodopa on cognition [15].
The objective of this study was to describe the neuropsychological profile of a relatively large sample of patients with mild DLB relative to patients with mild AD. In addition, we wanted to explore whether the cognitive profile could discriminate between AD and DLB. Finally, we explored the relation between neurocognitive measures and core diagnostic features for DLB: Visual hallucinations, cognitive fluctuations and parkinsonism.
Based on previous studies, we hypothesized that patients with DLB would exhibit more executive and visuospatial impairment and less memory impairment relative to AD patients and that visual hallucinations would be related to executive and visuospatial impairment [9, 16] and that tests of attention would be correlated with measures of cognitive fluctuation.
MATERIALS AND METHODS
Subjects
From March 2005 to March 2007, we screened all referrals to outpatient clinics in geriatric medicine and old age psychiatry in the counties of Rogaland and Hordaland in Western Norway for patients with a first time diagnosis of mild dementia, defined as Mini-Mental State Examination (MMSE)≥20 or CDR = 1. From April 2007 we selectively recruited patients with DLB and Parkinson’s disease with dementia (PDD). Additionally, the 3 neurology outpatient clinics in the same area were contacted, and agreed to refer new dementia cases to one of the participating centers. Routine blood tests were performed to exclude extra-cranial causes of dementia, and a structural MRI was performed to detect cerebrovascular disease or other structural causes of dementia. Patients without dementia or with acute delirium or confusion, terminal illness, recently diagnosed with a major somatic illness, previous bipolar disorder or psychotic disorder were not included. More details of the recruitment procedure are described elsewhere [17]. The patients and caregivers were first seen by the study clinician, who performed a structured clinical interview regarding demographical and clinical data. The comprehensive assessment procedure included a detailed history using a semi-structured interview, clinical examination including physical, neurological, psychiatric, and neuropsychological examinations. In addition, structural MRI and routine blood tests were performed on all cases; approximately 70 cases consented to lumbar puncture and CSF analysis. Thirty-five patients with clinically probable DLB had a dopamine transporter SPECT scan (Dat scan), with an abnormal scan in 27 patients, 7 normal and 1 probably normal scan, based on visual assessment [18]. Pathological diagnosis was available for 19 of the included patients in the present study (patients with probable DLB or probable AD). Seven out of 7 patients with clinical DLB had a matching pathological diagnosis, and 12 out of 12 patients with AD did. More details of the assessment program is provided elsewhere [17].
Clinical assessment
Standardized clinical assessments included the motor subscale of the Unified Parkinson‘s Disease Rating Scale (UPDRS) for parkinsonism, Neuropsychiatric Inventory (NPI) to assess psychiatric symptoms including visual hallucinations (VHs) and the Fluctuation Inventory–Scale [19] or Mayo fluctuation Questionnaire [20] for cognitive fluctuations (FC). Sleep disturbances including REM sleep behavior disorder (RBD) were monitored with the Mayo sleep Questionnaire [21]. Neuroleptic sensitivity was classified as previously reported [22]. The Montgomery and Åsberg Depression Rating Scale (MADRS) [23] was used to assess depression. The MMSE [24] was used as a cognitive screening test and the Clinical Dementia Rating scale (CDR) [25] was used to assess global dementia severity.
As previously reported [26] continuous scores for the core and suggestive DLB features were calculated; for visual hallucinations (frequency x intensity) using the NPI scale item 2 with a range of 0–12; parkinsonism on the UPDRS motor subscale (0–108); fluctuating cognition by the clinician assessment of cognitive fluctuations (0–16) (a subgroup on the Mayo fluctuation questionnaire (0–4)) and combined as previously described [26]. RBD was determined with the Mayo sleep questionnaire (0–4). Trained research clinicians performed all assessments, and bi-annual meetings were held to harmonize the procedures across the study centers.
Cognitive assessment
A battery of neuropsychological tests measuring verbal memory, executive, attention, and visuospatial functions was administered by a trained study nurse to corroborate a history and/or clinical signs of cognitive impairment.
Verbal memory
Verbal memory was assessed using the California Verbal Learning Test II (CVLT-2) [27], consisting of 16 words which were read five times, and after each time the subject was asked to recall as many words as possible. Total immediate recall (sum of trials 1–5), short-delay, and long-delay free/cued recall (after 20 min), as well as recognition, using d-prime, taking into account false positive rate [28].
Visuospatial abilities
The Silhouette test from the Visual Object and Space Perception Battery [12], and the pentagon test from MMSE were used. Silhouettes are blackened shapes of 15 objects and 15 animals as they appear at angular rotations affording a variable range of difficulty. The pentagon test differs from the silhouettes test in that it requires the subject not only to perceive and name shapes, but also to draw them, tapping into visuoconstructional abilities.
Executive functions
Executive functions were assessed using three tests: verbal fluency (category) [29] as a measure of initiation and semantic knowledge was administered by asking the patients to generate as many names of animals as possible within 1 minute, psychomotor speed and attention shift were assessed using the Trail Making Test (TMT) A and B [29]. As 60% of the patients were unable to complete the TMT-B test, the results were dichotomized into a binary able/not able to perform TMT-B variable. This was done according to the following criteria: a) the patient did not complete within a time-limit of 6 minutes, b) the patient did not understand the task, c) the patient started, but refused to complete the task, d) the patient was judged by the examiner to be unable to perform the task and the procedure was terminated or not started. While this algorithm introduces a certain degree of heterogeneity (i.e., different causes for non-completion of TMT-B), it is nevertheless a definition with clinical application. Scoring of the ability to perform TMT-B variable was done blinded to the clinical diagnosis. Attention and inhibition were assessed using the Stroop Color-Word Test [30]. Subjects name the colors of colored patches (1st card), read printed words (2nd card), and read printed color names in which the ink used for printing is a color different from the color designed by the printed name (3rd card).
Ethical issues
The study was approved by the regional ethics committee and the Norwegian authorities for collection of medical data. The patients provided written consent to participate in the study after the study procedures had been explained in detail to the patient and a caregiver, usually the spouse or offspring.
Statistical analyses
All continuous variables were inspected with histograms and were found suitable for parametric statistical analyses with the exception of TMT-A, which subsequently was log-transformed for the analyses. Univariate analyses were performed using t-tests and categorical variables were analyzed using chi-square. Effect size, Cohen’s d, was calculated for all comparisons. For the categorical variables, Cohen’s d was converted from the phi statistic. Conventionally, a Cohen’s ≥0.8 is considered a large effect size, 0.5 medium, and 0.3 a small effect size [31].
Sequential binary logistic regression with diagnostic group as outcome (AD:0; DLB:1) was subsequently performed with age, gender, education, and MADRS entered as covariates in the first block, followed by a stepwise block including all the neurocognitive variables which were statistically significant different in the AD versus DLB groups after Bonferroni-correction in the univariate comparisons. The stepwise procedure used forward stepwise selection with entry testing based on the significance of the score statistic (criterion set at p < 0.05), and removal testing based on the probability of a likelihood-ratio statistic based on the maximum partial likelihood estimates (criterion set at p < 0.10).
Finally, partial correlations were calculated between the neurocognitive variables which were significantly different between the AD and DLB group and visual hallucinations (NPI-2 for patients having visual hallucinations), cognitive fluctuations (Mayo or COGA score) and parkinsonism (UPDRS motor subscale).
RESULTS
Clinical and demographic characteristics
Of the 251 patients with a diagnosis of mild dementia, 77 (30%) patients had a clinical diagnosis of probable DLB, 113 (45%) probable AD, and 28 (14%) other dementia diagnoses (i.e., vascular dementia, PDD, frontotemporal dementia, and alcoholic dementia). Patients with DLB and AD were included in further analysis. Clinical and demographic characteristics of the groups are shown in Table 1. There was no difference in age or overall cognition as measured with MMSE or CDR between the groups. In the DLB group, UPDRS scores for parkinsonism were available for 68 patients (Mean = 14.62, SD = 13.11), NPI scores for visual hallucinations were available for 56 patients (Mean = 4.13, SD = 4.43), and cognitive fluctuation scores were available for 31 patients (Mean = 4.45, SD = 4.64).
In Table 2, univariate comparisons for all the neurocognitive variables are shown. According to a Bonferroni-corrected alpha limit of p < 0.003, DLB patients performed worse than AD patients on ability to perform TMT-B, log-transformed TMT-A, pentagon drawing from MMSE, and on all Stroop scores, while the AD group performed worse than the DLB group on short and long delay free recall from CVLT-2 as well as on the recognition score (D-prime). Medium to large effect-sizes were found for Stroop Word, long and short free recall, and ability to perform TMT-B.
Results from the sequential binary logistic regression with diagnostic group as dependent variable are shown in Table 3. The final, stepwise procedure selected reading words from Stroop, delayed recall and recognition score (D-prime) from CVLT-2, and ability to complete TMT-B as significant predictors of diagnostic group. Correct classification of DLB patients in the final model was 68% versus 85% for the AD group, resulting in a total classification accuracy of 79.1% with a sensitivity of 74% and a specificity of 82%. The final block with the set of neurocognitive variables included resulted a significantly better model fit than the first block which included only MADRS, age, sex, and education (χ2: 39.85, p < 0.001), with -2 log-likelihood of the full model of 113.75 (Nagelkerke R2 = 0.51) versus 153.64 for the first block (Nagelkerke R2 = 0.22).
To investigate the relations between the core diagnostic criteria for DLB (parkinsonism, visual hallucinations, and cognitive fluctuation) [7], we used partial correlations between these core features and the cognitive variables which differed significantly between the DLB and AD groups while controlling for age, sex, and education (see Table 4).
Parkinsonism correlated with all timed tests and with pentagon drawing, but not with the memory test scores. No other significant results were found. In order to assess whether the correlations between the timed tests and parkinsonism were mainly due to bradykinesia, we calculated separate average bradykinesia versus non-bradykinesia scores from UPDRS and correlated these measures with the cognitive variables. Mostly, the patterns of results were the same for bradykinesia and non-bradykinesia parkinsonism, e.g., for Trail-making A, the correlation with bradykinesia was 0.460 and with non-bradykinesia 0.402, p < 0.002 for both correlations. To mitigate the risk of making a type II error regarding the nonsignificant correlations between fluctuations and visual hallucinations with the cognitive variables, we dichotomized the scores for visual hallucinations and fluctuations, using a cut-off of zero. We then performed analyses of covariance using age, sex, and education as covariates, comparing DLB patients with and without visualhallucinations/fluctuations on the cognitive variables. Although visual hallucinations were associated with better performance on short delayed recall (p = 0.046) and long delayed recall (p = 0.049), this did not reach significance due to the Bonferroni corrected alpha limit of 0.002. None of the other analyses showed a tendency toward significance. It should be noted that as only a subgroup of the patients had valid data for these analyses, negative results may be due to low statistical power.
DISCUSSION
The main findings were that mild DLB was associated with worse performance on all timed tests involving attention and/or executive functions, except verbal fluency, and on pentagon drawing, but with superior performance on all measures of delayed verbal memory as compared to mild AD. In the DLB group, degree of parkinsonism was strongly correlated with the neurocognitive measures where performance was worse than in AD, both on the timed tests and on pentagon drawing. Our hypotheses of a relation of cognitive functioning to visual hallucinations and fluctuating cognition were not confirmed.
The finding of more impaired reading and color naming speed in DLB as compared to AD replicates the findings of Andersson and colleagues [32]. In the present study, the largest effect-size for all cognitive tests was found for simple word reading on the Stroop test. Reading words in literate subjects is a highly automatized skill, which is why the Stroop effect occurs in the interference condition. Thus, a specific deficit in word reading speed in the absence of deficits in visual perception, combined with less pronounced deficits in the interference condition and with no difference in verbal fluency, may indicate a phenomenon resembling the old bradyphrenia construct, a slowing of mental processes— in this case evident in a highly automatic mental task. The finding of a strong correlation with parkinsonism may indicate a shared pathophysiological substrate between parkinsonism and this bradyphrenic phenomenon, although it cannot be ruled out that the intensive speech requirements inherent in the task plays its part and that bradykinesia could explain the pattern of correlations between parkinsonism and the Stroop task as well as Trail-making A. However, DLB patients did not perform worse than AD on verbal fluency and bradykinesia can hardly explain the correlation with pentagon-drawing. Further, when analyzing bradykinesia and non-bradykinesia parkinsonism separately, no clear differences in the correlations with the cognitive variables were found.
The strong relationship between parkinsonism and cognitive performance on tests where DLB patients performed worse than patients with AD has not been reported earlier in the literature, based on a literature search in PubMed using the phrase “parkinsonism AND Dementia with Lewy bodies AND (cognit* OR neuropsychol*)”.
The finding of impaired pentagon drawing in DLB relative to AD replicates earlier studies [33, 34], and fall in line with results from other visuoconstructional tests [8]. However, in the present study, performance on the Silhouettes test from VOSP was similar to the AD group. The Silhouettes test consists of degraded visual stimuli depicting objects that are spatially distorted and rotated to varying degrees. Thus, this test resembles Boston naming test in that it involves confrontation naming and the results are quite similar. This may be interpreted as a confounded result in that two separate cognitive processes, verbal categorization and visual perception, may both contribute to the overall performance, leading to difficulty in interpreting the findings as the verbal naming component may be more impaired in AD than in DLB, while the opposite may be the case for earlier visual-perceptual processing stages.
The finding that the AD patients performed worse than DLB on all tests of delayed memory is also well established [8, 35–37]. The diagnostic criteria for AD require memory impairment, while this is not mandatory in the consensus criteria for DLB. Thus, there is an element of circularity that renders the findings difficult to interpret. However, in this light it is interesting to notice that in the present study, none of the memory measures correlated with any of the core diagnostic features in DLB.
This is one of the largest studies comparing the cognitive profile of people with AD and DLB, and one of very few studies focusing on patients with mild dementia. Our findings are consistent with previous studies, showing that AD is characterized by memory impairment, whereas impairments of attention, executive, and visuoconstructive functions were more marked in DLB patients. The cognitive profile correctly classified 79.1% of the patients. However, there was also a considerable overlap, and 32% of the DLB patients were misclassified, yielding a sensitivity of 74% and a specificity of 82% in the omnibus statistical model. Thus, as has been previously reported by the clinical diagnostic criteria for DLB, when the typical DLB pattern occurs there is a high likelihood that the diagnosis is DLB (i.e., high specificity), but a considerable proportion may not have the typical features, including cognitive profile. This is also consistent with our previous study, showing that memory complaints were the most common presenting symptom not only in AD but also in those later fulfilling DLB criteria [38].
Although few studies have explored the clinic-pathological relationship in DLB, there is evidence that the Lewy-body pathology is pronounced and widespread when a clinical diagnosis has been made, typically involving both fronto-subcortical circuits, likely associated with attention and executive cognitive changes, as well as posterior areas such as parietal and occipital cortex, likely affecting the visuospatial circuits. In addition, Alzheimer-type changes, in particular amyloid plaques, as well as dopaminergic and cholinergic deficits, contribute to the clinical profile of DLB. The association between cognitive and motor symptoms supports the notion that striatal changes, both dopaminergic fronto-striatal circuits [18] and striatal amyloid pathology [39, 40], contribute significantly to the cognitive impairment in mild DLB and in the present study, the relationship betweenparkinsonism and all attentionally and executively demanding tasks (except verbal fluency) probably may be linked to frontostriatal dysfunction due to the strong and specific correlation with parkinsonism.
A caveat in this study is that since the cognitive profile is part of the clinical criteria, there is a potential for bias due to circularity, i.e., the cognitive profile may for some patients have been used in setting the clinical diagnosis. In a recent autopsy-based study, i.e., without this potential bias, the cognitive profile at the stage of mild dementia or even before dementia was present, was compared in DLB and AD: Those with pure DLB (n = 12) had more visuospatial dysfunction and less delayed memory recognition than AD [8]. Thus, a weakness in the present study is the lack of autopsy confirmation in most subjects. However, our preliminary autopsy results indicate a total accuracy of 100% (DLB: 7/7; AD: 12/12) with regard to match between clinical and main pathological diagnosis. Further, our assessment of visuospatial functions was limited and did not capture the full spectrum of visuospatial functions. The use of VOSP silhouettes confounds verbal naming with visual perception, hampering interpretability. Thus, the present study cannot conclude regarding visuospatial functions in general.
Strengths of the study include the relatively large cohort of DLB patients with mild dementia, the use of a comprehensive battery of cognitive tests, the use of Dat scan, longitudinal follow-up with diagnostic re-evaluation to support the clinical diagnosis, and the available pathological diagnosis in a considerable proportion of patients. However, with no pathological verification of the diagnoses in most cases, misdiagnosis cannot be excluded. This suggests that our findings are conservative estimates and that the cognitive differences and their classification properties may be even higher.
Future studies should use novel imaging techniques to explore in vivo the mechanisms underlying cognitive and other clinical features in DLB, and more longitudinal studies describing the course of the cognitive domains in DLB. This may provide important background information in planning clinical trials in DLB.
Footnotes
ACKNOWLEDGMENTS
This work was founded through a postdoctoral grant to first author AR from the regional health authorities (Helse Vest). The project was funded by the Western Norway Regional Health Authority. We thank the DemVest research staff, the patients, and the caregivers.
