Verbal and visuospatial deficits in dementia with Lewy bodies

In contrast to the neuritic plaques and neurofibrillary tangles of Alzheimer disease (AD), dementia with Lewy bodies (DLB) is characterized by α-synuclein inclusions (i.e., Lewy bodies [LBs]) in neocortical, limbic, and subcortical regions. DLB and associated disorders of α-synuclein aggregation (synucleinopathies) are the second most common cause of neurodegenerative dementia after AD.¹ Although a minority of DLB cases have only LBs at autopsy (“pure” DLB), the more common finding is a mixture of LBs with sufficient burden of neuritic plaques and neurofibrillary tangles to also diagnose AD (mixed DLB/AD). The reverse is also true, as LBs are observed frequently in AD. A review of autopsied cases from the Washington University Alzheimer Disease Research Center (ADRC) suggests that as many as 25% of autopsied AD cases have cortical/limbic LBs in sufficient quantity to be classified as mixed DLB/AD, although most did not clinically manifest the distinctive symptoms of DLB.^2,3 Other autopsy series confirm the high co-occurrence of DLB and AD.^4,5

Although there are published criteria for the clinical features of DLB,³ these criteria appear to best capture the “pure” DLB cases. It is unclear how well the criteria capture the more common mixed DLB/AD group, many of whom may be misclassified clinically as probable AD.¹ Diagnostic criteria for DLB include progressive cognitive decline plus at least two of the following: parkinsonism, visual hallucinations, or cognitive fluctuation.³ Because neuropsychological studies of well-characterized samples with confirmed neuropathology are lacking, the specific cognitive features of DLB and its rate of progression remains unclear.^6–9 Some studies suggest that DLB is marked by specific declines in executive function^10,11 and visuospatial ability,^12–17 but others find a visuospatial deficit alone.^18–20 Interpretation of these findings is limited by failure to test across multiple cognitive domains, small sample sizes, and contamination of the DLB patient groups with varying levels of AD pathology.

This report describes cross-sectional and longitudinal performance on a battery of cognitive tests in people with autopsy-confirmed pure DLB, pure AD, and mixed DLB/AD. Our goals were to determine which tests differed among the groups, whether rates of cognitive decline varied, and whether differences could be attributed to either verbal, visuospatial, or executive domains. Analyses with autopsy-confirmed diagnoses, irrespective of antemortem clinical diagnosis, are an important first step in defining the cognitive deficits associated with the synucleinopathies and hence to improve clinical diagnostic criteria.

Methods.

Results.

Discussion.

Using data from a well–characterized longitudinal sample with neuropathologic confirmation, the current findings indicate that unlike other comorbidities of AD,^23,24 the presence of LB neuropathology significantly changes the cognitive presentation of AD but does not affect rates of cognitive decline once dementia is present.

First, results indicate that the presence of LB pathology alone is not associated with the verbal memory deficits witnessed in AD. These findings are consistent with other preliminary reports in the literature.⁴⁷ Pure DLB participants performed significantly better on all tests of verbal memory than did pure AD or mixed DLB/AD participants. Further, the DLB/AD performance was roughly equivalent to the AD participants on the verbal composite score and its subtests, indicating that LB and AD neuropathologic burden did not produce worse verbal memory in the combined group. The relative preservation of verbal skills in DLB may be an important feature of DLB that can aid in its clinical diagnosis in life.

Second, the pure AD and pure DLB groups were statistically equivalent on all tests of visuospatial ability, although individual subtest scores of the pure DLB group were generally lower than the scores of the pure AD group. This finding was surprising because current diagnostic criteria and findings of others^12–17 predict that those with pure DLB should perform much worse than those with pure AD in visuospatial ability. Only when a composite score of the individual tests was computed based on factor analysis did the groups differ significantly. Given the relatively low number of subjects in the pure DLB group, the purer measure of visual spatial ability was needed to attenuate error and increase sensitivity to detect the visuospatial decrement. Although a larger sample of pure DLB participants may increase the sensitivity of statistical tests for individual subtests, the composite scoring method is a more sensitive index than any single test. Given the rarity of pure DLB and the intensiveness of longitudinal study, the composite score is an efficient index of visuospatial deficits in DLB.

In contrast to the pure DLB participants, the combined DLB/AD participants performed significantly worse on the visuospatial composite score and all its component subtests. Thus, the presence of both diseases produced poorer visuospatial performance than either pathologic burden alone. Unlike the verbal memory factor, the co-occurrence of the two diseases appears to interact, as evidenced by the significantly lower visuospatial scores, both composite and individual test scores.

Finally, the three groups were equivalent in the longitudinal analyses, indicating that the rate of decline was comparable across all three groups in both the visuospatial and verbal memory domains. Somewhat surprisingly, the presence of both LB and AD neuropathologies did not change the rate of dementia progression over the 4.5-year course measured. This suggests that LBs in the presence of AD plaques and tangles adds a selective visuospatial cognitive deficit but does not accelerate the dementing process in either the visuospatial or verbal memory domains. Further, the significant main effects in the longitudinal analyses indicate that the observed differences at the selected cross-sectional moment were present at enrollment (presumably before the clinical detection of the dementia) and persisted throughout the course of the disease. These longitudinal results extend the findings reported by others investigators.^48,49

As a retrospective study, these findings have several limitations. First, volunteers who complete brain donation may not represent the population at large. Diagnosis of DLB in life is difficult (table 2), and the neuropathologic confirmed cases presented here may represent a sampling bias. Until diagnostic criteria for DLB are refined and the clinicopathologic features of the disease are known, this potential limitation will continue to be an issue in clinical DLB research.

Second, a valid criticism of the cross-sectional design of this study is that the observed lack of differences between the pure DLB and AD groups on the individual visuospatial subtests may have been due to the fact that the pure DLB participants were not matched well to the other groups. The pure DLB group contains more men, became demented at a younger age, and died earlier, which may have resulted in their relatively lower dementia ratings (see CDR and Blessed Dementia Scale³⁰ scores in table 1). The pure DLB group represents a convenience sample of those individuals presenting for evaluation; however, every attempt was made to match the overall sample at the cross-sectional moment on gender and dementia rating. The gender distribution of our sample is consistent with other reports of male predominance in DLB.⁵⁰ Further, the longitudinal analyses used a random effects mixed linear model because it is less susceptible to premorbid cognitive ability biases (intercept effects) by comparing average slopes of cognitive decline between groups. Thus, every individual serves as his or her own control reducing any biases that may occur with poor matching of the samples. Finally, all analyses were run with the dementia level statistically controlled. This did not substantially change any of the current findings. Ultimately, this clinical sample used all available participants who had come to postmortem examination, and one must judge the reliability of the test results by weighing the homogeneity of the sample (i.e., the pure DLB group were uniquely burdened by LB pathology) vs the statistical sensitivity of the tests and the effect sizes noted in table 3.

Third, variables selected for these analyses may not optimally capture the all clinical differences between the DLB and AD. For example, the pattern of visuospatial and verbal test scores across the psychometric battery suggests that there may be an important attention/executive function component to the cognitive deficits associated in mixed DLB/AD. Although an executive functioning factor was not indicated by this factor analysis, three visuospatial tasks with significant attention demands were significantly impaired in the mixed dementia group (Trailmaking A, Block Design, and Digit Symbol). Further, the verbal tasks that did not load very highly on the verbal factor have significant executive components as part of successful test performance. Several reports in the literature site attentional deficits,^12–17 but the tasks used are often speeded, thus confounding true cognitive impairment with motor slowing due to parkinsonism. Notably, all visuospatial differences remained significant even after motor slowing was statistically controlled.

Unfortunately, to test the attentional hypothesis robustly, a battery of more difficult visual attention tests is needed. For example, Trailmaking B may be a better test of executive function in the visuospatial domain, but it has not been consistently included in the ADRC longitudinal battery and thus data were not available for most of the participants in this report. Other tasks testing specific attentional modalities (e.g., Stroop task)⁵¹ are not part of the standard ADRC battery and therefore may have precluded us from replicating the attention deficits previously described.¹³ With a wider sampling of tests of visual attention and executive functioning domains, other cognitive changes associated with DLB might be detected and an executive factor may be indicated. Other tasks that have been used to discriminate DLB from AD include both unspeeded tasks (Wisconsin Card Sort,⁵² Ravens Progressive Matrices,⁵³ Embedded Figures Test,⁵⁴ and Clock Drawing)³⁴ and speeded tasks (Trailmaking A and B, Stroop task, and select WAIS subtests^55,56 Block Design, Digit Symbol, Picture Arrangement, and Object Assembly).

Overall, these cross-sectional and longitudinal analyses are encouraging because they indicate that LB pathology in the presence of AD results in dissociable deficits that may be used to identify participants with either or both dementia pathologies. The visuospatial/verbal memory differences observed here were consistent across groups, persisted across time, and occurred very early in the dementing process. Further, these data suggest that the presence of LBs alone may not produce the verbal memory deficits seen in AD or the frank visuospatial deficits suspected by current clinical criteria. When both diseases are present, the effect of LB and AD pathologies appear to be additive, wherein visuospatial performance suffers more in the presence of combined pathologies than with AD or LB pathologies alone. In contrast, there was no evidence that the combined pathologies accelerate the course of decline. Importantly, the combined longitudinal and cross-sectional results present an integrated snapshot of a broad set of findings from other cross-sectional clinical research.^20,57–59 Domain-specific testing may be one way to distinguish clinically the two types of dementia in life. Future investigations of cognitive impairment due to neuropathologic confirmed LB and AD pathology may help to refine the diagnostic criteria to discriminate better patients with pure DLB vs the more common mixed AD/DLB presentation. The application of quantitative pathology burden counts instead of categorical diagnostic groups may also add to future investigations of the impact that LBs have on cognitive function.

Acknowledgment

The authors thank Drs. Martha Storandt, Dan McKeel, Chengjie Xiong, and Elizabeth Grant of Washington University for assistance with manuscript preparation. They also thank the Clinical and Neuropathology cores of the Alzheimer Disease Research Center for providing diagnostic data.