Cognition at Each Stage of Lewy Body Disease with Co-occurring Alzheimer’s Disease Pathology 1

Participants

The current study used data from the U.S. National Alzheimer’s Coordinating Center’s (NACC) Uniform Data Set (UDS) collected between November 2005 and June 2019 from 35 past and present NIA-funded Alzheimer’s Disease Centers [15]. Participants were evaluated approximately annually at each Center using a standardized protocol beginning in 2005. Par-ticipants were enrolled with any level of cognition, ranging from normal to demented. Written informed consent was obtained from all participants and their study coparticipants; institutional review board approval was obtained from all individual Centers.

Participants selected for the current study met the following criteria (Fig. 1): 1) had both neuropath-ologic [16] and neuropsychological data, version C1 [17] available; and 2) time between the most recent UDS in-person visit and death was not greater than 2 years to ensure that the neuropsychological test results reflected cognitive status proximate to death, consistent with previous reports [18, 19]. Exclusion criteria included the presence of any of the follo-wing at autopsy: Down’s syndrome, prion disease, early-onset autosomal dominant genetic diseases, frontotemporal lobar degeneration, progressive sup-ranuclear palsy, cortical basal degeneration, multiple system atrophy, CADASIL, other 3R and/or 4R tauopathies, argyrophilic grains, chronic traumatic encephalopathy, TDP-43, tangle dominance disease, pigment-spheroid degeneration, malformation of co-rtical development, metabolic/storage disorders, leukodystrophy, multiple sclerosis or other demyelinating disease, neoplasm, cerebral microbleeds and microinfarcts (as the latter have the most support in the literature regarding impact on cognition) [20, 21].

OPEN IN VIEWER

Fig. 1

Flow Diagram of Exclusion Criteria. “Pure” Lewy body disease pathology (LBD) groups exhibited low Alzheimer’s disease neuropathologic change (ADNC). Low ADNC includes Neuritic plaque score (CERAD[26]) equivalent to C0 or C1, and Braak Neurofibrillary Tangle Score (Braak Stage; equivalent to B0 or B1 [27]). LBD groups were further delineated into brainstem (LBD-b; most commonly observed in the substantia nigra or locus coeruleus), limbic (LBD-l; cingulate, entorhinal, amygdala), and cerebral cortical (LBD-c) groups. Combined pathology groups (ADNC/LBD) exhibited intermediate or high ADNC and the presence of Lewy body neuropathology and were further delineated into brainstem (ADNC/LBD-b), limbic (ADNC/LBD-l), and cerebral cortical (ADNC/LBD-c) groups. “Pure” ADNC exhibited intermediate or high ADNC and no LBD. Healthy controls (HC) exhibited low ADNC and no LBD and no diagnosis of dementia at any evaluation. NACC, National Alzheimer’s Coordinating Center; UDS, Uniform Data Set; FTD, frontotemporal dementia; CERAD, Consortium to Establish a Registry for Alzheimer’s disease criteria.

Neuropathological characterization

Each Center conducted neuropathologic assessments following consensus guidelines and uploaded the data to the NACC using neuropathology forms 9 and 10 [22]. The presence and stage of ADNC [23, 24] and LBD [25] were used to characterize the extent of neuropathology and assign participants to one of the following eight groups (detailed in Fig. 1): Healthy controls (HC), LBD only [differentiated into brainstem (LBD-b), limbic (LBD-l), and cerebral cor-tical (LBD-c)], and combined pathology ADNC/LBD [differentiated by stage of LBD: brainstem (ADNC/LBD-b), limbic (ADNC/LBD-l), and cerebral cortical (ADNC/LBD-c)]. Specifically, consensus guidelines for LBD were used to determine the presence and stage of LBD with an NIA-AA Lewy body score of either 1, 2, or 3, corresponding to bra-instem, limbic/transitional, or cerebral cortical [25]. ADNC only and combined pathology groups (ADNC/LBD) exhibited intermediate or high ADNC. Low ADNC was defined as no/sparse neuritic plaques [Consortium to Establish a Registry for Alzheimer’s disease criteria (CERAD) score equivalent to C0 or C1] [26] and Braak stage 0-II for neurofibrillary degeneration [27] (equivalent to B0 or B1). Intermediate ADNC was defined as moderate or frequent neuritic plaques (CERAD C2 or C3) and Braak B2 (stage III-IV). High ADNC was defined as moderate or frequent neuritic plaques (CERAD C2 or C3) and Braak B3 (stage V-VI). A dichotomous variable quantifying the presence of vascular brain injury (other than those excluded for cerebral microbleeds and microinfarcts) was created to use as a covariate in statistical models.

Demographic and clinical characteristics

We report demographic data, including age, sex, education, and race/ethnicity, health histories, and clinical diagnoses as of the last study visit in the UDS (within 2 years of death). Diagnoses of normal cognition, mild cognitive impairment (MCI), or dementia were made at each Center either by a single clinician or consensus group of clinicians after a review of all evaluation information available.

Neuropsychological testing

As the majority of patients with neuropathological data available were administered the UDS C1 version of the neuropsychological assessment, only those participants were included in the current analyses. The Mini-Mental State Examination (MMSE) and neuropsychological tests administered as part of the UDS C1 are described in detail elsewhere [17]. We calculated normative z-scores using age- and sex-corrected demographic adjustments based on previously published regression norms derived using the UDS [28]. Tests were grouped into cognitive domains based on previous investigations [29]. In addition, we ensured that the within-domain subtest correlations were greater than the between-cognitive tests correlations within this sample (Fig. 2A). Domain scores were calculated by averaging the sum of the z-scores for the test scores within each of the following domains: executive/visuospatial, memory, language, and attention (Fig. 2A). A separate visuospatial domain score could not be generated since the UDS C1 neuropsychological test battery includes no stand-alone visuospatial test; however, the UDS C1 tests of executive function rely on visuospatial information processing and therefore this domain is referred to as the executive/visuospatial domain.

OPEN IN VIEWER

Fig. 2

Neuropsychological correlations and cumulative global deficit by group. A) Correlation plot depicting greater within domain correlations relative to between domain correlations among neuropsychological tests. Subtests (horizontal) contributing to each domain score (vertical). Executive Visuospatial domain includes: Trails A, Trail Making Test A time; Trails B, Trail Making Test B time; Digit Symbol, Digit Symbol Coding from the Wechsler Adult Intelligence Scale –Revised. Memory domain includes: Imm Recall, Logical memory Immediate Recall; Delay Recall, Logical Memory Delayed Recall. Language domain includes: number of correct words on Animal and Vegetable List generation; score on the modified Boston Naming Test (30 odd item version); Attention Domain includes: Digits Forward, Digit Span Forward total score Wechsler Memory Scale –Revised (WMS-R); Digits Backward, Digit Span Backwards total score from the WMS-R. B) Cumulative global deficit on Mini-Mental Status Exam (MMSE) and average MMSE Z-score at the participant’s last visit (within 2 years of death) by group. Cumulative global deficit is significantly greater for Alzheimer’s disease neuropathologic change (ADNC, red) and the combined ADNC and Lewy body disease groups (ADNC/LBD, purple) relative to Healthy Controls (HC, gray) and LBD only (blue), irrespective of stage of LBD (Brainstem, Limbic, or Cortical).

Statistical analyses

Participants were recruited at various stages of cognitive impairment with different lengths of follow up prior to their death, raising challenges to using traditional measures of longitudinal change (e.g., annualized rate of change). To overcome this, we cal-culated a measure of cumulative cognitive deficit derived using the estimated summary score trajectory based on the fitted mixed effects regression model. Specifically, we performed mixed effects regression analyses with longitudinal global cognitive ability (MMSE) or domain-specific summary scores (i.e., executive/visuospatial, memory, language, attention) as the outcome and time as the independent varia-ble of primary interest. Within-person correlations between the longitudinally measured summary sco-res are accounted for by subject-specific random intercept and slope, performed within each of the eight neuropathological subgroups. Models were adj-usted for age, gender, education, and the presence of vascular factors. The domain-specific models included the MMSE score as a covariate. Cumulative cognitive deficits are the average deviation for each cognitive domain between normative performance (a z-score of 0) and estimated summary score trajectory based on the fitted mixed effects regression model that was derived using available timepoints within a six-year time window prior to death. This allows us to summarize the cumulative impairment for patients who had less than six-year follow-up prior to their death. Additionally, the statistical model was used to estimate the underlying cumulative exposure after adjusting for a set of potential confounders.

We evaluated selected contrasts of interest with a focus on interactions between ADNC and various levels of LBD (e.g., the effect of ADNC on participants with LBD-b versus the effect of ADNC on participants with LBD-c) and applied appropriate follow-up simple effects contrasts. Statistical inferences were made via a nonparametric bootstrap method in which we repeatedly calculated the estimated cumulative deficit based on bootstrapped data. Planned sensitivity analyses were conducted to determine if results were consistent when the lowest possible score on a given test was substituted for data missing due to cognitive reasons. Two-sided probabilities of p < 0.0.0125 (0.05/4) were considered significant, after applying Bonferroni correction for the four comparisons within each model. All statistical analyses were conducted using R (R 3.6.1 The R Foundation for Statistical Computing).

Participants

Demographic, clinical, and pathological characte-ristics of the sample are presented in Table 1. Among the 2,433 autopsied UDS participants included in the current study, the mean age at death was 82.18 years (SD = 10.19 years) and the average interval between last clinical visit and death was 8.78 months (SD = 5.94 months). Across the entire sample, 111 (4.6) had LBD only [33 (1.4%) LBD-b, 43 (1.8%) LBD-l, and 35 (1.5%) LBD-c], 741 (30.5%) had combined ADNC/LBD [56 (2.3%) ADNC/LBD-b, 337 (13.9%) ADNC/LBD-l and 348 (14.3%) ADNC/LBD-c], 1,357 (55.8%) had ADNC only, and 224 (9.2%) had neither ADNC nor LBD and did not have dementia at any visit (HC) including their last visit within 2 years of death. Across all participants with LBD (N = 852), 111 (13%) had LBD only and the remainder had combined ADNC/LBD (N = 741). Clinical diagnoses for each of the neuropathology defined groups are presented in Table 1. The number and frequency of clinical diagnoses were as follows: (1) within LBD, 34 (30.6%) Parkinson’s disease, 50 (45.0%) dementia with Lewy bodies, and 20 (18.0%) Alzheimer’s disease; (2) within the ADNC/LBD, 70 (9.4%) Parkinson’s disease, 127 (17.1%) dementia with Lewy bodies, and 531 (71.7%) Alzheimer’s disease; (3) within the ADNC, 26 (1.9%) Parkinson’s disease, 36 (2.7%) dementia with Lewy bodies, and 1,089 (80.3%) Alzheimer’s disease.

Global cognitive ability

The ADNC only and all three combined pathology groups (ADNC/LBD-b, ADNC/LBD-l, and ADNC/LBD-c) all exhibited significantly greater cumulative MMSE deficit than the HC and LBD only groups (LBD-b, LBD-l, and LBD-c). Of the LBD only groups, only those with limbic and cerebral cortical stages of LBD (LBD-l and LBD-c) had significantly greater cumulative MMSE deficit than the HC group (Table 2; Fig. 2B).

Executive/visuospatial

There was a significant interaction between ADNC and LBD in the limbic and cerebral cortical stages of LBD (Table 2; Fig. 3). The limbic LBD and cerebral cortical LBD groups exhibited a significantly greater cumulative executive/visuospatial deficit than the HC group; however, only the cerebral cortical LBD group showed greater cumulative deficit than the ADNC only group. The reverse was seen for the LBD-b group, which showed significantly less cumulative executive/visuospatial deficit than the ADNC only group. Interestingly, limbic and cerebral cortical LBD groups (LBD-l and LBD-c) had sim-ilar cumulative executive/visuospatial deficit as com-bined pathology groups (ADNC/LBD-l and ADNC/LBD-c), but the LBD-b group exhibited less cum-ulative executive/visuospatial deficit as the combined ADNC/LBD-b group. Finally, the combined pathologies groups (ADNC/LBD-l and ADNC/LBD-c) also exhibited a significantly greater cumulative deficit than the ADNC only group, but only the ADNC/LBD-c group remained significant after Bonferroni correction.

OPEN IN VIEWER

Fig. 3

Barplots depicting the cumulative deficit for cognitive domains by group. In the Memory and Language domains combined LBD and ADNC (purple) showed a significantly greater cumulative deficit than pure LBD (blue) at each stage of Lewy body pathology. By contrast, in the Executive/Visuospatial domain combined LBD and ADNC showed greater cumulative deficit than pure LBD only when the Lewy body pathology was confined to the brainstem. Limbic and Cortical Lewy body pathology showed no differences in cumulative deficit with or without combined ADNC. HC, healthy control; ADNC, Alzheimer’s disease neuropathologic change; LBD, Lewy body disease; ADNC/LBD, Alzheimer’s disease neuropathologic change and Lewy body disease.

Memory

There was a significant interaction between ADNC and LBD in all stages of LBD (brainstem, limbic, and cerebral cortical). Unlike the executive/visuospatial domain, the ADNC only and combined pathology groups (ADNC/LBD-b, ADNC/LBD-l, and ADNC/LBD-c) all exhibited significantly greater cumulative memory deficit than the LBD only groups, regardless of pathologic stage (LBD-b, LBD-l, and LBD-c). The LBD-l and LBD-c exhibited significantly greater cumulative memory deficit than the HC group, but not the LBD-b group.

Language

There was a significant interaction between ADNC and LBD only in the cerebral cortical LBD stage. Similar to the memory domain, the ADNC only and combined pathology groups (ADNC/LBD-b, ADNC/LBD-l, ADNC/LBD-c) all exhibited greater cumulative language deficit than the LBD only groups, regardless of pathologic stage (LBD-b, LBD-l, and LBD-c).

Attention

No significant interaction was observed for ADNC and LBD (brainstem, limbic, and cerebral cortical examined separately). Simple effects indicated no significant differences.

Missing data and imputation

Findings were similar in models with neuropsychological test scores imputed as the worst score when data was missing due to cognitive reasons. Ra-tes of missing data due to cognitive reasons varied based on the cognitive test. For instance, 15%of ADNC, 13%of ADNC/LBD-l, 14%of ADNC/LBD-c, 10%of ADNC/LBD-b and 0%of the remaining groups were missing delayed memory test scores. In contrast, 21%of ADNC, 20%of ADNC/LBD-l, 24%of ADNC/LBD-c, 18%of ADNC/LBD-b, 12%LBD-c, 6%LBD-l and 0%of the remaining groups were missing scores on the Trail Making Test Part B.

Findings were also similar when limited to a subset of participants that were demented at their last visit (within two years of death). These analyses did not include the brain stem LBD groups since this subsample was too small. Findings were also similar when analyses included the time between last clinical visit and site as additional covariates.

Data availability

All data used in the current analyses is available for download from the National Alzheimer’s Coordinating Center (https://www.alz.washington.edu/).

The interaction between ADNC and LBD stage is domain-specific

Individuals with limbic and cerebral cortical LBD (LBD-l and LBD-c) exhibited executive/visuospatial deficits that are the same or worse than those of individuals with ADNC, suggesting that LBD uniquely impacts this cognitive domain to a greater degree, but only after it has progressed past the brainstem. Additionally, executive/visuospatial deficits did not significantly differ between brainstem LBD and HC, further suggesting impairment in executive/visuospatial function reflects a progression of LBD beyond the brainstem. In contrast, memory impairment is significantly greater when there is co-occurring ADNC across all stages of LBD. Interestingly, the deficit with combined pathology is not greater than when there is ADNC alone. This suggests that more severe memory impairment is driven by ADNC, regardless of the severity of the LBD pathology.

Executive and visuospatial impairments in LBD

As participants were recruited in the study with any stage of cognitive deficits (e.g., no cognitive impairment, mild cognitive impairment, dementia), examination of annualized rate of change in this cohort is not appropriate as many individuals may have significant impairment upon enrollment into the study. To overcome this, we used a measure of cumulative deficit that quantifies the extent of impairment modeled using the available data six years prior to the participant’s death. Patients with Lewy body dementias exhibit visuospatial and executive function impairment that is relatively more pronounced than in patients with Alzheimer’s disease dementia [13, 32–38]. The presence of early visuospatial deficits may be the most salient difference in cognitive profiles of dementia with Lewy bodies and Alzheimer’s disease dementia, and predicts the rate of decline in clinical and neuropathological cohorts [34, 40]. The executive/visuospatial domain score in the current study includes primarily executive functioning measures that rely on visuospatial information. Therefore, this measure is not specific to each of these domains and changes may reflect deficits in either of these functions. In contrast to memory, the executive/visuospatial deficit in the cerebral cortical LBD group was more pronounced than in the ADNC group and was equally severe in the limbic LBD and ADNC groups. These findings suggest that LBD is a determinant of executive/visuospatial dysfunction independent of ADNC. Further, the progression of Lewy Bodies from brainstem to limbic to cortical regions corresponds to a greater decrement in executive/visuospatial deficits. Additionally, combined pathology did not have a greater impact on executive/visuospatial deficits than LBD only. This is consistent with prior notions that the phenotypic expression of LBD may be less pronounced in the presence of co-pathology [35] and different LBD clinical subtypes may be associated with less ADNC [41].

Pathology underlying memory and language impairments

While memory impairment is a well-established predictor of dementia in Parkinson’s disease and dementia with Lewy bodies [42, 43], it is unclear to what extent memory impairment is due to LBD versus ADNC. In Parkinson’s disease patients, early memory changes may primarily be due to retrieval difficulties secondary to executive impairments (a non-amnestic memory impairment), although storage deficits related to medial temporal changes may also contribute (amnestic memory impairment) [42]. When characterizing patients based on amnestic MCI or non-amnestic MCI, prior work highlights that amnestic MCI have a propensity for increased neurofibrillary tangles in select anatomic regions, where whereas non-amnestic MCI exhibited increased Lewy body pathology [44]. The Logical Memory test used in the current study is not particularly sensitive to early memory changes [45] and may be less impacted by executive deficits, and therefore less sensitive to LBD, than other types of memory tests (e.g., word-list learning tests) since the to-be-remembered information is inherently organized and easier to retrieve [46, 47]. In the current study, we found significantly greater memory impairment in individuals with ADNC or combined ADNC/LBD than in those with LBD only, consistent with previous reports [14, 48]. While the memory deficit increased as LBD progressed across brainstem to limbic to cerebral cortical LBD groups, it never reached the same level as that of the ADNC only group. Combined ADNC/LBD did not have a significantly greater effect on memory than ADNC alone, perhaps because memory performance in the ADNC only group is near floor and any impact of LBD co-pathology cannot be discerned. With regard to language, the ADNC group had substantially greater language deficits than all LBD groups, consistent with reports of relatively preserved language in Lewy body dementias [49, 50].

Combined versus “pure” pathology in dementia syndromes

The prevalence of a dementia diagnosis at the final UDS visit varied across groups with the highest prevalence for the combined ADNC/LBD-l (93.5%) and ADNC/LBD-c (98.3%) groups. These values were higher than for those of the ADNC only (88.4%), ADNC/LBD-b (83.9%) or LBD only (LBD-b = 33.3%, LBD-l = 67.4%, LBD-c = 80.0%) groups. These results highlight that individuals with combined pathology are more likely than those with LBD only to be diagnosed with a dementia syndrome during life.

Only 13%of individuals with LBD exhibited “pure” LBD and the remainder exhibited combined ADNC/LBD with moderate or high levels of ADNC. Those with combined ADNC/LBD typically had more severe stages of α-synuclein than “pure” LBD (cerebral cortical and limbic). These results suggest a relatively lower prevalence of pure LBD relative to other studies [4, 51], likely due to the referral bias inherent in recruiting patients from memory clinics relative to movement disorders clinics. The more severe stages of LBD corresponding to greater rates of ADNC, however, is consistent with the existing literature [30]. An emerging framework proposes that the frequent co-occurrence of ADNC and LBD suggests a converging disease mechanism that promotes protein aggregation [2, 52]. Our findings of a greater global cumulative cognitive deficit (i.e., MMSE) across all combined pathology groups (ADNC/LBD-b, ADNC/LBD-l and ADNC/LBD-c) than in LBD alone and HC (Fig. 2) are consistent with this framework, however, this is given the limitation that the MMSE may have fewer items sensitive to the changes that is observed in LBD pathology. Examination of the individual cognitive domains suggests that the impact of individual and combined pathologies differs depending on the domain examined. This is likely due to the regional impact of neuropathology, and warrants further investigation.

Methodological considerations

The current study is the first to evaluate a novel measurement of domain-specific cumulative cognitive deficits in a large sample of individuals with neuropathologically confirmed LBD and ADNC; however, there are several methodological considerations. One major limitation is that the UDS test battery used for the current study cannot differentiate between visuospatial and executive function deficits. Despite this, we were able to examine the relative contributions of LBD and ADNC to deficits on executive function tasks that have a strong visuospatial component. Second, there is a referral bias due to the nature of participants recruited to Alzheimer’s Disease Research Centers, such that mixed ADNC and LBD is likely more common relative to what would be observed in a movement disorders or general neurology clinic. Therefore, the current cohort likely does not reflect the relative distribution of these pathologies in the community. Third, we cannot assess the impact of cognitive or neural reserve on the differential effects of the two pathologies and their interaction on deficits in various cognitive domains. Fourth, the available NACC forms only include limited vascular information, therefore, we cannot systematically examine vascular effects on our findings. Fifth, the cumulative impairment is estimated via random effects model to allow adjusting for selected baseline covariates and including patients whose follow-up time prior to death is less than six years in the analysis. The validity and precision of the estimated cumulative impairment depends on the model assumptions, which warrants further examination in a larger cohort. Finally, the lack of detailed information regarding the distribution of LBD (e.g., cortical pathology differentially impacting parietal relative to temporal lobe) precludes the ability to fully examine domain-specific cognitive-neuropathology correlates. Future studies examining stages and regional LBD and ADNC distributions would be helpful to further understand the impact of pathology location on domain specific cognitive impairment [53, 54].