Parkinson Disease Neuropathology: Later-Developing Dementia and Loss of the Levodopa Response

Objective  To investigate the neuropathologic substrate for dementia occurring late in Parkinson disease (PD).

Design  We identified 13 patients with a clinical diagnosis of PD who experienced dementia at least 4 years after parkinsonism onset (mean, 10.5 years) and subsequently underwent postmortem examination. Despite levodopa therapy, 9 patients later became severely impaired and nonambulatory, requiring total or near-total care; this included 4 patients treated with 1200 mg/d or more of levodopa (with carbidopa), which was consistent with loss of the levodopa response. These 13 patients were compared with 9 patients clinically diagnosed as having PD, but without dementia, who had undergone autopsies.

Results  Twelve of 13 PD patients with dementia had findings of diffuse or transitional Lewy body disease as the primary pathologic substrate for dementia; 1 had progressive supranuclear palsy. This pathology also apparently accounted for the levodopa refractory state. Among the 12 PD patients with dementia, mean and median Lewy body counts were increased nearly 10-fold in neocortex and limbic areas compared with PD patients without dementia (P≤.002). Alzheimer pathology was modest. Only one patient met the criteria defined by the National Institute on Aging and the Reagan Institute Working Group on the Diagnostic Criteria for the Neuropathologic Assessment of Alzheimer's Disease for "intermediate probability of Alzheimer's disease." There were, however, significant correlations between neocortical Lewy body counts and senile plaques as well as neurofibrillary tangles. Senile plaque counts did not significantly correlate with tangle counts in any of the analyzed nuclei. Arteriolar disease may have contributed to the clinical picture in 2 patients.

Conclusions  Diffuse or transitional Lewy body disease is the primary pathologic substrate for dementia developing later in PD. This same pathologic substrate seemed to account for end-stage, levodopa refractory parkinsonism. The occurrence of Alzheimer pathology was modest, but was highly correlated with Lewy body pathology, suggesting common origins or one triggering the other.

IDIOPATHIC Parkinson disease (PD) is a progressive neurodegenerative disease of undetermined cause with characteristic motor findings that include rest tremor, rigidity, bradykinesia, and postural disturbance. Degeneration of the substantia nigra pars compacta is the primary anatomic substrate for the motor symptoms; microscopic Lewy bodies within this nucleus are the pathologic hallmark.

Dementia may complicate PD after years of an otherwise typical course. This problem may overshadow the motor aspects of parkinsonism and become the primary source of disability. In PD, dementia is not a presenting feature, but is common later in the disease course. The frequency of dementia among patients from clinic-hospital cohorts ranges from 6% to 29% (Table 1).^1-7 However, this range likely underestimates the problem. Patients with dementia are less likely to be referred to or followed up in PD clinics. In community-based studies, prevalence figures are a little higher, ranging from 12% to 41% (Table 1^8-12), but these numbers too may underestimate the magnitude of this problem. By definition, prevalence studies ascertain cases at a single point in time. Dementia may lead to nursing home placement and earlier death; these patients will not be counted in subsequent prevalence analyses. Among PD patients who do not initially experience dementia, the yearly incidence of dementia ranges from 2.6% to 9.5% (Table 1^6,11,13-18). Thus, if PD patients live long enough, the risk is substantial. One series estimated that the cumulative risk of dementia among PD patients by the age of 85 years was more than 65%.¹⁴

The neuropathologic substrate for the dementia that develops in PD patients has been debated. In the last quarter of the 20th century, more than 20 investigations addressed this issue (Table 2).^19-40 Among these publications, there has been no consensus whether this dementia is primarily a cortical or subcortical process or whether Alzheimer disease (AD) or Lewy body disease is the primary contributor. In some patients, the pathologic cause for dementia was not apparent. Clinical heterogeneity may explain some of this lack of consensus. In most series, clinical details were sparse, including the temporal relationship between parkinsonism and dementia (Table 2). Some of these series may have included patients in whom the dementia was a presenting symptom. Early dementia may have an entirely different pathologic basis than dementia developing several years after the onset of levodopa-responsive PD. In fact, dementia developing concurrently with parkinsonism plus other characteristic features (including hallucinations and cognitive fluctuations) is now recognized to typically represent dementia with Lewy bodies.^41,42 Clinical parkinsonism may also complicate AD, in which case the dementia is prominent early in the disease course.^43-47

The uncertainties regarding the neuropathologic substrate for dementia that complicates PD may also relate to histologic techniques. Lewy bodies (especially cortical) and other cellular inclusions were underestimated based on microscopy with the older hematoxylin-eosin and silver impregnation techniques. With the application of ubiquitin immunohistochemical analysis, these were more easily identified.^48,49 Recently, α-synuclein immunohistochemical analysis was recognized to be a much more sensitive and specific technique for identifying Lewy bodies and distinguishing these from other cellular inclusions.^50,51 The studies presented in Table 2 were performed before α-synuclein immunohistochemical testing was available.

We were interested in a specific subset of patients whom we commonly see in our PD clinics—those with typical early PD who later experience dementia. We anticipated that the neuropathologic basis for this dementia should be apparent and may be fairly uniform if (1) we restricted our PD patients to those in whom parkinsonism preceded dementia by at least 4 years and (2) the histologic techniques included α-synuclein immunohistochemical analysis. A preliminary report of our findings has been published, documenting diffuse or transitional Lewy body disease as the primary substrate for the dementia in all but one of these patients (the exception having progressive supranuclear palsy).⁵² Hurtig et al⁵³ and Mattila et al⁵⁴ have also reported similar findings.

From the Mayo Health Sciences Research Database, we identified all patients undergoing autopsy between 1976 and 1997 with diagnoses in life of PD and dementia. All clinical histories were reviewed to identify those who met the following criteria: (1) clinical picture of idiopathic PD without features that suggest another parkinsonian disorder; (2) an initially favorable response to levodopa therapy, if administered; and (3) dementia developing at least 4 years after PD motor symptom onset. Initially, we selected 5 years as the minimum criterion interval between PD motor symptom onset and dementia; however, because of the small number of patients, we reduced this to 4 years (adding 2 patients). Parkinsonism onset was dated from the time the initial symptoms were experienced by the patient. Onset of dementia corresponded to the date when cognitive impairment was first noted. To qualify as PD, all patients were required to have at least 2 of the 3 cardinal PD motor symptoms (rigidity, rest tremor, and bradykinesia) and have been diagnosed as having PD by a Mayo Clinic physician. To qualify as having dementia, patients fulfilled criteria for dementia according to the Diagnostic and Statistical Manual of Mental Disorders, Revised Third Edition⁵⁵ as documented in the medical histories.

From the same Mayo database, we identified all PD patients undergoing autopsy between 1976 and 1997 who met criteria for absence of dementia and for whom adequate brain tissue specimens were available for analysis. Criteria for PD were the same as in the group with dementia. Nondementia status within the last year of life was documented by fulfilling 1 of 2 criteria: (1) achieving a score that was within normal limits on the Mini-Mental State Examination,⁵⁶ Short Test of Mental Status,⁵⁷ or formal psychometric examinations or (2) having been seen within a year of death at least twice by Mayo Clinic physicians and having medical records that documented the clinical impression of no dementia. These patients served as the control group. Both the patients with and without dementia were selected without knowledge of the patient's neuropathologic condition.

Brain areas examined included multiple neocortical regions (frontal, parietal, and calcarine), anterior cingulate cortex, nucleus basalis of Meynert, amygdala, entorhinal cortex, hippocampus, thalamus, and multiple brainstem nuclei, including substantia nigra and locus coeruleus. Microscopic sections were stained with hematoxylin-eosin, modified Bielschowsky and thioflavin S stains, and immunohistochemical stains for α-synuclein (polyclonal⁵⁸) and tau (monoclonal, PHF-1; Peter Davies, PhD, Albert Einstein College of Medicine, Bronx, NY). Microscopic features of these sections were evaluated independently by 2 observers (J.E.P. and D.W.D.). Although attempts were made to mask these observers to the clinical information, this was not consistently possible. Assessment of AD-type pathologic features included determination of Consortium to Establish a Registry for Alzheimer's Disease (CERAD)⁵⁹ neuritic plaque scores and Braak and Braak staging.⁶⁰ The National Institute on Aging and the Reagan Institute Working Group on the Diagnostic Criteria for the Neuropathologic Assessment of Alzheimer's Disease (NIA-Reagan) diagnostic criteria were used for the diagnosis of AD.⁶¹ Criteria for the pathologic diagnosis of Lewy body dementia were from the Consortium on Dementia With Lewy Bodies.⁴¹

Quantitative analyses included counts of Lewy bodies, senile plaques, and neurofibrillary tangles (×200 microscopic field) in brain regions from patients with and without dementia. Lewy body counts were performed in the substantia nigra, neocortex, limbic cortex, amygdala, and nucleus basalis of Meynert. Senile plaques and neurofibrillary tangles were counted in the neocortex, the CA1 region of the hippocampus, and the entorhinal cortex with thioflavin S fluorescent microscopy using an Olympus BH2 microscope (Olympus America Inc, Melville, NY) with a 490-nm bandpass filter.

Statistical analyses were performed using JMP computer software (JMP Software, version 4.0.4; SAS Institute Inc, Cary, NC).⁶²t Tests were used for comparison of the clinical demographic data between groups. Preliminary analyses (Shapiro-Wilks W test) indicated that much of the pathologic data were not normally distributed, and hence, nonparametric statistical tests were used for these assessments. Comparisons between groups were by Wilcoxon rank sum test or, where appropriate, χ² analysis. Spearman rank order statistics were used for correlative assessments.

P<.05 was considered statistically significant.

From the Mayo database, 162 PD patients were identified who had undergone autopsy between 1976 and 1997. Dementia had been coded in 64 of these patients. Of these, 13 patients (6 men, 7 women) met the inclusion criteria for PD with later-developing dementia (most of the remainder had dementia documented early in the disease course). From the cohort of 162 PD patients who had undergone autopsy, we included all 9 who met criteria for absence of dementia (5 men, 4 women). Mayo Clinic neurologists were involved in the diagnosis and care of all patients except for a single PD patient without dementia who had rest tremor–predominant PD. As given in Table 3, there was a trend toward earlier onset of parkinsonism in the group with dementia, whereas the age at death was similar. Parkinsonism duration was significantly longer in the demented group and much more severe in the last year of life (based on the Hoehn and Yahr scores). The mean interval from parkinsonism onset to dementia was 10.5 years. All patients in the PD group with dementia had been administered levodopa therapy and initially were at least moderately responsive, except for one. That patient (patient 3) developed parkinsonism in 1959 and did not receive levodopa until 13 years later, by which time she was experiencing dementia and hallucinating; levodopa was mildly beneficial in lower doses but exacerbated her psychosis. In view of the otherwise typical PD and the long interval from onset to levodopa treatment, we elected to include her. One patient without dementia had mild clinical symptoms with rest tremor–predominant parkinsonism and had not been administered levodopa therapy.

Dementia severity and the clinical state within a year of death are summarized in Table 4. Dementia was documented by formal psychometric or office mental status testing in 7 of the 13 patients. In the rest, the clinical records revealed confusion and unequivocal cognitive impairment; these observations are summarized in Table 4. Hallucinations or delusions were experienced by 10 of the 13 patients after dementia onset. Dementia was at least moderately severe in all, sufficient to substantially interfere with activities of daily living, and very apparent to family and caregivers. Fluctuations in cognition were clearly documented in 2; these could have been present in others but not documented. In the group without dementia, absence of dementia within a year of death was supported by normal performances on an office mental status examination (5 patients) or formal psychometrics (1 patient). Nondemented status was inferred from the medical records in the remaining 3 patients, all of whom had been seen at least twice within the last year of life by a Mayo Clinic physician.

Of the 13 patients with dementia, 8 had Hoehn and Yahr stage 5 disease at the end of their life, were severely impaired, and required total or near-total care. Only 3 were ambulatory without assistance (patients 7, 8, and 11), including 1 who was ambulatory until preterminal hip fracture. All 13 patients with dementia continued levodopa therapy but most had a poor response (Table 4). Among the patients with Hoehn and Yahr stage 5 disease, 4 were treated with relatively high doses of carbidopa levodopa-therapy (1200-2200 mg/d) and clearly had lost their levodopa responsiveness. This contrasts with the group without dementia in which the median Hoehn and Yahr score in the last year of life was 2 to 3. All 8 levodopa-treated patients who did not have dementia remained responsive to levodopa, although incompletely responsive in some patients, including 2 with medication refractory imbalance.

On neuropathologic evaluation, one patient with dementia who had an initially favorable response to levodopa therapy demonstrated typical histopathologic features of progressive supranuclear palsy. One patient without dementia had striatonigral degeneration (multiple system atrophy); she also had responded to levodopa therapy, although late in the disease course she developed marked imbalance. These 2 patients are not included in the subsequent analysis. The remainder of the patients in both groups had typical Lewy body PD within the substantia nigra.

Median and mean Lewy body counts were approximately 10 times higher among the 12 PD patients with dementia compared with the 8 PD patients without dementia in the neocortex, limbic cortex, and amygdala (all statistically significant; Table 5). In the nucleus basalis and substantia nigra, the median and mean Lewy body counts were almost twice as high in the group with dementia, but this failed to reach statistical significance (Table 5).

All 12 PD patients with dementia had pathologic findings consistent with either diffuse (neocortical) or transitional Lewy body disease.⁴¹ Transitional Lewy body disease, in which the Lewy body pathology is primarily confined to limbic areas with sparse neocortical involvement, was found in 7 patients; the remaining 5 patients had widespread Lewy body pathology that included neocortex.

Among the PD patients with dementia, there was a trend toward increased senile plaque and neurofibrillary tangle counts in the neocortex, the CA1 region of the hippocampus, and the entorhinal cortex (Table 6). However, compared with the patients without dementia, this was only significant for neurofibrillary tangles in the CA1 region. The CERAD⁵⁹ neuritic plaque scores ranged from sparse (CERAD A) to moderate (CERAD B).

Only one PD patient with dementia met criteria for early AD as defined by (1) NIA-Reagan criteria for "intermediate probability of AD,"⁶¹ with the presence of neocortical neurofibrillary tangles (Braak and Braak stage IV),⁶⁰ and (2) a moderate (CERAD B) plaque score. In all other patients, neurofibrillary tangles were confined to entorhinal and hippocampal areas (Braak and Braak stage II-III; NIA-Reagan "low probability of AD"). Coexisting large and small vessel disease (atherosclerosis, arteriolosclerosis) was present in all patients, with and without dementia, as commonly seen in elderly patients. Among the patients with dementia, 2 had white matter pallor and 1 of these had small white matter infarcts (lacunes).

Among the PD patients with dementia, 3 remained ambulatory before death (patients 7, 8, and 11), with less advanced parkinsonism than the other patients with dementia (Table 4). Pathologically, these could not be distinguished from the other PD patients with dementia. Mean Lewy body counts were similar to the group with dementia as a whole, as were the counts of senile plaques and neurofibrillary tangles. The neuropathologic diagnosis was diffuse Lewy body disease in 2 (patients 7 and 8) and transitional Lewy body disease in the other (patient 11).

Among the PD patients without dementia, no substantial neuropathologic change was found beyond the brainstem. Among these patients, Lewy bodies were sparse in the neocortex and limbic cortex; none had more than minimal evidence of AD pathologic change.

Lewy body counts in each of the analyzed brain regions were highly correlated with counts in the other areas (Table 7). Patients with high counts in one brain region were likely to have high counts in the others. The analysis given in Table 7 includes all PD patients (with and without dementia); however, similar values were also found when the analysis was restricted to only the patients with dementia (although less robust; data not shown).

In the neocortex, Lewy body counts were highly correlated with counts of senile plaques and, to a slightly lesser extent, neurofibrillary tangles (Table 8). This was not an artifact of age, since age at death was not correlated with neocortical counts of Lewy bodies, senile plaques, or neurofibrillary tangles (Spearman ρ values all ≤0.1 and P>.60). Furthermore, senile plaque counts were not correlated with counts of neurofibrillary tangles within any of the analyzed brain regions (neocortex, CA1 area of hippocampus, or entorhinal cortex). The analysis given in Table 8 included PD patients with and without dementia. When restricted to only those patients with dementia, the outcomes were similar, except that the correlation between neocortical Lewy body counts and neurofibrillary tangles did not reach statistical significance (ρ = 0.49, P = .11).

Lewy body disease, either transitional or diffuse, seemed sufficient to account for dementia in 12 of 13 patients. The exception was the single patient with progressive supranuclear palsy. Two patients also had white matter pallor, one with multiple lacunar infarctions, which could have contributed to the dementia. Overall, these findings suggest that the primary cause for dementia that develops later in the course of otherwise typical PD is the underlying Lewy body disease. This finding is consistent with 2 recent series that reported similar results.^53,54 In contrast to multiple prior publications (Table 2), major Alzheimer changes were not found among our PD patients with dementia. Only one patient met NIA-Reagan criteria for intermediate probability of AD⁶¹; the remainder of the patients fell into the NIA-Reagan category of "low probability of AD."

The mean age was similar in the groups with and without dementia, suggesting that simple aging was not the cause for the pathologic differences. However, the group with dementia had significantly longer disease durations. This was likely a factor in both the clinical and neuropathologic differences; in progressive neurodegenerative syndromes, longer disease durations typically result in more severe outcomes.

Although AD did not seem to be the primary substrate for dementia in these patients, it clearly was a component of the neuropathologic process. There were trends toward higher senile plaque and neurofibrillary tangle counts in most of the analyzed nuclei of the group with dementia, although this was only statistically significant for tangles within the hippocampal CA1 region (Table 6). Remarkably, neocortical Lewy body counts were significantly correlated with counts of senile plaques and, to a lesser extent, neurofibrillary tangles (Table 8). This is not explained by age-related changes, since there were no trends linking age at death to counts of any of these neuropathologic markers. Senile plaque counts did not correlate with counts of neurofibrillary tangles within any of the analyzed nuclei.

The present findings of predominant diffuse or transitional Lewy body disease, combined with lesser Alzheimer changes, are remarkably similar to the findings in patients prospectively diagnosed as having dementia with Lewy bodies.⁴² In the latter condition, patients present with dementia with visual hallucinations. Parkinsonian motor findings may develop early or later in the disease course, in contrast to the present series with initial parkinsonism and later dementia. Despite the difference in clinical course, the postmortem pathologic features are similar, with Lewy body pathology predominantly of the transitional and diffuse type, plus, a mixture of AD pathology that in most patients does not meet NIA-Reagan criteria for AD. The findings reported herein underscore the increasing recognition of the overlap of AD and Lewy body pathologies with a continuum of clinical deficits from pure dementia to pure parkinsonism, but often with combinations.^47,63-65

The results of this study suggest that a generalized process that simultaneously affects widespread brain regions is responsible for the dementia. This is in contrast to what has been suggested by some prior investigators who have focused on specific brain nuclei^22,24,28 (Table 2). Lewy body counts in the 5 analyzed brain regions were highly correlated (Table 7) with most Spearman ρ values more than 0.7 (neocortex, limbic, amygdala, nucleus basalis, and substantia nigra). The evolving clinical state of the PD patients with dementia also suggests a generalized brain disorder. Initially, the clinical picture was that of a focal process, with typical levodopa-responsive PD. However, with time, most patients in this series not only developed dementia, but also became levodopa refractory, nonambulatory, and unable to care for themselves.

These findings also provide insight into the neuropathologic substrate for underlying loss of the levodopa motor response. Prior PD clinicopathologic studies have obviously included levodopa refractory, bed-bound patients (Hoehn and Yahr stage 5). However, this specific group of PD patients has not been singled out for neuropathologic analysis. In the present study, 8 of the 12 PD patients fit into this category; 3 had been treated with more than 1200 mg/d of levodopa (with carbidopa), suggesting a levodopa refractory state. Thus, the same neuropathologic process that accounts for dementia seems to play a critical role in the loss of the motor response to levodopa therapy. Patient 13 also had Hoehn and Yahr stage 5 disease late in the disease course, was unresponsive to high-dose levodopa therapy, and had progressive supranuclear palsy.

Heterogeneous pathologic explanations for dementia developing within the context of PD have been reported (Table 2). This contrasts with the present study in which the predominant pathologic findings were fairly consistent among patients. At least 3 factors account for this discrepancy. First, many of the earlier investigations summarized in Table 2 were performed before the availability of more modern immunohistochemical techniques. In fact, all of the studies listed in Table 2 predate the availability of α-synuclein immunohistochemical testing. This is now recognized to be the most sensitive and specific method for identifying Lewy bodies.^50,51 Second, patients included in these prior studies were likely heterogeneous. As is apparent in Table 2, basic clinical features were typically not reported and perhaps may not have been available to the investigators. It was the rare series in which such basic information as latency from parkinsonism to dementia or levodopa responsiveness was reported (Table 2). For those series that included patients in whom dementia was a presenting feature, one might expect different neuropathologic conditions, including AD. In the present series, the patients were clinically homogeneous in that dementia was a later development and all initially presented with what seemed to be typical levodopa responsive PD. The only exception was patient 3, who presented with typical PD before levodopa was available; when initiated 13 years later, the response was limited by hallucinations. Third, many prior investigators reporting Alzheimer changes as the substrate for PD dementia (Table 2) used the older Khachaturian neuropathologic criteria.⁶⁶ Recent consensus criteria for the AD neuropathologic substrate⁶¹ emphasize the importance of neurofibrillary tangles, in addition to neuritic plaques. The Khachaturian criteria focused primarily on plaques, which are now recognized as much less specific.

In summary, Lewy body pathology was the consistent and prominent finding among these PD patients with dementia. However, the mixture of mild AD pathologic changes raises questions about the interplay of these disorders.

Accepted for publication August 30, 2001.

Author Contributions:Study concept and design (Drs Apaydin, Ahlskog, Parisi, Boeve, and Dickson); acquisition of data (Drs Apaydin, Ahlskog, Parisi, Boeve, and Dickson); analysis and interpretaion of data (Drs Apaydin, Ahlskog, Parisi, Boeve, and Dickson); drafting of the manuscript (Drs Apaydin, Ahlskog, Parisi, Boeve, and Dickson); critical revision of the manuscript for important intellectual content (Drs Apaydin, Ahlskog, Parisi, Boeve, and Dickson); statistical expertise (Dr Ahlskog); administrative, technical, and material support (Drs Apaydin, Ahlskog, Parisi, Boeve and Dickson); study supervision (Drs Apaydin, Ahlskog, Parisi, Boeve, and Dickson).

Supported in part by grants AG06786 and AG16574 from the National Institute on Aging, Bethesda, Md, and NS40256 from the National Institutes of Health, Bethesda.

Corresponding author and reprints: J. Eric Ahlskog, PhD, MD, Department of Neurology, Mayo Clinic, Rochester, MN 55905.