Prognosis and Neuropathologic Correlation of Clinical Subtypes of Parkinson Disease

Key Points

Question  Does clinical subtyping of Parkinson disease at diagnosis estimate disease progression and neuropathology?

Findings  In this cohort study, clinical subtyping of 111 autopsy-confirmed patients with Parkinson disease based on motor symptoms, rapid eye movement sleep behavior disorder, and autonomic and cognitive dysfunction at diagnosis showed different disease course and survival rates across subtypes. Parkinson subtypes were associated with different rates of Lewy pathology and Alzheimer disease–related neuropathology progression.

Meaning  Clinical subtyping in Parkinson disease is feasible in clinical practice and provides accurate estimation of disease progression and survival; different neuropathologies are important determinants of clinical subtypes and contribute to the clinical heterogeneity.

Importance  Clinical subtyping of Parkinson disease at diagnosis is useful in estimating disease course and survival. Severity and rate of progression of neuropathologies are important determinants of clinical Parkinson subtypes.

Objective  To provide longitudinal clinical disease-course data and neuropathologic correlation for newly proposed Parkinson disease subtypes.

Design, Setting, and Participants  Retrospective cohort study of consecutive patients with autopsy-confirmed Parkinson disease who were regularly seen throughout their disease course by hospital specialists in the United Kingdom and donated their brain at death to the Queen Square Brain Bank between January 2009 and December 2017. Patients with additional neuropathologic diagnoses, monogenic forms of parkinsonism, or insufficiently detailed clinical information were excluded. Based on severity of motor symptoms, rapid eye movement sleep behavior disorder, and autonomic and cognitive function at diagnosis, patients were classified adapting a subtyping classification into mild-motor predominant, intermediate, or diffuse malignant subtypes.

Main Outcomes and Measures  Time from diagnosis to disease milestones (recurrent falls, wheelchair dependence, dementia, and care home placement) and death were compared between subtypes, and their risk was estimated using Cox hazard regression models. Severity and distribution of Lewy pathology and Alzheimer disease–related pathology were assessed using staging systems.

Results  From a total of 146 patients, 111 patients were included (67 men [60.4%]; mean [SD] age at diagnosis, 62.5 [11.5] years). The diffuse malignant subtype had earlier development of milestones and reduced survival. Cox proportional hazard regression showed an increased adjusted risk of any disease milestone (hazard ratio, 10.90; 95% CI, 5.51-21.58; P < .001) and death (hazard ratio, 3.65; 95% CI, 1.98-6.75; P < .001) in the diffuse malignant group. Age at diagnosis was the only additional variable with statistical significance (adjusted hazard ratio for death, 1.14; 95% CI, 1.11-1.17; P <.001). Staging of Lewy pathology and Alzheimer disease–related pathology did not differ between subtypes, although they showed different rates of progression, and the latter was associated with age at death.

Conclusions and Relevance  Parkinson clinical subtypes at diagnosis may estimate disease course and survival, which may be useful in providing a more accurate prognosis in individual patients in clinical practice and helping to stratify subgroups in clinical trials. Different severity and progression of neuropathologies are important determinants of Parkinson subtypes, and age at diagnosis should be included in future subtype classifications.

Clinical heterogeneity is well recognized in Parkinson disease (PD),¹ and many attempts have been made to identify distinct clinical syndromic patterns. Initial empirical classifications have been gradually replaced by data-driven cluster analysis without a priori hypotheses,^2,3 and nonmotor symptoms with prognostic implications have been incorporated into PD subtypes.^4-9

However, new clinical subtypes have attracted some criticism. With very few exceptions,^5,9 most cluster studies have no longitudinal follow-up, and therefore, evidence to assess the long-term validity of the proposed clinical subtypes on disease progression is lacking. Classification of patients with PD is purely based on data association and may not reflect underlying pathophysiology driving clinical heterogeneity. Furthermore, there is very little information on neuropathologic findings and clinical subtypes,¹⁰ and only a few studies provide supporting data on biomarkers.^5,7Quiz Ref ID One of these studies identified 4 clinical features (motor severity, cognitive impairment, dysautonomia, and rapid eye movement behavior disorder [RBD]) as key determinants for subtyping. Using a comprehensive data set of clinical manifestations at diagnosis, with additional neuroimaging and cerebrospinal fluid biomarker data from the Parkinson’s Progression Markers Initiative study,¹¹ 3 distinct groups based on cluster analysis were identified: mild-motor predominant, diffuse malignant, and intermediate.⁵ The prognostic value of this subtype classification remains to be confirmed because patients were followed up for less than 3 years, and its usefulness in other populations has not been tested. By using clinical records that include detailed information regarding nonmotor symptoms on a large group of pathology-confirmed PD cases at the Queen Square Brain Bank (QSBB), we now provide longitudinal data throughout the disease course and neuropathologic correlation for the newly proposed clinical subtypes and suggest possible pathophysiologic mechanisms responsible for clinical and neuropathologic heterogeneity.

This is a retrospective cohort study of consecutive patients with a neuropathologically confirmed diagnosis of PD from the QSBB in London, England, between January 2009 and December 2017. Quiz Ref IDPatients were excluded if (1) a diagnosis of dementia before or within 1 year of the onset of motor symptoms was made,¹² (2) neuropathologic diagnosis of an additional neurodegenerative condition was found on postmortem examination, (3) it was monogenic form of parkinsonism, or (4) comprehensive clinical records throughout the entire disease course were not available. The brain donor program was approved by a London Multi-Centre Research Ethics Committee, and written informed consent was obtained from all donors.

Clinical records were systematically reviewed by a neurologist with expertise in PD (E. D. P.-F.). All patients were assessed by hospital specialists (neurologists or geriatricians) in the United Kingdom regularly throughout disease course. The presence and severity of the cardinal motor signs of PD at the time of diagnosis, including tremor, bradykinesia, rigidity, and postural instability, were recorded. The following nonmotor features were also included: (1) autonomic dysfunction defined by the presence of any 2 of the following symptoms or signs persistent for longer than 6 months (urinary symptoms, constipation, symptoms of upper gastrointestinal tract dysfunction, orthostatic hypotension, sweating abnormalities, or erectile dysfunction in men),¹³ (2) clinically diagnosed RBD defined as presence of repeated episodes of sleep-related vocalization and/or complex motor behavior, (3) cognitive dysfunction (graded according to the degree of impairment of tasks of daily living from mild cognitive impairment to severe dementia), and (4) symptoms of major depressive disorder. Grading of symptom severity was performed blinded to disease course and neuropathologic data, based on the clinical relevance of the symptoms on the premise that this would be well documented in the clinical notes. Using a 4-point semiquantitative scale based on the clinical impression by the treating physician, symptoms and signs were graded as follows: (0) absent; (1) mild symptom severity/mild distress to patient/no therapeutic intervention required; (2) moderate symptom severity/moderate distress to patient/good symptomatic control with therapeutic intervention; and (3) severe intensity/severe distress to patient/poor symptomatic control despite therapeutic intervention.

Quiz Ref IDTo assess disease progression, time of onset from diagnosis to specific disease milestones were recorded^14,15: (1) regular falls or (2) wheelchair dependence as milestones of motor disability; (3) dementia, defined as cognitive impairment severe enough to significantly affect tasks of daily living; and (4) placement in residential or nursing home care as a measure of global disability. Time from diagnosis to death (survival) was also recorded.

Definitions proposed by Fereshtehnejad et al⁵ were adapted to our retrospective, nonstandardized clinical data so that our subtypes corresponded to their criteria as closely as possible. Our semiquantitative severity scores of motor and nonmotor features were converted into 4 domains that included 1 composite for motor (a composite of the individual scores for tremor, bradykinesia, rigidity, and postural instability) and 1 composite for autonomic dysfunction (a composite of the individual scores for urinary symptoms, constipation, symptoms of upper gastrointestinal tract dysfunction, orthostatic hypotension, sweating abnormalities, and erectile dysfunction in men) and 2 individual scores for RBD and cognitive dysfunction. The corresponding 75th percentile was calculated for each domain, and patients were classified at the time of diagnosis into 3 subtypes:

• Mild-motor predominant. Motor and all nonmotor scores less than the 75th percentile.

• Diffuse malignant. Either motor score greater than the 75th percentile and at least 1 nonmotor score greater than the 75th percentile or all 3 nonmotor scores greater than the 75th percentile.

• Intermediate. All those individuals not meeting criteria for other subtypes.

As a consequence of adapting the original subtyping system using quantitative assessment tools to a semiquantitative scale with limited categories, groups of patients could have the same score corresponding to the 75th percentile. All those patients with the same score were allocated to the same group (either >75th or <75th percentile group) based on where most of them distributed. For example, the 75th percentile for RBD score was 1 = mild, and 2 patients with this score fell into the less than 75th percentile group while 22 fell into the greater than 75th percentile group; therefore, all patients with RBD score mild were allocated to the greater than 75th percentile group.

Formalin-fixed brain tissue samples were examined with immunohistochemistry using standard protocols. Representative sections of brainstem, basal forebrain, and neocortical areas were graded for distribution and severity of Lewy pathology, and each individual was assigned to a Lewy body subtype (brainstem, limbic, or diffuse neocortical) and a Braak stage (stage 1-6) based on Lewy body scores.^12,16,17 Severity of neurofibrillary tangle pathology was assessed using specific immunohistochemistry against hyperphosphorylated tau protein and the staging system described by Braak and Braak.^18,19 Amyloid β immunostaining was used for the assessment of amyloid β deposition, and neuritic plaques and their severity and extent were graded based on the classifications proposed by Thal et al²⁰ and the Consortium to Establish a Registry for Alzheimer’s Disease protocol.²¹ Global Alzheimer disease neuropathologic changes were assessed using the ABC scoring system proposed by the National Institute on Aging–Alzheimer’s Association guidelines (absent, low, intermediate, and high) based on the combination of neurofibrillary tangle, amyloid β deposition, and neuritic plaque severity.²²

Comparisons between groups were performed using Fisher exact test for categorical variables, and analysis of variance (ANOVA) for continuous variables as appropriate, with Bonferroni correction for multiple pairwise comparisons. Ordinal neuropathology stages were analyzed using the Kruskal-Wallis test.

Multivariable Cox proportional hazard regression models were used to estimate the risk of disease milestones and death for each Parkinson subtype. Other potential confounders, including age at diagnosis, sex, levodopa response, and depression severity, were included in the model as explanatory variables. Only those with a relevant association in the multivariate analysis were included in the final model. Adjusted hazard ratios (HRs) and 95% confidence intervals were estimated. Kaplan-Meier curves of disease milestones and survival were plotted for each Parkinson subtype. Visual inspection of Kaplan-Meier curves and plots of scaled Schoenfeld residuals against time were used to assess the proportional hazards assumption. Censoring was considered uninformative. Two-tailed tests were performed, statistical significance was set at P < .05, and Stata statistical software, version 12 (StataCorp), was used for statistical analysis.

One hundred eleven patients (67 men [60.4%]), with mean (SD) age at diagnosis of 62.5 (11.5) years, were included in the final analysis (eFigure 1 in the Supplement). The demographic data and clinical features of the whole group and by Parkinson subtypes are shown in Table 1.²³ The QSBB cohort did not differ in age at diagnosis (mean [SD], 62.5 [11.5] years vs 61.1 [9.7] years; t = −1.299; P = .20), sex (male sex, n = 67 of 111 [60.4%] vs n = 276 of 421 [65.6%]; χ² = 1.04; P = .31), or PD subtype distribution (mild-motor predominant/intermediate/diffuse malignant: 48.7%, 35.1%, and 16.2%, respectively, vs 52.5%, 38.5%, and 9.0%, respectively; χ² = 4.83; P = .09) compared with the reference Parkinson’s Progression Markers Initiative group. Patients with the mild-motor–predominant subtype patients were significantly younger at diagnosis, had a better response to levodopa, and received a higher levodopa equivalent dose. Patients with the diffuse-malignant subtype were older, were almost all men, had a poorer response to levodopa, and were more frequently misdiagnosed as having an atypical parkinsonian syndrome in life (Table 1 for all comparisons).

Quiz Ref IDThe PD subtypes showed different rates of deterioration, with the diffuse malignant subtype reaching all prognostic milestones earlier in the disease course and having the shortest survival (ANOVA; Table 1 for all comparisons; eFigure 2 in the Supplement). Non–PD-related causes of death were similar between groups. Despite differences in disease progression and survival, PD subtypes did not differ in disability at the time of death as judged by the presence of disease milestones, and all subtypes reached advanced stages, with 105 of 111 patients (94.6%) reaching at least 1 disease milestone (Table 1).

Only PD subtype and age at diagnosis remained significant in the multivariate model (Table 2). The diffuse-malignant group showed an increased risk of development of all disease milestones compared with mild-motor predominant. The probability of survival in the diffuse-malignant group was also reduced, with an adjusted HR of 3.65 (95% CI, 1.98-6.75; P < .001) (Table 2 and Figure 1).

Lewy pathology did not show significant differences across PD subtypes measured by Braak and Lewy pathology staging (Kruskal-Wallis test; Table 3 for specific values). The severity of comorbid Alzheimer disease–related neuropathologic changes was also similar among different PD subtypes measured using several accepted staging systems (Table 3).

To elucidate potential pathogenic mechanisms, we analyzed the association between neuropathologic findings and other demographic characteristics. Neurofibrillary tangle pathology (ANOVA; F_6,104 = 2.84; P = .01), Consortium to Establish a Registry for Alzheimer’s Disease neuritic plaques (ANOVA; F_3,107 = 3.56; P = .02), and Alzheimer disease–related pathology (ANOVA; F_3,107 = 4.12; P = .008) showed significant association with increased age at death. Association of neocortical Lewy pathology and disease duration did not reach statistical significance (ANOVA; F_1,109 = 3.64; P = .06).

This large, retrospective study of pathology-confirmed patients with PD provides, for the first time to our knowledge, entire life-course data on new Parkinson subtypes that incorporate nonmotor symptoms into their classification and also permits neuropathologic correlation. Our results showed that clinical subtyping at diagnosis accurately estimates disease progression and survival. Different rates of neuropathologic progression appear to be important determinants of clinical subtypes.

Multiple subtyping classification systems have been proposed.^3,24 We favored the scheme proposed by Fereshtehnejad et al⁵ because of its potential applicability in clinical practice, using an algorithm that translates groups to an individual level and the presence of imaging and biomarker supporting data.⁵ We adapted the classification to a semiquantitative grading assessment using nonstandardized collected clinical data, with similar representation of PD subtypes. Age at diagnosis was not included in this classification system, although it has been considered a key determinant in PD subtypes by others.²⁵ Age at diagnosis differed significantly among PD subtypes in our study, was an important independent factor associated with faster disease progression and reduced survival in the multivariate regression model, and was associated with comorbid pathologies. Based on our findings, we propose that age at diagnosis should be included in any PD subtype definition.

Very few studies have assessed the validity of Parkinson subtypes longitudinally,^4,9,26 and the short follow-up periods limit making firm conclusions. To our knowledge, this study is the first to provide life-course data on disease progression and shows that clinical subtyping estimates the development of clinically relevant motor, cognitive, and global disability disease milestones and reduced survival. It is worth emphasizing that these results were independent of other potential confounders and despite differences in age at the time of diagnosis.

Our study provided postmortem confirmation of the diagnosis, as opposed to previous PD subtype studies that may have an inherent diagnostic accuracy bias, particularly at early stages of the disease.^27,28 Most patients in our study reached advanced stages (Figure 2), and PD subtypes did not show any differences in the severity or distribution of Lewy pathology. Only one previous study, using a separate group of patients from the Queen Square Brain Bank, attempted to correlate postmortem findings with Parkinson subtypes, although direct comparisons cannot be drawn because the subtype classification did not include nonmotor features.¹⁰ In this previous study, clinical subtypes did not show a clear correlation with neuropathologic stages, although patients with nontremor subtype had more severe neocortical Lewy pathology.¹⁰ Previous clinicopathologic evidence suggests nonlinear progression in the late stages of the disease, with patients reaching a common pathologic end point.^14,29 Because most of the patients were at this pathologic end point, if there are topographic differences in Lewy pathology responsible for the clinical heterogeneity during life disease course, they are no longer detectable at post mortem. However, our results suggest that the speed of the pathologic process varies between subtypes because a similar pathologic end point is reached in all of them despite differing disease duration. Similar conclusions were reached by Kempster et al¹⁴ in previous clinicopathologic studies when correlating clinical progression with neuropathologic findings in patients with PD divided in groups by age at death¹⁴ and patterns of levodopa response.²⁹ Taken together, the results show that despite a heterogeneous disease course during early-middle stages, advanced stages of the disease are clinically very similar, with accumulation of disability in a similar time course followed by death without any differences on neuropathologic findings at post mortem (Figure 3). A nonlinear progression may govern the terminal stages of the neurodegenerative process, and age and other comorbid pathologies may play a synergistic effect.^14,29 Analysis of comorbid pathology in this study’s cases showed no significant differences in Alzheimer disease–related neuropathology among PD subtypes, which suggests a more rapid deposition in the group with shorter disease duration (diffuse malignant), although age seems to be an additional important determinant, as discussed in subsequent paragraphs.

Further analysis revealed that in addition to PD subtype, other demographic factors may influence neuropathologic progression. Quiz Ref IDIn this group of patients, while Lewy body pathology seems to be a primary disease-related process because neocortical Lewy stages were associated with disease duration, Alzheimer disease–related pathology showed a significant association with age at death (Figure 2B). Halliday et al³⁰ evaluated pathologically distinct Parkinson subgroups and correlated them with clinical findings. Three different patterns of pathologic progression were identified: a group with Lewy pathology progressing anatomically as described by Braak,¹⁶ who presented with young-onset and long-disease duration, a second group with early severe neocortical Lewy pathology presenting with dementia, and a third group with marked cortical Lewy pathology associated with Alzheimer disease–related and vascular changes with older onset and complex disease course. The analysis of the overall pathologic findings in our entire cohort of patients (which excluded patients with a presentation of Lewy body dementia) showed that some of the factors affecting the rate of progression are in agreement with the Halliday et al study.³⁰

Combining our findings with results from previous clinicopathologic studies,^14,29-31 several conclusions with regards to the mechanisms underlying the clinical and pathological variability in PD can be made (Figure 2).

• Young patients presenting with a more classic Lewy body phenotype with mild motor and nonmotor symptoms (mild-motor predominant) show a slow disease course and neuropathologic progression before they reach the terminal stages of the disease. Disease duration seems to be the main determinant of clinical and pathologic progression in this subgroup.

• Additional factors are likely to contribute to the more rapid spread of Lewy pathology throughout the brain in diffuse malignant subtypes. Moreover, other age-related pathologies contribute to the clinical picture in older patients with a more complex disease course (corresponding to the diffuse malignant subgroup) and more rapid clinical and pathologic progression, with a potential synergistic effect between comorbid pathologies.

• There may also be topographic differences in Lewy pathology distribution at the early stages of the disease, with those with an initial cognitive presentation consistent with Lewy body dementia showing early involvement of cortical areas,³⁰ although these patients were not included in either our study or the Parkinson’s Progression Markers Initiative cohort.

The main strengths of this study are the length of longitudinal regular follow-up and the detailed neuropathologic findings. The retrospective nature of the study and the assessment by different professionals without clear methodologic homogeneity (including the lack of validated questionnaires or investigations such as cardiovascular autonomic function tests or polysomnography) may potentially account for some limitations in the accuracy of the recording of the symptoms. To limit this potential bias, strict exclusion criteria were used, and only the most recent cases with detailed and regular clinical information from PD hospital specialists were selected. Despite this and the fact that clinical information at the time of diagnosis was documented in detail for all included cases, we acknowledge that some of the symptoms, particularly mild nonmotor symptoms without significant effect on clinical practice, may have been underreported. As discussed in previous paragraphs, the advanced clinical and pathologic stages of our patients may have limited detection of any potential neuropathologic topographic differences in earlier stages of the disease.

Clinical subtyping of PD based on motor symptoms, RBD, and autonomic and cognitive dysfunction at diagnosis is feasible in clinical practice and provides accurate long-term estimation of disease progression and survival. Different pathologies with differing rates of progression are important determinants of clinical subtypes, and age at diagnosis should be included in future subtype classification systems.

Corresponding Author: Thomas T. Warner, MD, PhD, Reta Lila Weston Institute of Neurological Studies, University College London Queen Square Institute of Neurology, 1 Wakefield St, London WC1N 1PJ, England (t.warner@ucl.ac.uk).

Accepted for Publication: November 10, 2018.

Published Online: January 14, 2019. doi:10.1001/jamaneurol.2018.4377

Author Contributions: Drs De Pablo-Fernández and Warner had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.

Concept and design: De Pablo-Fernández, Lees, Warner.

Acquisition, analysis, or interpretation of data: De Pablo-Fernández, Holton.

Drafting of the manuscript: De Pablo-Fernández, Lees.

Critical revision of the manuscript for important intellectual content: De Pablo-Fernández, Holton, Warner.

Statistical analysis: De Pablo-Fernández.

Obtained funding: Warner.

Supervision: Lees, Holton, Warner.

Conflict of Interest Disclosures: Dr Lees is a consultant for Genus; has received research support from PSP Association, Weston Trust, and the Reta Lila Howard Foundation; and has received honoraria from Roche, Novartis, Boehringer Ingelheim, Lundbeck, GE Healthcare, Servier, Teva, Solvay, GlaxoSmithKline, Ipsen, Allergan, Orion, Bial, and Abbvie Lucid. Dr Holton is supported by the Multiple System Atrophy Trust; the Multiple System Atrophy Coalition; Fund Sophia, managed by the King Baudouin Foundation; Alzheimer’s Research UK, and CBD Solutions. Dr Warner receives research support from the Reta Lila Weston Medical Trust, Brain Research Trust, Cure Huntington’s Disease Initiative, Medical Research Council, and Corticobasal Degeneration Solutions. No other disclosures were reported.

Funding/Support: Queen Square Brain Bank is supported by the Reta Lila Weston Institute for Neurological Studies, the Progressive Supranuclear Palsy Association, and the Medical Research Council UK. This research was supported by the National Institute for Health Research University College London Hospitals Biomedical Research Centre.

Role of the Funder/Sponsor: The funding sources had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.

Additional Contributions: We thank the patients and their families, without whose generous donation and support none of this research would have been possible. No compensation from a funding source was received.