Refining the Diagnostic Accuracy of Parkinsonian Disorders Using Metaphenomic Annotation of the Clinicopathological Literature

AbstractBackgroundThe diagnostic precision of Parkinsonian disorders is not accurate enough. Even in expert clinics up to one in five diagnoses are incorrect. This leads to cohorts with mixed pathologies, impacting our ability to understand disease heterogeneity and posing a major challenge for clinical trials. Gold standard diagnosis is post-mortem confirmation of the underlying proteinopathy, however many clinicopathological studies focus on either a single disease or frame analyses in one temporal direction (i.e., in-life diagnosis vs post-mortem or vice versa). Given Parkinson’s Disease (PD), Multiple System Atrophy (MSA), Progressive Supranuclear Gaze Palsy (PSP), Dementia with Lewy Bodies (DLB) and Corticobasal degeneration (CBD) can all mimic one-another, these may underestimate mis- and missed diagnoses.MethodsThe objective was to comprehensively map the mis- and missed diagnoses across the Parkinsonian disorders and use phenotypic features to develop a probabilistic model to refine diagnostic likelihoods based on clinical observations. We identified 125 published clinicopathological cohorts and case-reports since 1992, extracted phenotype information for ∼9200 post-mortem cases, and curated the data in a standardized machine-readable format.FindingsMSA diagnostic accuracy was highest (92·8%) and DLB lowest (82·1%). MSA and PSP were most frequently mis-labelled as PD in life (7·2% and 8·3% of cases), where-as the most common PD misdiagnosis was Alzheimer’s (∼7% cases). DLB age at diagnosis was older, CBD younger, and survival longer in PD. Clinical annotation was extremely variable, which represents a limitation with clinicopathological literature, however we created likelihood ratios for a range of features and demonstrate how these can refine diagnoses.InterpretationThis work delivers a harmonized, open-source dataset representing over 30 years of published results and represents a key foundation for more flexible predictive models that leverage different sources of information to better discriminate Parkinsonian disorders during the early and prodromal phases of the illness.FundingMedical Research CouncilResearch in contextEvidence before this studyThe diagnostic precision of Parkinsonian disorders is not accurate enough – estimated misdiagnosis rates, derived from clinicopathological studies, vary between 10 – 20% depending on the condition, context and criteria. However, many previous studies either focus on one single condition, or frame the analysis in one temporal direction. By the time Parkinsonian disorders manifest with motor symptoms, the conditions have been present for 10-20y. Previous work has proposed a probabilistic approach to identify prodromal Parkinson’s disease, but none exist for the range of common Parkinsonian disorders that often mimic one another.Added value of this studyThis study structures and standardises 30-years of clinicopathological data across all the main Parkinsonian syndromes, making it available in an open, machine-readable format, and also updates the Human Phenotyping Ontology for Parkinsonian syndromes. It uses these to comprehensively map the patterns of missed and mis-diagnosis across all of the conditions, and build a flexible multimodal probabilistic approach to help refine diagnoses of these disorders.Implications of all the available evidenceThis work provides a key foundation for a modular framework that can be flexibly adapted and combined with different tools, techniques and approaches to more accurately diagnose different Parkinsonian disorders during the early and prodromal phases of the illness.