Specialists Agree on the Core Signs, but Not on the Edges of Atypical Parkinsonism
Accurate antemortem diagnosis of atypical parkinsonian disorders remains one of movement disorders medicine’s most difficult tasks. Dementia with Lewy bodies, multiple system atrophy, progressive supranuclear palsy, and corticobasal degeneration can present with overlapping motor, cognitive, autonomic, and ocular features, and the preprint examines how specialists actually sort those features in day-to-day diagnostic reasoning.
The study did not test a treatment or a biomarker. It examined judgment itself.
Investigators surveyed movement disorders specialists on the diagnostic implications of 32 clinical features and compared those responses with a 2024 consensus algorithmic framework, the “General Neurologist’s Practical Diagnostic Algorithm for APDs.” The aim was not simply to ask whether clinicians know the criteria, but whether the published framework matches the way experts operationalize the same features when faced with an individual patient.
High Agreement for Archetypal Features, Less Certainty for Borderline Ones
The strongest signal in the survey was consistency. Respondents showed high inter-specialist agreement in feature attribution, with raw agreement of 81.9% and Gwet’s AC1 of 0.67, with a 95% confidence interval of 0.64 to 0.70. In a field where the same symptom can point in several directions depending on its timing, symmetry, and associated exam findings, that level of concordance suggests specialists share a stable internal map for the most recognizable clues.
Even so, the study makes plain where consensus weakens. Feature endorsement rose sharply as the framework’s association strength increased, moving from 12.9% for unmapped features to 36.2% for modest associations and 82.6% for moderate associations. In other words, when a sign is clearly linked to a disorder, specialists tend to agree. When the association is weaker or shared across disorders, the diagnostic signal becomes less uniform.
The point matters at the bedside. A patient with a mixed syndrome rarely arrives with a tidy textbook pattern, and clinicians must constantly decide whether a feature is essential, supportive, or merely compatible. The survey suggests that the published framework captures the strongest associations well, but its border zones remain subject to personal weighting.
Coverage Was Better for MSA, PSP, and CBD Than for DLB
Diagnosis-level coverage varied by disorder. The framework covered multiple system atrophy, progressive supranuclear palsy, and corticobasal degeneration more completely than dementia with Lewy bodies. That asymmetry likely reflects the way these syndromes are often recognized in clinic: MSA, PSP, and CBD can present with more syndrome-defining clusters, while DLB more often overlaps with other parkinsonian and cognitive phenotypes.
In practical terms, a framework with better coverage for some disorders than others may influence how confidently clinicians move from suspicion to working diagnosis. A clinician can be reasonably sure that vertical gaze palsy or a strongly suggestive autonomic phenotype should push the differential in a specific direction. By contrast, DLB often lives in the diagnostic gray zone where cognitive fluctuation, hallucinations, parkinsonism, and sleep-related features may not arrive in a clean sequence.
Exact-set agreement was more variable than the broader feature-attribution results and depended on whether modest associations were included in the framework definition. When a framework is forced to match an exact combination of findings, even small differences in feature weighting become decisive. When modest associations are admitted, the shared diagnostic space widens.
A Unified Framework Could Help, but Only if It Mirrors Clinical Reasoning
The survey found that respondents, despite reporting strong familiarity with and use of published diagnostic criteria, overwhelmingly wanted a unified diagnostic framework for APDs paired with automated clinical decision support. That finding is easy to understand from a clinical workflow standpoint. Movement disorders specialists already think in pattern-recognition terms, but the transition from intuition to standardized decision support requires a framework that reflects how they actually interpret mixed findings.
Automated clinical decision support will not solve the biology of these disorders, and it will not create a biomarker where one does not exist. What it can do is organize uncertainty, particularly when a patient sits between categories or presents with a partial syndrome early in the disease course. In a domain where the cost of misclassification is delayed counseling, distorted prognostic conversations, and possible misdirection of workup, a usable algorithm could tighten clinical reasoning without replacing it.
Still, the study also shows the limits of standardization. Specialist agreement was strongest around archetypal features and less consistent around weaker or overlapping associations. A rigid decision tool that treats every feature as equally weighted would risk flattening exactly the nuance that experts rely on. A more useful system would likely need to preserve graded associations rather than forcing binary yes-or-no rules.
What the Survey Says About Diagnostic Medicine in APDs
This preprint speaks to a familiar problem in neurology: the difference between published criteria and real-world cognition. Consensus criteria can be carefully constructed and still fail to capture how expert clinicians sort ambiguous signs at the bedside. The results here suggest that specialists share a strong conceptual core, but they diverge at the margins where atypical parkinsonian disorders are hardest to separate.
The authors’ comparison with the 2024 algorithmic framework therefore goes beyond a simple validation exercise. It exposes where a diagnostic schema aligns with specialist practice and where it may need revision. The gap is not trivial, because those modest associations may be exactly the findings that determine whether a patient is classified as having one syndrome rather than another.
The cohort was survey-based, and the study did not include patient-level outcomes, imaging, pathology, or longitudinal confirmation. It also reflects the reasoning of movement disorders specialists rather than general neurologists or other clinicians who first evaluate these patients. The preprint is explicitly not peer reviewed, and its findings should be read as hypothesis-generating rather than practice-changing.
Even so, the message is direct: expert consensus is strong for the classic signs of atypical parkinsonian disorders, but the weighting of weaker or overlapping features remains unsettled, and any future diagnostic support tool will need to account for that uneven terrain.