Validating the Accuracy of Parkinson's Disease Clinical Diagnosis: A UK Brain Bank Case-Control Study.

objectiveDespite diagnostic criteria refinements, Parkinson's disease (PD) clinical diagnosis still suffers from a not satisfying accuracy, with the post-mortem examination as the gold standard for diagnosis. Seminal clinicopathological series highlighted that a relevant number of patients alive-diagnosed with idiopathic PD have an alternative post-mortem diagnosis. We evaluated the diagnostic accuracy of PD comparing the in-vivo clinical diagnosis with the post-mortem diagnosis performed through the pathological examination in 2 groups.

methodsIn this retrospective case-control study, patients and healthy subjects who consented to the post-mortem pathological diagnosis at the UK Brain Bank were consecutively enrolled from the UK Brain Bank. Medical records were reviewed to classify participants and performance metrics were further calculated using neuropathological diagnosis as the gold standard.

resultsFour thousand five hundred seventy one subjects were eligible for the study. The clinical diagnosis group was: 1,048 Parkinson's patients and 1,242 healthy subjects. Pathology diagnosis group were: 996 Parkinson's patients and 1,288 subjects with no post-mortem abnormality. For the group of clinical diagnosis, PD diagnosis showed: sensitivity of 99%, specificity of 86%, accuracy of 90.96%, F1-Score 0.89, and a receiver operating characteristics area under the curve (ROC AUC) 0.925 (SE ± 0.006) [95% confidence interval [CI]: 0.913, 0.937], 𝑝<0.001. In this group, the most frequent pathology diagnosis among clinically misdiagnosed PD (false positive) patients was dementia with Lewy bodies (19.4%). Conversely, the most frequent clinical diagnosis among PD missed clinical diagnosis (false negative) patients was Alzheimer's disease (18.5%).

interpretationOur findings confirm a still significant diagnostic error and emphasize the need for more fine and homogeneous criteria to classify idiopathic Parkinson's patients correctly. ANN NEUROL 2025;97:1110-1121.