Accuracy of four clinical diagnostic criteria for the diagnosis of neurodegenerative dementias.

To evaluate the inter-rater reliability and validity of clinical diagnostic criteria for neurodegenerative dementias. Inter-rater accuracy of the diagnosis of AD has been explored, but there are few accuracy studies for progressive supranuclear palsy (PSP) and frontotemporal lobe dementia (FTD). Furthermore, there have been no simultaneous accuracy studies in a mixed sample of patients with cortical and subcortical neurodegenerative processes. Four experienced clinicians reviewed first-visit clinical data abstracted from the records of 40 pathologically diagnosed demented subjects. They were asked to apply the NINCDS-ADRDA criteria for AD, the NINDS-SPSP clinical criteria for PSP, the Lund and Manchester criteria for FTD, and the Consensus Guidelines for the Clinical Diagnosis of Dementia with Lewy Bodies (DLB). The generalized K for AD was 0.73, for PSP 0.82, for FTD 0.75, and for DLB 0.37. The K pool test showed a statistically significant difference between DLB and the other disease processes, and no differences were observed among AD, FTD, and PSP. The mean sensitivity for AD was 95%, for PSP 75%, for FTD 97%, and for DLB 34%. The mean specificity for AD was 79%, for PSP 98.5%, for FTD 97%, and for DLB 94%. We found improved inter-rater reliability for the diagnosis of AD among clinicians compared with earlier studies. Similarly, there was a near-perfect and substantial inter-rater agreement for the diagnosis of PSP and FTD. The sensitivity for the diagnosis of AD was high, although clinicians overdiagnosed this condition. However, there was a reasonable accuracy for the diagnosis of PSP and FTD. Heterogeneity of the clinical presentation of DLB significantly affected inter-rater agreement and accuracy. The use of multiple diagnostic criteria for cortical and subcortical dementia increases the level of clinical diagnostic accuracy.