Diagnosis of dementia: clinicopathologic correlations.

Based on 54 demented patients consecutively autopsied at the University of Pittsburgh, we studied the accuracy of clinicians in predicting the pathologic diagnosis. Thirty-nine patients (72.2%) had Alzheimer's disease, while 15 (27.7%) had other CNS diseases (four multi-infarct dementia; three Creutzfeldt-Jakob disease; two thalamic and subcortical gliosis; three Parkinson's disease; one progressive supranuclear palsy; one Huntington's disease; and one unclassified). Two neurologists independently reviewed the clinical records of each patient without knowledge of the patient's identity or clinical or pathologic diagnoses; each clinician reached a clinical diagnosis based on criteria derived from those of the NINCDS/ADRDA. In 34 (63%) cases both clinicians were correct, in nine (17%) one was correct, and in 11 (20%) neither was correct. These results show that in patients with a clinical diagnosis of dementia, the etiology cannot be accurately predicted during life.