Questionable dementia: clinical course and predictors of outcome.

OBJECTIVE: To evaluate the clinical course and predictors of outcome in outpatients with cognitive impairment who do not meet criteria for dementia at initial evaluation.

DESIGN: Naturalistic longitudinal study.

METHODS: Cognitively impaired patients in a memory disorders clinic who fell between the "normal" and "dementia" categories were defined broadly as "questionable dementia" (QD). Of 127 consecutive QD patients, 75 were followed for a minimum of 1 year (mean 2.5 years, SD 1.7). Baseline neuropsychological testing was conducted in 62 of these 75 QD patients.

RESULTS: At the final follow-up time-point, 41.3% met diagnostic criteria for dementia (27 of 31 patients with dementia had possible or probable Alzheimer's Disease, AD), 44% were rated as not demented, and 14.7% remained as "uncertain" dementia. Increased age was associated with the final diagnosis of dementia, but duration of follow-up, Clinical Dementia Rating, and modified Mini Mental State (mMMS) scores were not predictive. Low scores on the mMMS delayed recall subtest, consistent long-term retrieval on the Selective Reminding Test, category naming for animals, and the WAIS-R digit symbol, picture arrangement, and block design subtests were predictive of the final diagnosis of dementia (all P < or = .01). mMMS delayed recall showed 66.7% sensitivity and 71.4% specificity, the other five neuropsychological subtests together showed 66.7% sensitivity and 66.7% specificity, and the six tests together showed 81% sensitivity and 76.9% specificity. Similar predictive accuracy was obtained for the final diagnosis of AD.

CONCLUSIONS: In QD patients, poor performance on the mMMS delayed recall item may be a useful predictor of the diagnosis of dementia (and AD) on follow-up. Combining a screening instrument like the mMMS with specific neuropsychological tests may provide good predictive accuracy. In QD patients, the observed heterogeneity in diagnostic outcome, with most patients in the "dementia" and "no dementia" categories at follow-up, enhances the feasibility of evaluating early markers with predictive accuracy for dementia and AD.