Do primary and supplementary measures of semantic memory predict cognitive decline by patients with Alzheimer's disease?

Although the average performances of large groups of patients with probable Alzheimer's disease (AD) exhibit steady deterioration in cognition, there are substantial differences in rates of decline for individual patients. Several previous studies suggest that poor initial performance on measures of language-semantic memory may forecast imminent cognitive deterioration, but the findings are inconsistent. In the present study, we examined the ability of several primary (total correct on naming and verbal fluency) and supplementary (e.g., cluster size, switches, error rate) measures of semantic memory to distinguish between deteriorated patients and stable patients (Total N = 158). Deterioration was defined by the magnitude of decline on the Dementia Rating Scale (DRS) over I year. Regardless of the cutoff criterion (5, 7, 10 or 15 points), logistical regression revealed that only age and score on the Boston Naming Test (BNT), and no other measures of semantic memory included in the study, predicted deterioration status accurately. Linear regression analysis showed that the best combination of predictors (Age, BNT) accounted for less than 7% of the variance in change in DRS scores. None of the measures of semantic memory predicted which patients, with an initial DRS score of at least I 10, would decline to a score of less than 100. Although measures of semantic memory are sensitive to differences between patients with AD and healthy elderly controls, they do not predict cognitive decline among individual patients.