Clinical trials in dementia: learning effects with repeated testing.

The possible confounding role of learning effects during multiple test administration in clinical trials in demented subjects remains uncertain. Seventeen mildly to severely affected patients with Alzheimer's disease (AD) and 16 controls were evaluated with an extensive neuropsychological battery of manually administered and computerized tests. Subjects received 3 weekly sessions using alternate test forms, to mimic a baseline, placebo, and drug condition. Mean scores of AD patients as a group showed no stable improvement, but more subtle learning effects were suggested by an association of dementia severity and change scores in verbal and visual learning. Controls evidenced consistent learning in 3 of 9 tests. These results suggest that clinical trials data from AD patients, especially those with moderately severe impairment, are not contaminated by learning. Degree of cognitive impairment may be related to learning capacity, suggesting caution for learning effects in controls and possibly in patients with mild AD.