Recovery of cognitive function after stroke.

Abstract

Previous studies have suggested that recovery of cognitive function after stroke is maximal within the first 3 months after onset. We performed the present study to investigate the long-term course and clinical correlates of improvement in generalized cognitive function after ischemic stroke. We administered a battery of neuropsychological tests to 151 patients (age, 70.4 +/- 7.7 years; education, 10.4 +/- 4.6 years) at 3 months and then annually after stroke. We transformed their test results into z scores based on the performance of a stroke-free normative group, averaged those scores to create a summary score, and defined improvement in annual examinations as an increase in that summary score greater than two standard deviations above the mean first annual change of the normative group. We then used logistic regression to determine whether stroke location, syndrome, or recurrence; vascular risk factors; dementia status; depression; or demographic variables were associated with improvement. We found that 19 of the 151 patients exhibited improvement, which was evident only at the first annual examination in most cases. Logistic regression determined that improvement was significantly related to left hemisphere infarction relative to brain stem/cerebellar infarction (odds ratio [OR], 5.57), while the presence of a major hemispheral stroke syndrome showed a trend toward significance (OR, 3.32). Diabetes mellitus was significantly associated with a failure to exhibit improvement (OR, 0.12). Based on the logistic model, the probability of long-term improvement would be 54.0% for a patient with a left hemisphere infarct and a major hemispheral syndrome but only 11.9% if diabetes was also present. Long-term improvement in generalized cognitive function may be evident after stroke in association with left hemisphere infarction and severe hemispheral syndromes, while it may be compromised by diabetes, possibly because of an increased burden of cerebrovascular disease.