Abstracts

RATIONALE: Prediction of changes in verbal memory after epilepsy surgery is important, but prediction formulas are not commonly used because of the complexities which are typically involed. The objective of this paper is to present an easily usable method for predicting memory change after epilepsy surgery.
METHODS: Patients were 48 adults (27 women, 21 men) with intractible partial seizures all of whom received resection surgery for their seizures which involved either the left temporal lobe (n=29) or the right temporal lobe (n=19). Variables considered as predictors of memory loss included age (M 31.96, SD 11.07), age at onset of epilepsy (M 14.90, SD 12.02), level of preoperative verbal memory (sum of WMS Form 1 Logical Memory immediate and delayed, Paired Associates easy and hard immediate and delayed plus Rey AVLT sum of 9 trials; median split into HIGH and LOW groups), mesial temporal sclerosis (PRESENT vs. ABSENT by high resolution MRI), surgery side vs. side of speech as shown by the Wada or intracarotid amobarbital test (SAME vs. DIFFERENT), Wada memory (three memory assessment techniques considered as detailed in Brain and Cognition 1997;33:210-233), and extent of resection (lateral, mesial, hippocampal). All patients received post-operative memory evaluations 11-53 months after surgery (M 21.85, SD 8.40) and the percent of change in the overall verbal memory score was calculated. Predictive variables were examined individually, and those which made a significant contribution to prediction of memory change after surgery were placed in a stepwise multiple regression paradigm. The results of these variables were then presented in a simplied flow diagram to show memory change.
RESULTS: Variables demonstrating statistically significant independent relationships with verbal memory change after surgery were as follows (in order of most to least potent): 1) preoperative verbal memory level (HIGH level predicted loss); 2) side of surgery vs. side of speech (SAME side predicted loss); 3) hippocampal sclerosis on MRI (ABSENT predicted loss). The other variables did not predict memory change after surgery although Wada memory (Seattle procedure) would have been a significant predictor had not the side of surgery vs. side of speech variable been included. Percent changes in verbal memory were as follows for each combination of the three predictors, respectively: LOW, OPPOSITE, PRESENT +32%; LOW, OPPOSITE, ABSENT, +14%; LOW, SAME, PRESENT -9%; LOW, SAME, ABSENT -7%; HIGH, OPPOSITE, PRESENT +1%; HIGH, OPPOSITE, ABSENT -12%; HIGH, SAME, PRESENT -11%; HIGH, SAME, ABSENT -31%.
CONCLUSIONS: Preoperative verbal memory level, whether or not surgery was on the same side as speech, and presence or absence of mesial temporal sclerosis were the most potent variables in predicting changes in verbal memory after surgery. These variables are easily obtained and when they are considered in a dichotomous fashion, they predict changes in verbal memory with a considerable degree of accuracy.