The effects on cognition of carbamazepine, phenytoin and sodium valproate were compared in 64 new cases of childhood epilepsy treated at the Leeds General Infirmary, England. The children age 5-14 years had no neurological deficit and were allocated randomly and assessed prospectively over a 12 month period. They were seizure-free throughout the 12 months. Psychological tests included visual recall, auditory recall (memory), visual scanning (vigilance), Stroop test (concentration), speed of information processing, and standard scales of intellectual functioning and reading. Serum levels of carbamazepine (CBZ) showed a significant negative correlation with changes in scores on the sum of five memory tests from before medication to 6 months and a year later. No significant correlations with memory scores were found for children treated with phenytoin (PHY). Only 10 of 126 serum level estimations of CBZ gave a value greater than 8.2 mcg/L. In the sodium valproate (SV) group significant positive correlations were found between serum levels and the sum of five memory tests at 1 month and at 6 months after starting treatment. The memory test clearly showed that there was impaired recent recall on CBZ apparent by 6 months of treatment and even more definite after a year. Reading scores were also lower in the CBZ compared with the PHY group after a year on drug treatment. Doses of CBZ were usually in the lower half of the accepted therapeutic range of 5-15 mcg/L. 
COMMENT. In this study involving children with epilepsy previously untreated, carbamazepine in moderate dosage adversely affected memory but sodium valproate and phenytoin did not. The tendency to favor carbamazepine in preference to phenytoin on the basis of reported cognitive deficits seems questionable and unfounded. A phenytoin-induced improvement in auditory memory has been demonstrated in children with EEG dysrhythmias.