To the Editor.
—Dr Scolnik and colleagues1 report a "clinically important negative effect of phenytoin on neurobehavioral development" and "recommend that whenever clinically possible, switching therapy to carbamazepine should be considered." However, we question whether their experimental design and data analysis support such a strong conclusion.The authors used a case-control design in which mothers receiving either phenytoin or carbamazepine were paired to control mothers who were not receiving antiepileptic drugs (AEDs). The IQs of children exposed to AEDs in utero were compared with IQs of nonexposed children for their respective control group. However, we calculate a -10.3 point (phenytoin-phenytoin control) difference and a -3.4 (carbamazepine-carbamazepine control) difference rather than -10.6 and -1.8 as reported in Table 3.The principal conclusion was based on informally contrasting a significant (phenytoin-phenytoin control) IQ difference with a nonsignificant (carbamazepine-carbamazepine control) IQ difference. However, the appropriate statistical analysis is to compare directly the