A phase III study was performed to compare the efficacy and safety of lamotrigine (Lamictal1), desipramine (Norpramin1), and placebo in the treatment of unipolar depression. Desipramine is extensively metabolized by cytochrome P450 2D6 (CYP2D6), and kinetics of this compound are altered in poor metabolizers. Genotyping was utilized to exclude poor metabolizers in order to increase subject safety and to eliminate the need to continuously monitor plasma desipramine levels. As part of screening, subjects were genotyped for the *3(A), *4(B), and *5(D)alleles, which identify approximately 95% of poor metabolizers. Extensive metabolizers were eligible for randomization to the lamotrigine, desipramine, or placebo arm. Follow-up genotyping for the *6(T)and *7(E)alleles was performed after study enrollment and was used to identify poor metabolizers who may have been incorrectly identified as extensive metabolizers upon initial three-allele screening. Of 628 subjects screened for *3(A), *4(B), *5(D)alleles, 590 (93.9%) were classified as extensive metabolizers. The remaining 38 (6.1%) subjects were poor metabolizers and excluded. Subsequent *6(T)and *7(E)testing revealed that two poor metabolizers had been enrolled, and the follow-up genotyping provided an explanation for the high desipramine plasma concentrations in one subject. No differences in phenotypic or allelic frequencies were found between the study population and literature populations. However, the frequency of poor metabolizers varied among clinical sites (0–15%). For a compound that is extensively metabolized by CYP2D6, prescreening subjects for *3(A), *4(B), *5(D), *6(T)and *7(E)alleles can increase subject safety and eliminate the need to continuously monitor drug plasma concentrations.