Cytochrome P4502D6 (CYP2D6) is a highly polymorphic gene locus with > 50 variant alleles which lead to a wide range in enzymatic activity. So called poor metabolizers are carriers of any two non-functional alleles of the CYP2D6 gene. CYP2D6 genotyping is cumbersome and the question of how much genotyping is necessary for an accurate phenotype prediction is still debated. The goal of this study was to determine the optimum amount of genotyping required to accurately predict the phenotype at a reasonable cost in a white North American population. To address this issue, we designed a polymerase chain reaction (PCR)/restriction fragment length polymorphism-based genotyping strategy to detect 'key' mutations linked to extensive metabolizer or poor metabolizer associated alleles in combination with extra-long PCR (XL-PCR). All mutations with the exception of gene deletions and duplications are detectable by simple restriction digestion analysis and agarose gel electrophoresis. In addition, we utilized a genotyping algorithm based on our own and published allele frequency data and phenotype analysis to calculate the probability of a correct genotype (and thus, phenotype) assignment. As little as one XL-PCR reaction followed by a maximum of six reamplification reactions allows an accurate prediction of an individual's genotype to 99.15%. As few as four reamplification reactions identify 97.9% of poor metabolizer individuals. We evaluated our model in 208 white North Americans by testing for the presence of 'key' mutations linked to CYP2D6*2, *3, *4, *6, *7, *8, *9, *10, *11, *12, *15, *17 and *18 alleles and the *5, *13 and *16 gene deletions. For all individuals, the correct phenotype has been predicted. Discordant phenotype assignment occurred in only two individuals which subsequently was attributed to CYP2D6 inhibition by concomitant drug therapy.