Sharon Marsh Washington University School of Medicine, Campus Box 8069, 660 S. Euclid Ave., St. Louis, MO 63110, USA Tel.: +1 314 747 5186; Fax: +1 314 362 3764; E-mail: smarsh@ im.wustl.edu ‘Allele frequencies for most polymorphisms show significant differences between populations, and within common population classifications.’ The release of the first working draft of the human genome was quickly followed by the prediction that approximately 1.42 million SNPs would be identified in the human genome [1]. This estimate was rapidly exceeded. Currently, the National Center for Biotechnology Information (NCBI) polymorphism repository dbSNP, has well over 4 million polymorphisms [101]. Endeavours, such as the International HapMap project [102], have provided insights into not only the quantity of polymorphisms, but also the variability in polymorphism frequency between world populations (specifically Europe, Yoruba, Japan and China) [2]. The available data set for pharmacogenetics research will be further expanded with the 1000 Genomes Project [103], which aims to sequence the entire genome of 1000 individuals from multiple populations (chiefly the extended HapMap sample set). This will provide a highly detailed map of polymorphisms that occur at a frequency of 1% or more in the human genome. The ultimate goal of pharmacogenetics is to define robust markers that can be used in the clinical setting to aid in individualized treatment selection. However, despite the plethora of genetic information at our fingertips the true number of established pharmacogenetic markers ready for integration into clinical practice is remarkably small [3,4], and their usefulness may be geographically limited. The lack of robust clinical pharmacogenetic markers could be owing to a multitude of reasons, including environmental factors, diet, concurrent medications, sample size, treatment regimen, population variation in allele frequencies and the lack of available sample sets to perform validation analyses. In addition to the reasons above, commonly published polymorphisms often suffer from contradictory results, making their role in the clinic unclear. Sometimes this can be clarified by taking a closer look at the studies. For example, conflicting data for UGT1A1*28, an FDA approved marker for predicting toxicity in patients receiving the chemotherapy drug irinotecan, can be clarified by including irinotecan dose in the analysis [5]. The polymorphism appears to be a marker for neutropenia in patients receiving high-dose irinotecan, but is a less useful marker for patients receiving low-dose irinotecan. Another factor involved in the use of UGT1A1*28 as a clinical marker is the allele frequency difference between populations. Although UGT1A1*28 is a functional polymorphism and has been associated with irinotecan toxicity in multiple populations [6], the allele frequency is lower in Asian populations (approximately 15% compared with 30–45% in Caucasian and African populations) [7]. Other potentially functional variants in UGT1A1 (e.g., UGT1A1*6) are prevalent in Asian populations and also demonstrate an association with irinotecan toxicity [8]. So although UGT1A1*28 appears to be a relevant clinical marker for all populations, some information would be missed by screening this alone in every patient. There is still work needed to define a panel of markers for irinotecan toxicity in all populations.