Statistical power of association using the extreme discordant phenotype design

Ge Zhang, Daniel W. Nebert, Ranajit Chakraborty, Li Jin

Research output: Contribution to journalArticlepeer-review

28 Scopus citations


BACKGROUND: Selective genotyping has been proven to be an effective design for mapping quantitative trait loci (QTL), either by linkage or by allelic association, wherein the individual trait values can be used as the indices for phenotype selection. It has also been proposed that association studies of dichotomous traits can benefit from such design. When there is no quantitative measurement for phenotype available, cases and/or controls having extreme discordant phenotypes (EDP) can still be selected, based on their exposure status to a drug toxicity or environmental risk factor. The advantage of EDP design is intuitive and it has been successfully used in a number of studies. METHODS: In this report, we developed a statistical method to calculate the power of EDP methodology, using a mixture model of genotype-specific distributions of a single biallelic susceptibility locus. We also compared the power of three statistical tests commonly used in association studies - including the χ2 test of allelic frequencies, the χ2 test of genotype frequencies, and the Armitage trend test. The power of two different EDP designs was evaluated under a range of scenarios. RESULTS AND CONCLUSION: Our results indicate that the χ2 test of genotype frequency is a robust, though less powerful, test for single-locus association, and that EDP methodology is a powerful design for genetic association studies - especially those of common diseases caused by quantifiable drug toxicity or environmental risk factors.

Original languageEnglish
Pages (from-to)401-413
Number of pages13
JournalPharmacogenetics and Genomics
Issue number6
StatePublished - Jun 2006


  • Association study
  • Environmental exposure
  • Extreme discordant phenotype
  • Selective genotyping
  • Statistical power


Dive into the research topics of 'Statistical power of association using the extreme discordant phenotype design'. Together they form a unique fingerprint.

Cite this