1. ¹Center for Human Genome Variation, Institute for Genome Sciences & Policy, Duke University, Durham, NC 27708, USA
2. ²Department of Psychiatry, Duke University Medical Center, Durham, NC, USA
3. ³Bioinformatics Research Center, North Carolina State University, Raleigh, NC, USA
4. ⁴Biological Psychiatry, John Umstead Hospital, Butner, NC, USA

Correspondence: Dr DB Goldstein, Center for Human Genome Variation, Institute for Genome Sciences & Policy, Duke University, Durham, NC 27708, USA. Tel: +1 919 684 0896; Fax: +1 919 668 6787; E-mail: d.goldstein@duke.edu

⁵Current address: Pharmacogenetics Consultancy, Cabernet Pharmaceuticals, R. David Thomas Center, One Science Drive, Box 90344, Durham, NC 27708, USA

Received 11 June 2008; Revised 11 October 2008; Accepted 9 December 2008; Published online 21 January 2009.

Abstract

The Clinical Antipsychotic Trials of Intervention Effectiveness (CATIE) Phase 1 Schizophrenia trial compared the effectiveness of one typical and four atypical antipsychotic medications. Although trials such as CATIE present important opportunities for pharmacogenetics research, the very richness of the clinical data presents challenges for statistical interpretation, and in particular the risk that data mining will lead to false-positive discoveries. For this reason, it is both misleading and unhelpful to perpetuate the current practice of reporting association results for these trials one gene at a time, ignoring the fact that multiple gene-by-phenotype tests are being carried out on the same data set. On the other hand, suggestive associations in such trials may lead to new hypotheses that can be tested through both replication efforts and biological experimentation. The appropriate handling of these forms of data therefore requires dissemination of association statistics without undue emphasis on select findings. Here we attempt to illustrate this approach by presenting association statistics for 2769 polymorphisms in 118 candidate genes evaluated for 21 pharmacogenetic phenotypes. On current evidence it is impossible to know which of these associations may be real, although in total they form a valuable resource that is immediately available to the scientific community.