The transmission disequilibrium test (TDT), an alternative to case-control analysis that is not influenced by population stratification, focuses on affected child trios (ACTs) comprising affected cases and their parents. Unaffected child trios (UCTs) have also been proposed but mainly to rule out segregation distortion. To explore situations when UCTs are preferable to detect transmission distortion, we compared the number of UCTs and ACTs needed to achieve 80% power for a wide variety of scenarios. For a given genetic model, UCT sample size declined rapidly with increasing disease prevalence, whereas ACT sample size remained constant. Furthermore, at some prevalence value (40-60% depending on model parameters), detection of transmission distortion could be accomplished with fewer UCTs than ACTs. Such high prevalence may be found in special populations (diabetes among Pima Indians), secondary conditions (renal/retinal complications in diabetes), and pharmacogenetics (responders to treatment). Also, because exposure to an additional risk factor can increase the disease prevalence in an exposed sub-group, we explored how sample size requirements vary by exposure status. Whereas power differences between exposed and unexposed ACTs could be explained solely by genetic risk ratios, sub-group-specific disease prevalence also played an important role in UCTs. Finally, we considered the impact of including 5, 10, 20, or 30% misclassified ACTs in the UCT sample and found that a 24, 58, 180, or 550% larger sample would be required. In conclusion, UCTs can detect transmission distortion more effectively than ACTs when disease prevalence reaches 40-60%, although some efficiency may be lost owing to misclassification. Moreover, focusing on particular sub-groups defined by exposure status can potentially increase power, but such gains depend heavily on the nature of the gene-exposure interaction.
Copyright 2000 Wiley-Liss, Inc.