We demonstrated that sampling approaches that increase the proportion of resistant cases versus random sampling or case-control sampling lead to increases in the power to detect resistance effects. The high-risk design, and modified approaches, yield power similar to, or much greater than, case-control sampling by increasing the presence and variability of factors that impact the liability distribution. Notably, a sampling approach that ascertains case and control fathers from below and above the mean on an environmental risk factor respectively yields dramatic increases in power to detect resistance factors. However, this approach, which is contingent both on parental phenotype and measured parental environment would rely on a much larger sample from which sub-sampling would occur, thus substantially limiting its feasibility. The obvious exception to this is extant cohorts with parental data. More importantly, the low-risk design which ascertains based solely on measured liability irrespective of parental phenotype, and is hence much more feasible, yields increases in power similar to the high-risk designs with the previously mentioned exception of extreme environmental risk sampling. We contend that the low-risk design is substantially more
