thousands of cases and controls become prohibitively expensive, a multi-stage design can be more economical 32. In such a design, all SNPs are tested in a random subset of cases and controls, and those exhibiting a nominal pre-determined significance level are taken through to be tested in the remainder of the study sample. Subsequent analysis needs to be carried out for the different stages combined to maintain power level 33. The optimal multi-stage design depends on the relative cost of stage 1 vs. stage 2 genotyping, but also on the underlying (unknown) disease model, which presents a problem in the design. If budgetary reasons are not a strong issue, one-stage designs are preferred over multi-stage designs. With genotyping costs for whole genome panels ever decreasing, the case for multi-stage designs for budgetary reasons is becoming less strong.
