The most favourable condition shown is , that is there are 1000 markers with effects on disease. Figure 3 and Table 4 show that a few thousand cases and controls could yield a clinically useful AUC, but under most conditions several tens of thousands are needed. Under less favourable conditions – low heritability, low proportion of null markers – several hundred thousand cases and controls are needed to obtain an AUC within 10% of the achievable level, and even an AUC of 0.75 requires some tens of thousands of subjects. In the worst case the order of magnitude is of the millions.
