Figure 2

Cumulative odds ratios (ORs) for 4, 8, and 30 genomic markers to predict type 2 diabetes (T2D). (a) The x-axis shows the cumulative OR for the four genomic markers with the most credible evidence (TCF7L2, CDKAL1, CDKN2A/B, and PPARG) in a simulated population of individuals developing T2D (dashed line with open circles) and a corresponding population of individuals who do not develop T2D (solid line with filled circles). Details of the model are described in the methods. Each circle represents the number of individuals per 1,000 in each of the 20 intervals (equally spaced on the log scale) from cumulative ORs from 0.1 to 10. There is considerable overlap with the two groups, indicating poor test performance (false-positive rate: 4.5% and detection rate: 7.2%). (b) Includes the same information for the same four genomic markers along with the four next most commonly used markers (HHEX, SLC30A8, KCNJ11, and WFS1). Performance does not improve appreciably. (c) Includes all 30 genomic markers identified in the evidence review and results in the best, but still poor, performance. DR, detection rate; FPR, false-positive rate.