Effect of sequence variants on variance in glucose levels predicts type 2 diabetes risk and accounts for heritability

Daniel Gudbjartsson, Kari Stefansson and colleagues assess the effect of variants associated with mean fasting glucose levels on the variance in fasting glucose levels. They find that variants that increase both the levels and variance of fasting glucose increase type 2 diabetes risk, whereas those that increase the levels but reduce the variance do not. Sequence variants that affect mean fasting glucose levels do not necessarily affect risk for type 2 diabetes (T2D). We assessed the effects of 36 reported glucose-associated sequence variants 1 on between- and within-subject variance in fasting glucose levels in 69,142 Icelanders. The variant in TCF7L2 that increases fasting glucose levels increases between-subject variance (5.7% per allele, P = 4.2 × 10 −10 ), whereas variants in GCK and G6PC2 that increase fasting glucose levels decrease between-subject variance (7.5% per allele, P = 4.9 × 10 −11 and 7.3% per allele, P = 7.5 × 10 −18 , respectively). Variants that increase mean and between-subject variance in fasting glucose levels tend to increase T2D risk, whereas those that increase the mean but reduce variance do not ( r 2 = 0.61). The variants that increase between-subject variance increase fasting glucose heritability estimates. Intuitively, our results show that increasing the mean and variance of glucose levels is more likely to cause pathologically high glucose levels than increase in the mean offset by a decrease in variance.