We used the net reclassification index to assess the extent to which adding genotype data to the phenotype based risk models resulted in the movement of participants into risk categories that more accurately reflected their eventual outcome. In these analyses, we used the same four risk categories (≤5%, 5-9.9%, 10-14.9%, and ≥15%) and did the analyses separately for people diagnosed as having type 2 diabetes on follow-up and those who remained healthy. Among the eventual cases, a shift to a higher risk category on the addition of genetic information is classed as a correct reclassification and a downward shift as an incorrect reclassification. The opposite is the case for people who remained healthy. The addition of the simple gene count to the Cambridge risk score resulted in a net reclassification improvement of −1.1% (95% confidence interval −6.0% to 3.8%) (table 4), and the addition of information from the genetic risk function produced a net reclassification improvement of 4.6% (−2.0% to 11.3%) (P=0.17) (web table L). Neither the simple gene count score (net reclassification improvement 0.2%, −5.1 to 4.7) (table 5)
