Subjects <50 years old (n = 144 diabetes cases) | ||||
---|---|---|---|---|

Model without genetic information | Model using 40 individual SNPs | Model using 40-SNP weighted risk score | Model using prior 17-SNP weighted risk score | |

Men (vs. women) | 0.45 (0.30–0.68) | 0.43 (0.28–0.67) | 0.46 (0.30–0.70) | 0.46 (0.30–0.70) |

Family history of diabetes vs. not | 2.26 (1.55–3.30) | 2.22 (1.49–3.29) | 2.20 (1.50–3.22) | 2.18 (1.49–3.19) |

BMI (kg/m^{2}) | 1.10 (1.06–1.14) | 1.11 (1.07–1.15) | 1.11 (1.07–1.15) | 1.11 (1.08–1.15) |

Fasting plasma glucose (mg/dl) | 1.14 (1.11–1.16) | 1.13 (1.11–1.16) | 1.13 (1.11–1.16) | 1.13 (1.11–1.16) |

Systolic blood pressure (mmHg) | 1.02 (1.01–1.03) | 1.03 (1.01–1.04) | 1.02 (1.01–1.03) | 1.02 (1.01–1.03) |

HDL cholesterol (mg/dl) | 0.96 (0.95–0.98) | 0.96 (0.95–0.98) | 0.96 (0.95–0.98) | 0.96 (0.95–0.98) |

Fasting triglycerides (mg/dl) | 1.00 (1.00–1.00) | 1.00 (1.00–1.00) | 1.00 (1.00–1.02) | 1.00 (1.00–1.00) |

Genetic risk score | — | — | 1.24 (1.13–1.36) | 1.39 (1.22–1.59) |

C-statistic (95% CI) | 0.908 (0.884–0.932) | 0.920 (0.898–0.941) | 0.911 (0.887–0.935) | 0.909 (0.884–0.933) |

P value for difference in C-statistic | 0.02 | 0.3 | 0.89 | |

Calibration χ^{2} (P value) | 4.37 (0.8) | 6.60 (0.6) | 9.78 (0.28) | |

NRI (%) | 11.4 | 10.2 | 7.5 | |

P value | 0.002 | 0.001 | 0.01 |

Subjects ≥50 years old (n = 302 diabetes cases) | ||||
---|---|---|---|---|

Model without genetic information | Model using 40 individual SNPs | Model using 40-SNP weighted risk score | Model using prior 17-SNP weighted risk score | |

Men (vs. women) | 1.03 (0.76–1.38) | 1.04 (0.76–1.41) | 1.05 (0.78–1.41) | 1.05 (0.78–1.41) |

Family history of diabetes vs. not | 2.09 (1.54–2.85) | 2.18 (1.58–3.00) | 2.11 (1.55–2.88) | 2.12 (1.56–2.88) |

BMI (kg/m^{2}) | 1.08 (1.05–1.11) | 1.09 (1.06–1.12) | 1.09 (1.06–1.12) | 1.09 (1.06–1.12) |

Fasting plasma glucose (mg/dl) | 1.14 (1.13–1.16) | 1.14 (1.12–1.16) | 1.14 (1.12–1.16) | 1.14 (1.12–1.16) |

Systolic blood pressure (mmHg) | 1.01 (1.00–1.02) | 1.01 (1.00–1.02) | 1.01 (1.01–1.02) | 1.01 (1.01–1.02) |

HDL cholesterol (mg/dl) | 0.98 (0.97–0.99) | 0.98 (0.97–0.99) | 0.98 (0.97–0.99) | 0.98 (0.97–0.99) |

Fasting triglycerides (mg/dl) | 1.00 (1.00–1.00) | 1.00 (1.00–1.00) | 1.00 (1.00–1.00) | 1.00 (1.00–1.00) |

Genetic risk score | — | — | 1.11 (1.03–1.19) | 1.13 (1.02–1.25) |

C-statistic (95% CI) | 0.883 (0.863–0.903) | 0.888 (0.869–0.908) | 0.884 (0.865–0.904) | 0.884 (0.865–0.904) |

P value for difference in C-statistic | 0.02 | 0.2 | 0.18 | |

Calibration χ^{2} (P value) | 10.97 (0.2) | 15.01 (0.06) | 8.46 (0.39) | |

NRI (%) | 5.7 | 0.4 | 0.02% | |

P value | 0.001 | 0.7 | 0.98 |

Data are OR (95% CI) unless otherwise indicated. Data in bold represent statistical significance.

↵†The simple clinical variables–adjusted model included sex, family history of diabetes (self-report that one or both parents had diabetes), BMI, fasting glucose level, systolic blood pressure, HDL cholesterol, and fasting triglycerides levels (3). No age adjustment was done in the age-stratified models.

To evaluate the individual contribution of each SNP, we entered one term per SNP (total 40 terms plus terms for sex or clinical variables) in the logistic regression models.

We constructed a weighted genetic risk score using 40 SNPs currently associated with type 2 diabetes and a weighted genetic risk score using 17 SNPs that we used in our previous report (1). rs689 at

*INS*on chromosome 11, previously included in our 18-SNP genetic risk score (1), was not replicated in posterior meta-analyses and is therefore not included in the current 17-SNP or 40-SNP analyses. Moreover, rs5945326 at*DUSP9*on chromosome X (10) is not included in the analysis because there are no available genotyping or imputation data for this SNP in the Framingham Offspring Study.For the construction of the weighted risk scores, we counted risk alleles (0, 1, 2) for each genotyped SNP—or its dosage when imputed—(actual distribution ranging from 28 to 53) and multiplied each SNP genotype by its published β coefficient for diabetes risk (10). We added up the product of that multiplication at each SNP, divided the sum by twice the sum of the β coefficients, and multiplied the result by the number of SNPs.

ORs, 95% CIs, and C-statistics for the 144 cases of diabetes in 6,763 person-observations in subjects <50 years old and for the 302 cases of diabetes in 4,595 person-observations in subjects ≥50 years old were calculated using pooled logistic regression with generalized estimating equations. Mean age at diabetes onset was 49.30 years for subjects <50 years old at baseline and 66.07 years for subjects ≥50 years old at baseline. We took 50 years as the age cutoff point because of the low incidence rate of diabetes in younger subjects when lower values were chosen. Sensitivity analyses using a cutoff age of 45 years (84 cases in 5,095 person-observations) showed a lower NRI in younger subjects (3.59%;

*P*= 0.2), though this result should be taken with caution because of the low number of cases.For NRI evaluation, we established three risk categories (low, intermediate, and high). The percentages of low, medium, and high risk of diabetes are based on the distribution of the cumulative incidence of diabetes across our population, in which cumulative incidence was low for a predicted risk <2%, intermediate for predicted risks ≥2% and ≤8%, and high when predicted risk was >8% (this assumption is an a priori requirement for the NRI calculation) (15). NRI is better if more people who develop diabetes are reclassified as higher risk when the genotype score is added to the model, and more people who remain free of diabetes are classified as lower risk when the score is added. The NRI is penalized for misreclassification; for instance, if many people who develop diabetes are classified as lower risk by adding the genetic risk score to the model.

Data for the sex-adjusted model in age-stratified analyses are shown in supplementary Table A3. Complete data for the population overall are shown in supplementary Table A4.