Table 1

HRs and 95% CIs for the risk of developing type 2 diabetes according to baseline levels of TGF-β1

Tertile 1Tertile 2Tertile 3Pfor trend
Median (lower–upper limit) (men)27.80 (6.10–31.92)35.17 (31.93–38.44)43.18 (38.45–60.60)
Median (lower–upper limit) (women)28.43 (9.17–31.92)34.75 (31.93–37.63)42.18 (37.64–59.29)
n of case/noncase subjects143/481143/491174/502
Model 11.01.08 (0.83–1.42)1.41 (1.08–1.83)*0.012
Model 21.01.09 (0.82–1.46)1.42 (1.05–1.93)*0.019
Model 31.01.09 (0.80–1.48)1.40 (1.02–1.91)*0.032
Model 41.01.02 (0.74–1.42)1.35 (0.94–1.93)0.088
  • Data are HRs (95% CIs) for tertiles of TGF-β1 (ng/ml) unless otherwise indicated. HRs and 95% CIs were estimated by Cox proportional hazards models. Models contained continuous variables unless otherwise indicated. Because of the case-cohort design, correction of the variance estimation was performed based on the sampling weights to give SE estimates for the parameter estimates. The inverse of the sample sizes for the subcohort by the cohort of interest yielded survey- and sex-specific sampling weights. If required, we additionally differentiated between case and noncase subjects. Sex-specific tertiles of the weighted distributions in the subcohort were used.

  • *P < 0.05 compared with tertile 1. Model 1: adjusted for age, sex, and survey; model 2: adjusted for factors in model 1 plus BMI and lifestyle factors (i.e., smoking status [never smoker, former smoker, and current smoker], alcohol consumption [0, 0.1–39.9, and ≥40 g/day for men and 0, 0.1–19.9, and ≥20 g/day for women], and physical activity [inactive and active]); model 3: adjusted for factors in model 2 plus systolic blood pressure, total/HDL cholesterol, and parental history of diabetes (negative, positive, and unknown); model 4: adjusted for factors in model 3 plus C-reactive protein, MIF, IL-8, soluble E-selectin, and RANTES (all included in the models stratified in sex-specific tertiles; sample size after exclusion of subjects with missing values for additional biomarkers: n = 1,847).