TABLE 2

Univariate and multivariate analysis

∼1 SD changeUnivariate hazard ratio (95% CI)PMultivariate hazard ratio (95% CI)P
Age5 years1.03 (0.88–1.21)0.681.00 (0.84–1.19)1.00
BMI3 kg/m21.64 (1.43–1.88)<0.00011.28 (1.10–1.49)0.0012
Ln(WCC)0.25 ln(109 cells/l)1.41 (1.20–1.65)<0.00011.14 (0.94–1.38)0.19
Systolic blood pressure20 mmHg1.22 (1.00–1.48)0.051.02 (0.83–1.25)0.86
Ln(triglyceride)0.5 ln(mmol/l)2.25 (1.81–2.79)<0.00011.42 (1.03–1.96)0.031
Total cholesterol0.5 mmol/l1.26 (1.10–1.45)0.00081.04 (0.87–1.24)0.69
HDL cholesterol0.25 mmol/l0.60 (0.49–0.75)<0.00010.84 (0.64–1.09)0.18
Alcohol10 units/week1.02 (0.89–1.16)0.830.95 (0.82–1.10)0.48
SmokingY/N1.13 (0.80–1.60)0.50.95 (0.66–1.49)0.97
Pravastatin therapy0.70 (0.49–0.99)0.0450.72 (0.50–1.03)0.070
Ln(CRP)1 ln(mg/l)1.55 (1.32–1.82)<0.00011.30 (1.07–1.58)0.0075
Glucosemmol/l
    Quintile I: <4.3Referent levelReferent level
    Quintile II: >4.3 to 4.50.98 (0.35–2.75)0.90 (0.32–2.54)
    Quintile III: >4.5 to 4.70.42 (0.11–1.85)<0.0001*0.38 (0.10–1.42)<0.0001*
    Quintile IV: >4.7 to 5.01.53 (0.85–3.58)1.31 (0.56–3.08)
    Quintile V: >5.011.9 (6.0–23.7)8.78 (4.37–17.6)
  • Univariate and multivariate hazard ratios with P values and approximate 95% CIs for baseline predictors of development of diabetes. The given hazard ratio for continuous covariates is for an approximate 1 SD change; for example, for log(triglyceride) the hazard ratio is for a 0.5 unit change in the logarithm of triglyceride (with triglyceride measured in millimoles per liter). For categorical variates, the hazard ratio is with or without the stated attribute, e.g., smoker or nonsmoker. Ln, natural logarithm.

  • *

    * P for the χ2 test on four degrees of freedom for the equality of the five quintiles.