Table 4

OR* showing natural direct and indirect effects† of study (KYH vs Tromsø 7) on diabetes‡ prevalence assessed from mediation analyses and mediated percentage for different sets of risk factors (BMI, waist circumference, smoking, hsCRP), by sex

Model 1
BMI and waist circumference included as mediators
Model 2
BMI, waist circumference, smoking, and hsCRP included as mediators
 Natural direct effect2.02 (1.70, 2.40)1.87 (1.57, 2.24)
 Natural indirect effect0.99 (0.95, 1.04)1.07 (0.99, 1.16)
 Total effect2.01 (1.68, 2.40)2.01 (1.69, 2.38)
 Percentage mediated−1.1% (−8.9, 5.5)9.9% (−0.6, 20.8)
 Natural direct effect1.99 (1.70, 2.35)1.77 (1.49, 2.11)
 Natural indirect effect1.81 (1.68, 1.94)2.04 (1.85, 2.26)
 Total effect3.66 (3.13, 4.26)3.62 (3.10, 4.21)
 Percentage mediated46.0% (39.6, 53.8)55.5% (46.5, 66.0)
  • *Adjusted for age.

  • †Total effect of exposure is decomposed into natural direct and indirect effects. Natural indirect effect means effect of exposure that is mediated by a specific set of risk factors. Natural direct effect is the remaining effect of an exposure after quantifying the natural indirect effect. In our analysis, the study (KYH vs Tromsø 7) was considered the exposure, while diabetes risk factors were considered possible mediators.

  • ‡Diabetes defined as HbA1c ≥6.5% (48 mmol/mol) and/or self-reported diabetes and/or use of medication with ATC code A10 (antidiabetics) according to the Anatomical Therapeutic Chemical (ATC) classification.

  • BMI, body mass index; HbA1c, glycated hemoglobin; hsCRP, high sensitivity C reactive protein; KYH, Know Your Heart.