Figure 2.

Mediation effects of skin autofluorescence (SAF) on the relationship between HbA1c and a-PWV across four analyses: (a) full sample unadjusted, (b) subsample unadjusted, (c) full sample adjusted for covariates, (d) subsample adjusted for covariates. Observations: The variables in the mediation model were HbA1c (X), SAF (M) and a-PWV (Y). The adjusted model is adjusted by the minimum sufficient adjustment set provided by the directed acyclic graph, which was corroborated with the multiple regression model for a-PWV. The subsample was derived from the full sample by excluding participants identified as having definite CV risk according to SAF risk. The results showed unstandardised beta coefficients and error standards. Statistical significance is indicated by asterisks, where ** indicates p ≤ 0.01 and *** indicates p ≤ 0.001. Abbreviations: % Med: mediation percentage; a-PWV, aortic pulse wave velocity; HbA1c, glycated haemoglobin; IE (95% CI): indirect effect (95% confidence interval); IM: mediator variable; SAF, skin autofluorescence; X, independent variable; Y, dependent variable.