Figure 2.

Differences in metabolite concentrations from cross-sectional analysis of KORA S4. Plots (A, B) show the names of metabolites with significantly different concentrations in multivariate logistic regression analyses (after the Bonferroni correction for multiple testing with P<3.6 × 10⁻⁴) in the five pairwise comparisons of model 1 and model 2. Plot (C) shows the average residues of the concentrations with standard errors of the three metabolites (glycine, LPC (18:2) and acetylcarnitine C2) for the NGT, IGT and dT2D groups. Plot (A) shows the results with adjustment for model 1 (age, sex, BMI, physical activity, alcohol intake, smoking, systolic BP and HDL cholesterol), whereas plots (B, C) have additional adjustments for HbA_1c, fasting glucose and fasting insulin (model 2). Residuals were calculated from linear regression model (formula: T2D status∼metabolite concentration+model 2). For further information, see Supplementary Table S4.