Metabolomic profiles associated with bone mineral density in US Caucasian women

Nutrition & Metabolism

Table 3 The AUC and comparisons of AUC of the ROC curves from different predictive models

Predictive models	AUC of the ROC curve (95% CI)	Difference of AUC (95% CI)	P value for difference
Model 1: traditional risk factors^a	0.88 (0.83–0.94)	Reference	–
Model 2: Model 1 + PLS-DA-derived score^b	0.95 (0.92–0.98)	0.07 (0.02–0.11)	0.002
Model 3: Model 1 + Individual analysis-derived score^c	0.95 (0.92–0.98)	0.07 (0.03–0.11)	0.002
Model 4: Model 1 + Both methods-derived score^d	0.97 (0.94–0.99)	0.08 (0.04–0.13)	0.0004

AUC Area under the curve, CI Confidence interval, PLS-DA Partial least squares-discriminant analysis, ROC Receiver operating characteristic
^aIncluding age, age², body mass index, current smoking, alcohol drinking, physical activity, and dairy intake
^bGenerated using γ-aminobutanoate, threonine, taurine, stachydrine, isovalerylcarnitine, lysoPE (16:0), cholic acid, and N-acetylneuraminate
^cGenerated using L-cysteine, taurine, stachydrine, L-glutamic acid, formylkynurenine, isovalerylcarnitine, ursodeoxycholic acid, tauroursodeoxycholic acid, succinate, and N-acetylneuraminate
^dGenerated using all the metabolites identified by the PLS-DA method and individual metabolite analysis

ISSN: 1743-7075