Table 4.
Area under the curve (AUC), sensitivity, specificity, accuracy and positive predictive value (PPV) of ROC models (with 95% CI), from the main PLS analysis (N = 336, X-set = 21, Y-set = 285)
| AUC | AUCbb | Sensitivity | Specificity | Accuracy | PPV | |
|---|---|---|---|---|---|---|
| Adjustment covariates (ADJ)a | 0.842 (0.794, 0.891) | 0.821 (0.766, 0.868) | 0.752 (0.662, 0.829) | 0.802 (0.743, 0.852) | 0.785 | 0.0015 |
| X1 scores + ADJ | 0.846 (0.797, 0.894) | 0.825 (0.766, 0.875) | 0.743 (0.653, 0.821) | 0.838 (0.783, 0.884) | 0.806 | 0.0018 |
| X1 + X2 scores + ADJ | 0.854 (0.808, 0.900) | 0.831 (0.772, 0.881) | 0.743 (0.653, 0.821) | 0.824 (0.768, 0.872) | 0.797 | 0.0017 |
| X1 + X2 + X3 scores + ADJ | 0.859 (0.811, 0.907) | 0.836 (0.778, 0.887) | 0.796 (0.710, 0.866) | 0.788 (0.729, 0.840) | 0.791 | 0.0015 |
| Y1 scores + ADJ | 0.841 (0.793, 0.890) | 0.817 (0.760, 0.865) | 0.735 (0.643, 0.813) | 0.820 (0.763, 0.868) | 0.791 | 0.0016 |
| Y1 + Y2 scores + ADJ | 0.845 (0.795, 0.894) | 0.820 (0.762, 0.872) | 0.735 (0.643, 0.813) | 0.851 (0.798, 0.895) | 0.812 | 0.0020 |
| Y1 + Y2 + Y3 scores + ADJ | 0.853 (0.804, 0.902) | 0.827 (0.771, 0.877) | 0.726 (0.634, 0.805) | 0.883 (0.833, 0.922) | 0.890 | 0.0025 |
aThe model is run on the ADJ including the C-reactive protein concentration, alpha-fetoprotein concentration and a composite score for liver damage.
bAUCb is the bootstrapped-cross validated estimate of the AUC. X1, X2 and X3 are the lifestyle component scores of the first, second and third PLS factors, respectively. Y1, Y2, and Y3 are the metabolomics component of the first, second and third PLS factors, respectively.