Skip to main content
. 2017 Mar 16;8(17):29116–29124. doi: 10.18632/oncotarget.16249

Figure 3. The ROC curves of the logistic regression model and five serum metabolites.

Figure 3

(A) The ROC curves of CA19-9. ROC curves for five metabolites with the highest AUROC: (B) 3-ketosphingosine (AUC=0.818, P<0.001, 95% CI 0.708–0.928), (C) dodecanoylcarnitine (AUC=0.831, P<0.001, 95% CI 0.714-0.948), (D) keto_palmitic_acid (AUC=0.885, P<0.001, 95% CI 0.799–0.971), (E) tauroursodeoxycholic_acid (AUC=0.826, P<0.001, 95% CI 0.720–0.949), (F) taurocholic_acid (AUC=0.834, P<0.001, 95% CI 0.716–0.937). (G) ROC curve analysis for the predictive power of combined plasma biomarkers for distinguishing PC from DM controls. The final logistic model included 2 plasma biomarkers (shown in Table 2).