Skip to main content
. 2023 Aug 28;6(8):e2331197. doi: 10.1001/jamanetworkopen.2023.31197

Figure 2. Discrimination and Calibration of the Prediction Model for Early-Stage PDAC vs Benign Disease.

Figure 2.

A, Receiver operating characteristic (ROC) curves for the risk prediction model, CA19-9, bilirubin, and both CA19-9 and bilirubin. B, LOESS-smoothed flexible calibration curve (light blue) with 95% CIs (shaded area), grouped proportions (dots) with 95% CIs, and kernel density estimates of the predicted risks for patients with PDAC (dark blue density plot), and patients with benign disease (orange density plot). The gray oblique reference line indicates perfect calibration with an intercept of 0, a slope of 1, an E:O ratio of 1, and an ICI of 0. AUC refers to area under the receiver operating characteristic curve; CA19-9, carbohydrate antigen 19-9; E:O, ratio of expected to observed number of patients with PDAC; ICI, integrated calibration index; LOESS, locally estimated scatterplot smoothing; PDAC, pancreatic ductal adenocarcinoma; and ROC, receiver operating characteristic.