Figure 1

Biomarker modeling in training (Sample Set B) and validation (Sample Set A) sets. Biomarker models were generated for all two and three marker combinations in Sample Set B (Table 3). These models were then validated in PDAC versus healthy/non-cancer group of Sample Set A (Table 4). Displayed are three models which showed a significant improvement in AUC to that of CA19.9 alone in both training (1a) and validation sets (1b). Confidence intervals (CI) for AUC were calculated using DeLong’s method. P-values are given in Table 4.