Abstract
To identify non-invasive biomarkers of non-metastatic pancreatic cancer (PC), the blood from 186 patients (PC n=28; DM-diabetes mellitus n=60; ChP-chronic pancreatitis n=47; healthy controls n=51) was analyzed for 58 candi-date biomarkers. Their effectiveness to identify PC was compared with CA19-9.
Panel defined by Random-forest (RF) analysis (CA19-9, AAT, IGFBP2, albumin, ALP, Reg3A, HSP27) outperforms CA19-9 in discrimination of PC from DM (AUC 0.92 vs. 0.82). Panel (S100A11, CA72-4, AAT, CA19-9, CB, MMP-7, S100P-s, Reg3A) is better in discrimina-tion PC from ChP than CA19-9 (AUC 0.90 vs. 0.75).
Panel (MMP-7, Reg3A, sICAM1, OPG, CB, ferritin) is better in discrimination PC from healthy controls than CA19-9 (AUC 0.89 vs. 0.78). Panel (CA19-9, S100P-pl, AAT, albumin, adiponectin, IGF-1, MMP7, S100A11) identifies PC among other groups better than CA19-9 (AUC 0.91 vs. 0.80).
Panel defined by logistic regression analysis (prealbumin, IGFBP-2, DJ-1, MIC-1, CA72-4) discriminates PC from DM worse than CA19-9 (AUC 0.80 vs. 0.82). Panel (IGF-1, S100A11, Reg1alfa) outperforms CA19-9 in discrimination PC from ChP (AUC 0.76 vs. 0.75).
Panel (IGF-2, S100A11, Reg3A) outperforms CA19-9 in discrimination PC from healthy controls (AUC 0.95 vs. 0.78). Panel (albumin, AAT, S100P-serum, CRP, CA19-9, TFF1, MMP-7) outperforms CA19-9 in identification PC among other groups (AUC 0.89 vs. 0.8).
The combination of biomarkers identifies PC better than CA19-9 in most cases. S100A11, Reg3A, DJ-1 were to our knowledge identified for the first time as possible serum biomarkers of PC.