Detection of resectable pancreatic cancer with SOMAmer proteomic technology.

Abstract

162 Background: To improve the current dismal 5-year survival rates for pancreatic adenocarcinoma (PC) it is essential to develop strategies focused on detection in high-risk populations. This study employed a unique SOMAmer proteomics platform to identify a biomarker panel that discriminates between pancreatic cancer and control patients. A secondary goal was to compare the panel with CA 19-9. Methods: Plasma samples (total < 20 ul) were analyzed for 825-proteins using a SOMAmer proteomics platform on training set of 100 cases and 69 controls from a single institution. The control group contained individuals with GI symptoms, including acute and chronic pancreatitis, and normal controls. An independent blinded validation set from another center consisted of 43 PC patients and 47 controls. About two thirds (96/143) of the study cases were in resectable stages I-IIb. Biomarkers were identified by Random Forest backwards selection, using a false-discovery-rate corrected value of p<0.001. CA19-9 levels were available on a subset of the patients from the training set (87 PC and 29 controls). A CA 19-9 level of > 40 U/ml was defined as positive. Results: In the training set, 37 markers were significantly different between case and controls. A 10 marker Random Forests classifier had an AUC of 0.91 for training and 0.90 for the independent validation set. Comparison of the SOMAmer results with CA 19-9 levels are shown in the table. There was 78% agreement between the two tests. One could maximize sensitivity with some sacrifice of specificity by combining either test positive (95% and 66%, respectively) or conversely maximize specificity (97%) with lower sensitivity of 77% if both tests must be positive. Conclusions: 1) The 10 marker SOMAmer biomarker panel is promising for the detection of resectable pancreatic cancer with an AUC of 0.91 on a training set, which was confirmed on an independent validation set. 2) Preliminary data suggests that this test is complementary to CA 19-9 with improved combined sensitivity of up to 95%. 3) Studies to assess further applications of these panels, including for risk assessment in individuals with a family history of pancreatic cancer, are underway. [Table: see text]