Diagnostic Performance of a Tumor Marker Gene Test to Personalize Serum CA19–9 Reference Ranges

Abstract

Abstract Purpose: CA19–9 synthesis is influenced by common variants in the fucosyltransferase (FUT) enzymes FUT3 and FUT2. We developed a clinical test to detect FUT variants, and evaluated its diagnostic performance for pancreatic ductal adenocarcinoma (PDAC). Experimental Design: A representative set of controls from the Cancer of the Pancreas Screening study was identified for each FUT functional group. Diagnostic sensitivity was determined first in a testing set of 234 PDAC cases, followed by a 134-case validation set, all of whom had undergone resection with curative intent without neoadjuvant therapy. Tumor marker gene testing was performed in the Johns Hopkins Molecular Diagnostics Laboratory. CA19–9 levels were measured in the Hopkins Clinical Chemistry lab. Receiver operating characteristic (ROC) curve analysis was used to evaluate the discriminative ability of CA19–9 alone versus with the gene test. Results: Applying the CA19–9 standard cutoff (<36 U/mL) to all 716 subjects yielded a 68.8% sensitivity in the test set of cases, 67.2% in the validation set, at 91.4% specificity. Applying 99th percentile cutoffs according to each individual's FUT group (3, 34.9, 41.8, and 89.2, for the FUT3-null, FUT-low, FUT-intermediate, and FUT-high groups, respectively) yielded a diagnostic sensitivity for CA19–9 in the first set of cases of 66.7%, 65.7% in the validation set, at 98.9% specificity. ROC analysis for CA19–9 alone yielded an AUC of 0.84; with the tumor marker gene test, AUC improved to 0.92 (P < 0.001). Conclusions: Using a tumor marker gene test to personalize an individual's CA19–9 reference range significantly improves diagnostic accuracy.