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

Mohamad Dbouk,Toshiya Abe,Chiho Koi,Yohei Ando,Helena Saba,Elizabeth Abou Diwan,Anne MacGregor-Das,Amanda L Blackford,Evelina Mocci,Katie Beierl,Ali Dbouk,Jin He,Richard Burkhart,Anne Marie Lennon,Lori Sokoll,Marcia Irene Canto,James R Eshleman,Michael Goggins
DOI: https://doi.org/10.1158/1078-0432.CCR-23-0655
2023-10-13
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.
What problem does this paper attempt to address?