New CA19-9 Cutoff Value Helps Identify High-Risk Pancreatic Cancer Patients
Posted on 22 May 2026
Pancreatic ductal adenocarcinoma (PDAC) is frequently diagnosed at an advanced stage and remains one of the most lethal solid tumors. Clinicians commonly use serum carbohydrate antigen 19-9 (CA19-9) to assess risk and monitor therapy, but around 10% of patients genetically do not produce this marker, complicating interpretation. These cases can be misclassified as standard risk when CA19-9 appears “normal.” New findings demonstrate that adopting a lower CA19-9 cutoff can better flag these high-risk patients.
At the National Health Research Institutes, working with National Cheng Kung University Hospital and Kaohsiung Medical University Hospital in Taiwan, researchers evaluated a dual-threshold model for the pancreatic tumor marker CA19-9. The approach is designed to identify patients who naturally produce little or no CA19-9 because of inherited genetic differences, which can make their results appear low even when risk is present. The model proposes adding a lower cutoff of 7 units/mL or less to help flag these patients, while still using the existing interpretive ranges, including the conventional 37 units/mL threshold.
The approach specifically focuses on Lewis antigen-negative “nonproducer” patients whose FUT3 gene polymorphisms impair fucosyltransferase (FUT) activity and CA19-9 synthesis. The team used whole-exome sequencing to determine FUT2/FUT3 genotypes for 615 patients with pancreatic cancer treated at the two Taiwanese hospitals. Participants were stratified by FUT genotype into four categories spanning FUT3-null (nonproducers) to FUT-high. Half of the cohort served as a training set to integrate genotype with CA19-9 levels and derive thresholds, which were validated in the remaining participants.
The analysis established a CA19-9 level of 7 units/mL or lower as a proposed cutoff for identifying pancreatic cancer patients with a Lewis antigen-negative genotype. In the validation cohort, this threshold identified Lewis antigen-negative patients with 87.9% accuracy. The study also found that patients in this very low CA19-9 group had outcomes similar to those with markedly elevated CA19-9 levels above 200 units/mL.
When overall survival (OS) was analyzed by CA19-9 level, the longest survival was seen in the two intermediate groups: 23.2 months for patients with CA19-9 levels between 7 and 37 units/mL, and 22 months for those with levels above 37 to 200 units/mL. By comparison, patients with CA19-9 levels of 7 units/mL or less had a median OS of 13.5 months, close to the 12.8 months seen in patients with levels above 200 units/mL, suggesting that very low CA19-9 may help identify a high-risk subgroup rather than indicate lower disease burden.
The findings, published in Clinical Cancer Research on May 21, 2026, underscore that the standard “normal” range may mask aggressive disease in nonproducers and that a dual-threshold model may reduce misinterpretation. Reported limitations include the single-country setting and inter-laboratory variability in CA19-9 assays, with an international validation study planned as the next step.
“These data tell us that the conventional normal CA19-9 range of less than 37 does not distinguish between true low tumor burden and Lewis-negative status,” said Yung-Yeh Su, MD, Ph.D., assistant investigator and attending physician at the National Health Research Institutes, National Cheng Kung University Hospital, and Kaohsiung Medical University Hospital in Taiwan.
“Using a CA19-9 cutoff of 7, we achieved about 88% accuracy in identifying Lewis antigen-negative patients, whose prognosis is poor despite ‘normal’ or very low CA19-9 values,” added Dr. Su. "Using this cutoff was a practical surrogate for identifying these patients without the need for genotyping, thereby improving their risk stratification and preventing underestimation of their disease severity."
Related Links
National Health Research Institutes
National Cheng Kung University Hospital
Kaohsiung Medical University Hospital
.jpg)
