February 11, 2025 | Two glycans are better than one when it comes to detecting pancreatic cancer that currently lacks a biomarker capable of accurately and reliably spotting the disease. The gold standard biomarker CA19-9, approved by the U.S. Food and Drug Administration (FDA) for monitoring response to treatment, correctly identifies only 44% of cases but the addition of CA199.STRA ramped that up to 71% while maintaining a low false positive rate in a recent case control study.
The hope is that the dual biomarker assay will find clinical utility in another few years at an early enough point in the disease process to give patients the possibility of a cure, according to Randall E. Brand, M.D., professor of medicine at the University of Pittsburgh, academic director of the gastrointestinal division of UPMC Shadyside and co-lead on the rigorous, multi-laboratory study published in Cancer Letters (DOI: 10.1016/j.canlet.2024.217245). In clinical testing to date “things are definitely trending in the right direction,” says Brand.
CA199.STRA was identified a decade ago in the lab of Brian Haab, Ph.D., professor of cell biology at the Van Andel Institute, who was acting on a hunch that glycans structurally related to CA19-9 might also be elevated in pancreatic cancer patients. He subsequently developed and patented the assay technology that involves an enzymatic preconditioning step to modify the glycan to make it detectable.
The results of this study have shown that by itself, CA199.STRA functions “at least as well and probably better than the existing gold standard biomarker, CA19-9,” says Brand. The exceptionally high performance of the new two-marker assay has been repeatedly demonstrated in different settings, including at ReligenDx, certified under the Clinical Laboratory Improvement Amendments (CLIA) program.
ReligenDx recently licensed the technology and will continue its validation work. The CLIA lab spent several years reproducing results achieved with Haab’s antibody microarray methods on its own microplate assays. This initial testing phase included control samples but expanded to 200 “brand new samples” sent to the lab in a blinded fashion, says Haab.
The next step for Haab and Brand is to find out how the dual assay works for the intended clinical use. To that end, they have started testing samples from high-risk populations for developing pancreatic cancer including patients with hereditary predisposition for developing this malignancy or pancreatic cysts.
The goal here is to be able to detect pancreatic cancer before symptoms appear, says Haab. But addressing that question with research on people who don’t have cancer would require a huge study because only a small percentage of them would eventually get it over a long period of time.
Published studies to date have all been done using samples taken at the time of diagnosis. Researchers are now moving to the analysis of samples collected from patients in large screening studies from a few months to a few years prior to their cancer diagnosis. “You can do a reasonably sized study and still look at the value of detecting in blood the presence of cancer before it is clinically evident,” Haab says.
When tumors are smaller, they don’t secrete as many of the molecules of diagnostic interest, including CA19-9, adds Haab. It follows that the percentage of patients showing biomarker elevations is smaller the further ahead of diagnosis they are being measured.
In the screening setting, a test need not be perfect but good enough to justify it being used—meaning, better than existing options. Encouragingly, he says, early unpublished studies show performance improvement with the addition of CA199.STRA holding up in the pre-diagnostic samples.
Haab and Brand have been working together since meeting more than two decades ago through the Early Detection Research Network (EDRN) of the National Cancer Institute (NCI). EDRN is a consortium of labs interested in the topic that regularly brings together clinicians, basic scientists, statisticians, and people with innovative technology, says Haab.
One of Haab’s specialties is using monoclonal antibodies to detect antigens, notably CA199.STRA, which is not detectable by the existing CA19-9 assay because “any difference in that glycan causes the antibody not to bind.” This microarray technology enabled Haab and his team to do parallel screening of different combinations of antibodies and other proteins that bind glycans to identify CA199.STRA as the “top hit.”
Brand says that Haab has worked diligently in his efforts to make CA199.STRA a commercially viable biomarker that is reproducible without depending on a given tool. Haab has established standardized protocols and procedures to prevent the “drift” in results that have been produced over time with other biomarker assays even when analyzing the same sample.
Labs are run much differently in the commercial world than the academic world, notes Haab. “We realized early on that you need a really good assay to do this work because otherwise it is just going to eat through your results. You’re not going to see the differences between the cases and controls, and you’re not going to reproduce it,” pointing back to the partnership with ReligenDx.
As a diagnostics company that does this for a living, Haab says, ReligenDx follows its own set of standards to guarantee reproducibility over time. It is also set up to run prospective studies with real-time results and the ability to handle private patient data.
Finding a suitable CLIA lab was admittedly challenging, says Haab. Attempted collaborations with two other certified labs proved not to be a great fit because other projects were being prioritized. ReligenDx was of the size not to be overwhelmed with other work and interested in a promising new test that could be a moneymaker for them in the future.
Brand, like most physicians, doesn’t use CA19-9 for early detection of pancreatic cancer due to its suboptimal performance characteristics. “There is a subset of the population that doesn’t make it, and that could be 5% to 10% or even higher depending on your ethnic background,” he says. “Another subset of the population secretes CA19-9 at an artificially high level and get falsely flagged as having cancer since the cutoff value for separating the positives from the negatives will generally be lower for early detection purposes.”
The specificity of the old standard biomarker certainly isn’t perfect. CA19-9 levels can rise from multiple causes beyond cancer, Brand continues, including patients with other types of cancers, or benign disease such as a patient with jaundice due to a gallstone blocking their bile duct.
“We don’t have a great biomarker for pancreatic cancer; this is what drives us to find a better biomarker,” he adds. To overcome the challenges in detecting the biological variability of pancreatic cancers, the research team is leaning toward a panel of biomarkers for early detection.
CA19-9 and CA199.STRA work better together because different patients make high levels of each one, Haab says, while others are high in both biomarkers. “That has been shown very clearly in studies of the tumors... even at the cell level.”
CA199.STRA was found to give the most additive performance when combined with the base CA19-9 biomarker in the latest study, based on the sensitivity test—the ability to detect the highest number of true positives—while not adding any false positives, says Haab. But there were a few other previously discovered biomarkers that showed additive diagnostic potential, including the proteins THSP2 and LRG1, and are deserving of further study.
“Based on the unique biology of a tumor, it's possible that some of these other biomarkers could pick up a few cancers that CA199.STRA would miss,” says Brand, getting back to the value of a panel test. “The problem is if you add more biomarkers, you run the risk of picking up a lot more false positives, so it affects your specificity. There is always something that you sacrifice when you expand,” he notes.
From a practical standpoint, adding more biomarkers can also “exponentially increase” the noise with the assay, says Haab. “If you can keep the number of markers in your panel as low as possible, and still get good performance, that’s very advantageous and it keeps the cost down.”
With a greater number of biomarkers in an assay comes the inherent risk of “overfitting,” an issue that in the latest study fell to Fred Hutchinson biostatistician Ying Huang to monitor. It is noteworthy that the performance characteristics for almost all the panels (up to four biomarkers) were highly aligned between the training and validation sets, says Brand, which is quite unusual.
It is more typical to see a 5% to 10% decline in performance when a biomarker is being independently validated, he says.
In terms of go-to-market plans, the first option is the two-marker CA19-9/CA199.STRA panel, since any additions would need further laboratory validation, says Haab. “Certainly, that’s a possibility going forward. If other biomarkers show value in specific settings, then they can be added.”
The goal with the clinical assay currently is to validate its use for surveillance of patients who are at elevated risk for developing pancreatic cancer or incipient pancreatic cancer, he continues. “They could potentially have a blood test to evaluate whether they would go on to some sort of an imaging evaluation."
If all goes well, it would be offered for this purpose in the CLIA lab and, as the data builds, provide support for FDA approval in another few years. The research team is looking to launch phase 4 trials to show why the dual biomarker assay deserves adoption and can therefore change medical practice, says Brand. “It can be as simple as using less resources on imaging.”
It all circles back to the patients. “We recognize that pancreatic cancer is a challenging disease to treat and cure, and we are working hard to give more people the opportunity to identify it at an earlier stage when they are asymptomatic and the tumors are confined to the pancreas and potentially curable,” says Brand.