NEW YORK – New validation data this month have demonstrated that testing for certain gene variants that affect the synthesis of a commonly used cancer protein biomarker, CA19-9, can improve the marker's performance in the early detection of pancreatic cancer by creating personalized reference ranges for each individual being tested.
The findings, published in Clinical Cancer Research, highlight a unique and relatively simple methodology, especially amidst a huge research and commercialization push for the development of much more complex cancer screening tools using circulating cell-free DNA.
According to Michael Goggins, the study's senior author and professor of pathology at Johns Hopkins University School of Medicine, the basic knowledge of the gene variants that affect the synthesis and circulation levels of CA19-9 has been around for decades. "There's an occasional variant identified now and then in an isolated population, but the synthetic pathway has been known for many, many years," he said.
What hasn't been appreciated, he explained, is how to use this clinically to classify people into functional groups with respect to these variants.
"People have done many studies identifying new biomarkers, often genetic variants, but having it matter to a significant percentage of the population [is a challenge] … because the magnitude is small. But here you have common variants having a significant effect on synthesis and levels in circulation," Goggins said.
Goggins and his team had explored using SNPs to tune reference ranges for several different protein markers in prior work, but they didn't see the same impact for CEA and CA-125 that they did for CA19-9, so they decided to focus on that marker and the variants in the fucosyltransferase (FUT) enzymes FUT3 and FUT2 that affect it.
Past studies had attempted to use the protein for pancreatic cancer early detection, but it didn't show high enough specificity to be applied clinically.
"For us, there was an understanding, despite issues regarding specificity, that CA19-9 had sensitivity characteristics that could lead to good performance and so that eventually led to this moment of considering how individualized reference ranges might improve things," Goggins said.
"There are people walking around who have tumors and maybe they're shedding a little bit of CA19-9, but they're still within the 'normal range' … which means there are people whose [individual] normal range is probably below what the conventional normal range would be," he added.
In their study, Goggins and his colleagues developed what they call the "CA19-9 tumor marker gene test," sequencing samples from control subjects enrolled in their ongoing Cancer of the Pancreas Screening (CAPS) study to establish clinical laboratory reference ranges for CA19-9 for four functional FUT groups, (FUT3-null, FUT-low, FUT-intermediate, and FUT-high), which reflect the predicted function of each FUT3 allele and their FUT2 functional status.
They then assessed diagnostic performance first in a testing set of 234 pancreatic cancer cases, and then a 134-case validation set, all of whom had undergone resection with curative intent without neoadjuvant therapy, comparing their personalized FUT reference range method to the standard CA19-9 cutoff.
Using the standard cutoff yielded a 68.8 percent sensitivity in the test set and 67.2 percent in the validation set, at 91.4 percent specificity.
Applying 99th percentile cutoffs according to each individual's FUT group, the team reported 66.7 percent sensitivity in the training group and 65.7 percent sensitivity in the validation set at a much-improved 98 percent specificity. In a receiver operating curve analysis, testing using a conventional CA19-9 cutoff showed an area under the curve of 0.84, compared to 0.92 for the personalized method.
Overall, the tumor marker gene test reclassified the CA19-9 result in about 7 percent of the PDAC cases compared with CA19-9 alone — 10 from negative to positive and 17 from positive to negative. Among controls, 8.6 percent were reclassified, mostly from false-positive to true negative.
Goggins and his coauthors wrote that their results are promising enough to merit evaluation of their genetics-informed CA19-9 testing method in prospective studies enrolling high-risk individuals undergoing imaging surveillance.
The group has already begun doing this in its CAPS study, where Goggins said the team already has close to a thousand people recruited at Hopkins, with a rollout underway to other institutions.
"The idea is that it's not a standalone test for the people who come to us. The people in the study are those who meet criteria for regular, typically annual, pancreatic surveillance," he said.
"Scans have superiority in detecting small tumors so this is going to provide us an opportunity to evaluate the performance," Goggins added, "But there are people who we'll see after a year and they have a stage II tumor or they have breakthrough symptoms six months out. Thankfully, that doesn't happen that much, but having a blood test that you could eventually have between scans would be valuable. That's part of what the prospective study is trying to do."
Regarding the potential to boost sensitivity further, Goggins said that exploring complementary biomarkers and refining or enhancing FUT-based binning would make sense.
"We do have a pretty good understanding of the factors affecting CA19-9 but it's possible that additional refinements would [help]," he noted. "But truly, there is some limit. Some patients have small tumors that happen to shed enough CA19-9, but others won't until they get bigger."
The team's boosted CA19-9 method could also potentially improve sensitivity for pancreatic cancers in the context of multi-cancer screening tests now being developed using epigenetic and other technologies.
With the intent for these tests to be applied to the general population, specificity is especially crucial. But with assays tuned to maximize specificity, sensitivity can suffer, especially for early-stage tumors and for cancer types that tend to shed less DNA into the blood, one of which is pancreatic cancer.
Goggins highlighted that one of these multi-cancer screening tests, CancerSEEK, did already incorporate protein markers, including CA19-9, so that could potentially be a direct application for his team's FUT binning method.