NEW YORK (GenomeWeb) – In an encouraging development for the pancreatic cancer space, researchers have shown that using a broader liquid biopsy approach — including both proteomic and circulating DNA analysis — can significantly increase the number of early cancer cases detected in blood samples.
As researchers have come to recognize that treatable genotypes don't turn up equally often in all cancers, it has also become clear that that circulating tumor DNA is not as prevalent or detectable in all cancer types and stages.
And while certain tumor types like non-small cell lung cancer have become poster children for genomic profiling and personalized treatment — and also more recently for blood-based cancer detection — others, like pancreatic cancer, have not shown as much promies despite a grim prognosis and a great need for better treatments and earlier detection methods.
In a study published this week in the Proceedings of the National Academy of Sciences, researchers from Johns Hopkins University reported that combining circulating DNA sequencing with measurement of four blood-based protein biomarkers enabled them to detect early-stage pancreatic cancer in 64 percent of a cohort of patients, compared to only 30 percent for ctDNA analysis and 49 percent for protein analysis.
Bert Vogelstein, the study's principal investigator, said that while his and other groups have shown that they can detect circulating tumor DNA in more than 80 percent of patients with advanced cancer, early detection is still an untapped frontier for the field.
Moreover, Vogelstein said, though pancreatic cancer is one of the more difficult tumor types, all cancers pose problems for purely ctDNA-based screening.
For example, in a study in 2014 that Vogelstein was also a part of, researchers reported that even at sensitivities that would detect a single mutated DNA molecule in an entire blood sample, only up to 47 percent of stage I cancers appear to have detectable ctDNA.
"If it isn't in the plasma you can't detect it. It doesn't matter what your technology is," Vogelstein said. "So we knew from the get go that if you are going to try to get sensitivity up you are going to need to combine with something. In this case it's proteins, but that's not the only thing you can do. You can combine with any of the analytes people have been working on — transcripts, miRNA, glycoproteins, the whole variety."
In the study published this week, the Hopkins team designed a PCR-based assay targeting the two codons (12 and 61) of the KRAS gene that are most frequently mutated in pancreatic cancer, along with some surrounding sequences.
The team's approach relies on a technology called Safe-SeqS (Safe-Sequencing System), which was initially developed at Hopkins and uses a barcoding strategy to distinguish amplification or sequencing errors from real variants.
The same barcoding technique has yielded exciting results recently in colorectal cancer patients, for detecting signs of recurrence more quickly than existing clinical and imaging methods.
But in that recurrence monitoring setting, investigators and clinicians have prior knowledge of patients' tumor genetics that they can use to design more sensitive assays based on their individual mutational profiles. With true early detection, you don't know what mutations you are looking for ahead of time, Vogelstein explained.
Vogelstein and colleagues' Safe-SeqS KRAS ctDNA assay alone got them 66 of the 221 pancreatic cancer patients — just 30 percent. To try to boost this, the researchers looked to CA19-9, a protein biomarker that has showed promise for pancreatic cancer recurrence monitoring.
Because high CA19-9 can also be seen in individuals without cancer, its screening potential is potentially limited, but the team believed that it they set a high-enough threshold for positivity they could avoid false positives.
And indeed, using a threshold of 100 U/mL the team found that could detect CA19-9 in about half of the 221 pancreatic cancers without picking up any of 182 healthy controls. The overlap between the patients detected using the protein and using the KRAS assay was also only partial, suggesting that an approach combining the two would be even more sensitive.
"If you are going to do real early detection you can't tolerate anything less than near perfect specificity," Vogelstein said. "Because the number of false positives in a general population screening will be huge."
"Having high thresholds so there are no [false positives], that will be instrumental for combining with any other analytes I mentioned or that may be discovered in future," he added.
Encouraged by the results with CA19-9 alone, the researchers then looked to see if they could boost their detection rate even further by adding more proteins. After various analyses of individual sensitivity, the team added CEA, HGF, and OPN to CA19-9, and found that this multi-marker combination detected 141 cases — 64 percent — of the 221-patient cohort.
According to the researchers, patient characteristics offer interesting added insight. For example, about 20 percent of the subjects in the study had no pancreatic cancer symptoms. The combination assay picked up 60 percent of this subgroup, and nearly three-quarters of those patients remained disease-free at about 12 months follow up.
The test only picked up 41 percent of patients considered to have the earliest stage of disease as measured by clinical staging criteria. Almost 60 percent of these patients remained disease-free at the termination of the study.
A sobering extract from the data though, was that patients with poorer outcomes were more likely overall to test positive using the blood-based combination assay, suggesting that there remain significant hurdles to increasing the sensitivity of liquid biopsy to the point that it can detect cancers at a truly curable stage.
The JHU researchers are not alone in looking outside of circulating DNA to solve the problem of cancer screening.
A firm called OTraces recently said it is developing cancer screening tests based on a mathematic strategy of enhancing signals and suppressing noise for a collection of well-known and relatively well-characterized proteins.
And although companies like Freenome have highlighted ctDNA as the backbones of their planned products, the machine learning approaches being used by such firms are fundamentally agnostic, and so could potentially yield test strategies that combine multiple types of analytes.
For their part, Vogelstein and his colleagues are pushing forward with the combination of ctDNA and proteins or potentially other biomarkers.
In pancreatic cancer specifically, the researchers are now planning a study in which they hope to be able to test their approach prospectively in newly diagnosed type-2 diabetes patients, who have an elevated risk of developing pancreatic cancer.
The researchers are also working on a study extending their most recent publication, in which they hope to add additional genes beyond KRAS to broaden the assay to different cancer types.