NEW YORK (GenomeWeb) – Exact Sciences, the company that's already well known for its at-home stool-based colorectal cancer test, has set its sights on improving the early detection of pancreatic cancer, a disease that is often diagnosed too late for most patients.
During a recent earnings call, Exact CEO Kevin Conroy highlighted that the company was conducting a study in collaboration with the Mayo Clinic for pancreatic cancer detection from blood. Although this test is several years away from commercialization, data from that collaborative study recently published in the journal Gastroenterology shows how researchers are trying to improve diagnosis of early-stage patients compared to currently available methods.
According to the nonprofit Hirshberg Foundation for Pancreatic Cancer Research, an estimated 56,770 people will be diagnosed with pancreatic cancer in the US in 2019, and more than 45,750 will die from the disease. Pancreatic cancer is the third leading cause of cancer-related deaths in the US and is expected to become the second leading cause by 2020.
That's mostly because a large majority of pancreatic cancer cases aren't caught until the disease is already in its later stages and much harder to treat. And the reason for that is twofold. Although the common symptoms for pancreatic cancer can include yellow skin, abdominal or back pain, or unexplained weight loss, these signs usually don't appear in the disease's early stages or aren't specific enough to suggest cancer. Further, the biomarker that is currently used to test for the presence of pancreatic cancer, carbohydrate antigen 19-9 (CA 19-9), is inconsistent.
"One of the real-world struggles with this disease is that we do not have good tools for early detection, and we know that if the disease is detected early it results in better outcomes for our patients," said Shounak Majumder, a gastroenterologist at the Mayo Clinic and first author on the Gastroenterology study. "The only currently available blood test for pancreatic cancer is CA 19-9, and we know that in up to 10 to 15 percent of patients with pancreatic cancer, the CA 19-9 values are normal or undetectable. It's not always a very reliable test in that setting."
Patients who lack a blood type antigen on red blood cells called Lewis antigen A, which is about 10 percent of the Caucasian population, don't produce CA 19-9 in any of their cells. In fact, the American Society of Clinical Oncology's current guidelines for pancreatic cancer testing, issued in November 2018, state that there are no approved biomarkers for screening and surveillance of the disease, and that "CA 19-9 is not recommended as a screening test in the general population due to low specificity and sensitivity."
With those difficulties in mind, Majumder and his colleagues began researching possible alternative biomarkers several years ago, beginning with experiments looking at cancerous and normal pancreatic tissue taken from patients, comparing them to identify any methylated DNA biomarkers that could help discriminate between cancer and normal. They eventually moved on to do a similar experiment in pancreatic cyst fluid.
These analyses yielded 13 candidate novel methylated DNA markers — GRIN2D, CD1D, ZNF781, FER1L4, RYR2, CLEC11A, AK055957, LRRC4, GH05J042948, HOXA1, PRKCB, SHISA9, and NTRK3 — that the researchers found to be strongly associated with pancreatic ductal adenocarcinoma (PDAC), the most common subtype of pancreatic cancer. For the recently published paper, the team used 170 archival plasma samples from pancreatic cancer patients who had donated samples to the Mayo Clinic biorepository and 170 control samples.
Of those 340 plasma samples, they tested for the 13 methylated DNA markers and for CA 19-9 levels in 120 control cases and 120 advanced stage PDAC cases, 60 of which were Stage 3 and 60 of which were Stage 4. The researchers were aiming to train a prediction algorithm to detect pancreatic cancer in these samples using these specific markers at 97.5 percent specificity. They then applied a locked algorithm to an independent blinded test set of 50 early-stage PDAC cases — five Stage 1 cases and 45 Stage 2 cases — and 50 controls.
Subsequently, data from all 340 patients was recombined. The researchers then applied a cross-validation approach by randomly splitting the entire data set into two groups for training and testing and putting the algorithm through 500 repetitions with the training set.
In the initial training set, the combined methylated DNA marker CA 19-9 panel detected 90 percent of Stage 3 and 98 percent Stage 4 PDACs at 97.5 percent specificity, the authors reported. Combining all 340 cases and controls, the cross-validated sensitivity of the panel was 79 percent in Stage 1, 82 percent in Stage 2, 94 percent in Stage 3 and 99 percent in Stage 4 PDAC at a specificity of 92 percent.
"Across all stages, the combination of the methylation markers and CA 19-9 gave us a sensitivity of 92 percent at 92 percent specificity, and the cross-validated area-under-the-curve for the combination was .97," Majumder said. "This was significantly higher than the AUC for either CA 19-9 by itself or the methylation markers by themselves."
Given the results, the team is currently conducting a separate tissue-discovery study to see if it can identify the DNA methylation markers underlying the other major pancreatic cancer subtype — pancreatic neuroendocrine tumors.
Despite the positive results, though, there's still a way to go before people at risk for pancreatic cancer can start asking their doctors for a blood test. For one thing, Majumder noted, the sensitivity of the panel in the earlier stages was much lower partly because of a lack of Stage 1 and Stage 2 samples.
"We are limited by the access to some of these really early-stage samples. Pancreatic cancer is typically detected late, so even in a biorepository that collects patient samples, the sample pool is heavily tilted in favor of more advanced disease," he said. "One of the key next steps is to do a prospective study and include more early-stage cases, to further validate the performance of this biomarker panel."
The researchers are currently conducting such a validation study, using more early-stage pancreatic cancer samples. Majumder estimated that the team will likely need several hundred samples, most of which will come from Mayo's repository. He noted that the validation study could take up to two years, and that further testing would likely be needed after that to ensure that the panel has been robustly tested for safety and efficacy.
Indeed, in an email, Conroy said that a potential launch and commercialization of an actual test "are likely years away."
If a test does materialize, however, Majumder has some ideas of how it could be deployed in the clinic. "We have multiple possible opportunities here for clinical applications. One of them would be as a screening blood test in patients who are at higher-than-average risk for pancreatic cancer," he said. "There are many risk factors that put an individual at a higher-than-average risk, and our attempts to screen and identify cancer early in this cohort have remained somewhat suboptimal. So, a marker like this, if validated in the larger pool of early-stage cancers, can be informative."
Conroy agreed, noting that because no major guidelines recommend routine pancreatic cancer screening for average-risk patients, part of Exact's collaboration with Mayo is to determine the clinical utility of a possible test. "This test would likely be used in a high-risk population," he added.
Indeed, the population that could benefit from such a test is larger than it might seem. "[Other than] family history, we know that new-onset diabetes is a risk factor for pancreatic cancer. We know that the presence of germline mutations is a risk factor for pancreatic cancer," Majumder said. "So, there are many things that put our patients at a higher-than-average risk, and the idea is that in that high-risk group, this blood test will eventually serve as an early detection tool."
He further noted that such a biomarker test could be used for disease monitoring as well. "Oftentimes, we'll treat pancreatic cancer patients with chemotherapy and try to make decisions about how long to treat them, and we'll often end up using CA 19-9 as the guide in that setting," he said. "But our surgeons and oncologists are always looking for novel biomarkers that can help them assess disease activity. And so, a biomarker like this can fit into that space and add to the level of care that we're providing for our patients."