NEW YORK – An international group led by the City of Hope has built a transcriptomic signature that identifies lymph node metastasis in high-risk submucosal colorectal cancer, or T1 CRC, patients, with the goal to help minimize unnecessary surgeries and potential costs following treatment.
The team will commercialize the method as a multiplex, reverse-transcriptase quantitative PCR assay that uses less than a milliliter of a CRC patient's blood sample to establish metastatic risk.
Oncologists typically recommend that patients with aggressive, early-stage CRC to undergo radical surgery — which removes most of the patient's colon — because of their metastatic risk. Current criteria for stratifying CRC patient risk include identifying the depth of submucosal invasion, the presence of lymphatic or vascular invasion, high-grade tumor budding, and a poorly differentiated tumor cell history.
If the patient has at least one of the pathological risk factors, they are considered high risk and advised to undergo surgery. However, Ajay Goel, chair of the molecular diagnostics and experimental therapeutics department at City of Hope, argued that the criteria may not be entirely accurate and can lead to overtreatment and long-term damage.
He also pointed out that colon tissue biopsies are typically only performed once or twice prior to making a clinical decision and can be highly inaccurate as well. His team previously identified four microRNAs and additional gene expression biomarkers that they could use as part of a tissue-based assay to help identify lymph node metastasis, or LNM, in CRC patients.
"Any kind of biopsy is just a small piece of the puzzle … which may or may not reflect the environment," Goel explained. "If you take a biopsy from the wrong area, it might lead to a misdiagnosis and a false negative."
Because the miRNA biomarkers that Goel's team found would ideally be used to diagnose patients with high-risk T1 CRC prior to surgery (before the tissue samples are typically available), Goel therefore wanted to build a noninvasive blood-based panel that maintained relatively high accuracy.
"Biopsies are good, but not perfect," Goel said. "If you look at these biomarkers in the blood, you eliminate the [sampling] bias … and all physical intervention, which takes place when you do the procedure."
Goel explained that the assay begins by extracting RNA from less than 1 ml of a patient's blood sample using Qiagen's miRNeasy Kit. The platform then uses RT-qPCR on the RNA to separately identify markers of metastatic disease including four miRNAs (miR-181b, miR-193b, miR-195, miR-411) and five mRNAs (AMT, FOXA1, PIGR, MMP1, MMP9).
The blood-based panel requires three to six hours to analyze a patient's blood sample and generates a patient-specific report for a physician within roughly one business day, Goel said. The report may help the clinician decide whether to perform a colectomy, to hold off from removing the organ, or to follow up with an additional test every three to six months based on indeterminate results.
As part of a retrospective study published in Gastroenterology earlier this month, Goel and his colleagues looked at a cohort of 330 samples from Japanese patients with high-risk T1 CRC, including 188 serum samples from a training (46) and validation (142) cohort, and a matched formalin-fixed, paraffin-embedded cohort (142) obtained after endoscopic or surgical tumor resection.
The group performed RT-qPCR on the different sample types, followed by logistic regression analysis to build an integrated transcriptomic panel.
In the 46-training set, Goel's team ran expression profiling of LNM-positive and -negative serum samples to build an optimized transcriptomic panel using the four miRNAs and five mRNAs. The group saw that the nine-biomarker panel has an area under the operating curve of 0.86 in the cohort.
Goel and his colleagues then validated the panel on the 142-patient cohort, producing a sensitivity of 83 percent and specificity of 76 percent. The researchers further established a risk-stratification model combining the nine RNA-based biomarkers and pathological risk factors.
The group found that the integrated model was more accurate than the original panel and served as an independent identification predictor of metastasis, with a 90 percent sensitivity and 81 percent specificity.
"We did not expect the blood-based assay to be as good as the tumor biopsy test," Goel said. "We were surprised that the blood test was identical, or even slightly better than the tissue biopsies."
While all patients enrolled in the study were designated as high risk for LNM and received radical surgery, post-surgical pathologies indicated that only a total of 17 were lymph node-positive and 91 percent underwent unnecessary surgery. The new model limited potential overtreatment to only 18 percent of all patients, which Goel said was a clear improvement to currently used pathological features, which result in overtreatment of 92 percent.
However, Michael Stamos, dean of the school of medicine at the University of California, Irvine and who is not affiliated with the study, argued that Goel's team mistakenly indicated that the endpoint was based on lymph node positivity — what Stamos labels as a surrogate marker for outcome — rather than disease-free survival.
"Yes, lymph node positivity is important, [since] if you knew about it before you did the operation, it will influence your decision," Stamos said. "But if I'm going to operate on someone that has one of these early cancer markers … I tell them that statistically there's a 90 percent chance that if we do nothing, you're going to have a good outcome, unless there's some really bad prognostic feature."
In response, Goel said in a follow-up email that the first clinical endpoint for deciding treatment in T1 CRC patients is still LNM status.
"Obviously, if someone is LNM-positive, then the clinical endpoint would be poor overall or disease-free survival," Goel said. "In other words, both answers are correct and tied to each other."
While the assay reclassified about 75 percent of high-risk T1 CRC patients into low-risk groups, Stamos wonders about the consequences for patients placed into the low-risk groups who were labeled incorrectly.
"[The assay] is not 100 percent sensitive and specific, so don't forget those patients that are now misclassified don't get treatment, and that then succumb to the disease," Stamos said. "You need a balance between the [treatment] risk and benefit."
Goel recognized that his team's test is not perfect, but he emphasized that the study suggests that any clinical decision should be made with both the assay and current clinical standards to lower "this 80 percent inaccuracy to maybe 20 percent, which is a huge deal."
In addition, Stamos pointed out that there will be patients who benefit from a colectomy, though oncologists may not know that for sure "until you follow them for five to 10 years."
"If I were to do a colectomy on a T1 cancer patient, and the pathologist said there's no residual cancer in the colon and no lymph nodes involved … I tell the patient that this is great news," Stamos said. "However, that doesn't mean the surgery is unnecessary or unhelpful, [as] some of those patients will have actually had some positivity that was underdiagnosed because the pathologist can't possibly look at every lymph node."
Goel also acknowledged that his team faced several limitations while performing the study. Because the sample size was relatively low, the group evaluated the signature in a moderately sized patient group. Goel also said that the model's ability to risk-stratify patients will need to be further validated in a larger cohort.
Marian Waterman, deputy director of UC Irvine's Chao Family Comprehensive Cancer Center, highlighted that Goel's team did a "great job" of approaching the accuracy element of the assay within the study. However, she pointed out that every CRC patient's tumor is unique, especially between ethnic, age, and racial groups.
"This was a very uniform small subset, as we need to start somewhere, but the challenge is to broaden the applicability, and you will probably have to [use] different biomarkers and tissue landmarks or pathology … for combined risk stratification," Waterman, who was not affiliated with the study, explained.
Acknowledging the discrepancies between different populations, Goel said his team has launched a new trial to establish the risk model's utility in roughly 300 patients in Japan, Spain, and the US to potentially replicate the results of the retrospective study.
While the current model only looks at miRNAs, mRNAs, and a patient's clinical factors, Goel's team may also explore DNA mutations and consensus molecular subtypes in future studies to potentially update the signature. The group also expects to integrate the separate RT-qPCR assays into a multiplex version, which Goel said will "significantly" lower the cost of the panel.
Goel envisions oncologists eventually applying the transcriptomic panel in a similar manner to how prostate-specific antigen, or PSA, is used to screen for prostate cancer in high-risk patients. He believes the cost of the test will be dramatically lower than what patients would pay for colonoscopies and certain long-term financial impacts following radical surgeries.
"I see the future of this assay [akin] to PSA and other methods … where there is a noninvasive and inexpensive test that can be done on a lot of patients [and help] clinicians make informed decisions," Goel said. "We're not saying it's a replacement [for biopsy testing] but given the cost and noninvasive nature of the test, it can be adjunct to what we're doing in the clinic to help inform clinical decision making."
Goel's team has filed IP for the transcriptomic panel with the US Patent and Trademark Office and is working with potential biopharma partners to determine the best commercialization route. However, he declined to elaborate regarding further commercial plans.
While Goel was unable to provide an estimated price for the anticipated commercial assay, he noted that the current panel requires roughly $20 to $25 to perform per sample in the lab. He expects the cost to fall once his team multiplexes the platform by placing the primers together and pooling all the targets in a single PCR reaction.
Goel believes that the RT-qPCR assay is unique due to its focus on detecting RNA-based biomarkers for T1 CRC patients. While the researchers used Thermo Fisher Scientific RT-PCR machines in the study, he said his team is "not wedded" to the instrument and may potentially use another company's platform for the commercial version of the assay.
"There aren't too many transcriptomic-based assays in the liquid biopsy space out there," Goel said. "All liquid biopsy tests, to the best of my knowledge, are mostly DNA or protein-based, and nothing [is focused] around T1 CRC."
Waterman believes that the cancer diagnostic space is moving toward a broader use of liquid biopsy assays that "are sensitive, accurate," and offer "fast throughput," when combined with current cancer-specific risk stratification tools.
"Artificial intelligence coupled with bioengineering diagnostics, I think is the future … because you can get more [of a] asymptotic approach with improved accuracy," Waterman said.