NEW YORK – German RNA-based testing firm Hummingbird Diagnostics is making strides to develop a microRNA assay for the early detection of lung cancer.
Last month, the company published promising data on the development of an 18-miRNA signature dubbed miLung in the Journal of Thoracic Oncology.
Hummingbird CSO Bruno Steinkraus said that the firm is very focused on early detection, even though its first foray into diagnostics was the development of a cancer immunotherapy response predictor.
"The [immunotherapy application] was more of a side project, but it was significantly easier to do," he said. "With early detection, there is very little signal in the peripheral blood, and hence it is a much harder task."
"In our overall strategy, [early detection] is the more important program. If we can prevent cancer in the first place or find it early, it's much better than to wait till it's there and then optimize the treatment regimen," Steinkraus added. "That is also important in itself of course. But our long-term goal is to find it early when it can be cured by surgery."
Although the approach taken in each project was similar, the team had to block "significantly more RNAs to essentially unearth the biomarkers required for early detection."
This blocking of background microRNA from red blood cells is a core component of the company's technology, Steinkraus said. Overall, the firm's approach combines this erythroid RNA removal step, ultra-deep small RNA sequencing, and machine learning tools to derive a predictive signature. In the immuno-oncology study last year, investigators blocked three miRNAs. For their new lung cancer detection effort, they were able to block 20 highly abundant molecules, "opening bandwidth for the detection of less abundant species" necessary for the detection of early-stage cancers.
With the new lung cancer data, Hummingbird believes it has overcome this challenge, demonstrating for the first time that its RNA approach has promise to supplant currently used imaging tools to detect early-stage cancers in high-risk individuals.
The multicenter study involved collaborators from Massachusetts General Hospital, Beth Israel Deaconess Medical Center, LungenClinic Grosshansdorf, and University Medicine Essen.
Overall, investigators screened whole blood samples from 1,384 individuals meeting risk-based eligibility criteria for low-dose CT screening. In an initial training set comprising patients diagnosed with lung cancer and controls who met criteria for screening but had no diagnosis, Hummingbird used machine learning analysis to develop an 18-marker signature of small RNA targets. Validated in an independent cohort of more than 400 additional cases and controls, the assay performed well, with accuracy around 83 percent based on an area under the receiver operating curve of 0.83.
This performance suggests viability for the signature as an alternative to low-density CT, the authors wrote, and one that could be deployed in the primary care setting where it could ease access and improve uptake of screening in high-risk populations. Overall, the team estimated that of 9 million eligible Americans only 300,000 take part in lung cancer screening.
They added that while lung cancer screening programs have not yet been implemented in Europe, they would likely face similar participation challenges. "A blood test, if integrated into routine annual checkups within primary care, could significantly lower the barrier to participation," the authors wrote.
According to Steinkraus, this implied improvement in adoption counterbalances the possibility that its test performance may fall short of what has been reported for CT imaging in the clinic.
"The problem with low-density CT is that it's not being done," he said, citing Exact Sciences' Cologuard stool test as an analog. "It isn't as good as colonoscopy. However, the overall impact is significant because people are using it."
At the recent annual meeting of the Association for Diagnostics & Laboratory Medicine (formerly the American Association for Clinical Chemistry), Hummingbird also shared new data featuring a version of miLung that uses digital PCR rather than NGS. One advantage for the firm over potential competitors employing genome-wide methylation and fragmentation technologies for early cancer detection is that it has focused on small RNA signatures, which can be assayed on cheaper and more ubiquitous platforms.
Companies developing molecular cancer tests for early detection have largely been pursuing either individual cancer indications or pan-cancer screening tests. For multi-cancer screening, Steinkraus said, the emphasis has to be on specificity because of the negative impact of false positives. But in single-cancer screening, especially where tests are directed toward known high-risk populations, high sensitivity is the more important factor.
One likely player in the future market for molecular early detection tests is Delfi Diagnostics, which has also targeted lung cancer. That firm's early data for its test showed similar performance to miLung, although new data shared this week has boosted its accuracy to 93 percent.
Steinkraus highlighted Hummingbird's larger cohort, and its efforts to ensure a more relevant control arm of unhealthy smokers rather than healthy controls, as advantages of its own results.
He added that Hummingbird is currently setting up a validation study, which will recruit 2,000 US patients — in the same high-risk population — across 15 to 20 sites.
The company also has no intention to abandon its immunotherapy predictor and is working with the German Lung Cancer Network to validate that test in a specific, PD-L1-high population. "There is currently clinical equipoise whether these PD-L1-high patients should get immunotherapy alone or whether they should get immunotherapy combined with chemo, so we are setting up a trial to address exactly that," Steinkraus said.
The immuno-oncology application likely offers an easier path to the clinic, he added, but the lung cancer screening validation should proceed quickly. All in all, the firm aims to have data in hand within about two years to advance both tests commercially.
In the current lung screening cohort, Hummingbird also reported an interesting finding that could potentially open up a cancer monitoring application. Among the 18 RNAs in its signature, one, a plasma-bound ribosomal RNA fragment of the L1 stalk, appeared to be providing the bulk of detection sensitivity. When the investigators looked at blood samples from cancer-positive patients taken after surgery, the abundance of this specific fragment dropped significantly.
For one, this boosts the company's confidence that the signal it is detecting is indeed coming from lung tumors. It also suggests that this molecule could be used to monitor patients for recurrence.
Although Hummingbird's investigators believe that tumor-derived RNAs do account for a significant proportion of the small RNAs in its panel, the combination of tumor- and immune cell-derived molecules is a crucial part of its strategy. As evidence of this, the authors reported that the tumor-linked ribosomal fragment was unable to distinguish stage I and II cancers from controls when assayed on its own.
Finally, the study also provided a proof of concept that Hummingbird's test could be performed on dried-blood spots instead of frozen samples, which would make adoption even easier across primary care settings, especially those with lower financial and technological resources, and could enable at-home sampling if a monitoring application does emerge.