NEW YORK – AI-driven cancer detection firm Freenome said this week that it has begun a prospective clinical trial to clinically validate the multi-omic and epigenomic test it has developed for blood-based detection of colorectal cancer and pre-cancerous lesions.
The trial, called PREEMPT CRC, will enroll 14,000 asymptomatic, average-risk patients between the ages of 45 and 85 undergoing a screening colonoscopy, and is intended to support a bid for premarket approval by the US Food and Drug Administration.
Gabriel Otte, Freenome's CEO, said that the company believes the trial, as designed, should end up being the largest of its kind ongoing in this setting.
"There is sort a reason for that," he added, which is that the company plans to seek approval for the test for the population beginning at age 45, rather than age 50.
"You want to have statistical significance, or at least a signal for every age group that you are actually testing and the big difference [with] starting at 45 and up as opposed to 50 [is that] the incidence rate between 45 and 50 is very different than the incidence above 50," Otte argued.
"It's far fewer people getting colorectal cancer age 45 to 50, so in order to make sure that we generate enough evidence to basically say 'this test does perform as expected in that age group,' it's incredibly important to have [a large-enough recruitment]," he added.
According to Otte, this is important not just for FDA approval, but also to make a case to payors and to convince the US Preventive Services Task Force to embrace a test for that shifted age group.
"While the American Cancer Society has lowered the age guidelines down to 45, most payors and the USPSTF still have it at 50," Otte said. "Right now, there hasn't been a study that's been done that is large enough that actually shows a utility in that population."
Freenome has yet to quantify in any detail how various analytes or biomarkers contribute to its cancer detection method, but it has reported that its approach relies on a combination of cancer-derived and non-cancer derived signals gleaned from DNA sequencing, bisulfite sequencing, and protein quantification methods.
The company's most recent data, shared at the American Society of Clinical Oncology's Gastrointestinal Cancers Symposium earlier this year, demonstrated 91 percent sensitivity overall, and an even higher 94 percent sensitivity for early-stage tumors (Stage I and II) at 94 percent specificity.
Notably, in announcing the new PREEMPT trial, Freenome hinted that it has also been able to tune its detection algorithm to detect not only cancers but also colorectal adenomas: cancer precursor lesions whose removal during colonoscopy has been shown to reduce individuals' risk of developing colorectal cancer down the line.
Detection of these adenomas is something rival firm CellMax Life has recently begun to highlight as a differentiator from potential competitors such as Freenome and Guardant Health. But Otte suggested this week that his firm is now also expecting its test to sensitively detect pre-cancerous growths.
Although Freenome has not shared any data yet on pre-cancerous lesion detection, Otte said the company has developed such a capability and has submitted some data to the FDA as part of its ongoing dialogue with the agency.
In its poster presentation at the ASCO GI meeting, the firm suggested that its internal data reflects what CellMax Life has also reported: that different biomarkers contribute to detection of precancerous polyps versus fully developed cancers — with the implication that methods that rely only on certain signals, like DNA methylation or mutations patterns, won't be able to detect precancers as well as cancers.
This unique contribution of different analytes is also something the company has seen in the context of detection of earlier versus later-stage cancers, Otte added. "We have [such] high sensitivity in the early stages, partly because those are different analytes than the ones that gives us high sensitivity in late stages."
Furthermore, Otte said, Freenome is confident that it can combine different signals without them detracting from one another or muddying the waters of the combined assay. For example, he said, "the ones that give us high sensitivity in late stages are largely tumor burden signatures and the ones that give us a performance in the early stages are largely immune-driven signatures. You can add those together and not have a performance loss, and it's no different for [what we are doing with] the advanced adenomas."
"It took us a little bit of time to figure out exactly how to combine the analytes so that it's not a direct tradeoff between sensitivity and specificity. But … if you do it the right way you don't see a loss in performance," he added.
According to Otte, although the company has locked in certain crucial variables of its method so that those can be kept consistent through the trial, cutoff points for various analytes can continue to be tuned to create the best possible sensitivity across cancer stages and now pre-cancers as the trial progresses.
"There are certain things that you definitely don't want to change — for example, the protocol on how the blood sample is collected …. and the analytes themselves — if you really want to make sure that the trial that you're doing is representative of what a test is actually going to be like in the real world. So that is what we have done," he said. "The thing that has not been locked down completely is, for example, the specific statistical cutoffs … that we're using … because we're continuing to generate new data."
For example, Otte said, after the firm read out its ASCO GI cohort in a blinded fashion, it then fed that new data back in to further train its detection model.
For precancerous adenomas, Otte said Freenome is lucky in that its approach is multi-omic, so even though the analytes have been locked in, they already include the signals necessary to detect these lesions.
Freenome hasn't provided a precise timeline for PREEMPT, but Otte said that the trial is starting with some established recruiting sites, with the goal of adding more sites over time.
"I think I could have given a more definite answer before COVID, but genuinely at this point it's really uncertain," he added. "Yes, shelter in place is lifting in a lot of places. Yes, elective procedures are starting back up in a lot of clinics. But does that mean that people are going to feel safe enough to actually go into the clinics to get a colonoscopy? How long does that take? We're trying to gather some data on that, but I ... would say we're in unknown territory."
As the company moves forward with PPREEMPT, Freenome is flanked by several direct competitors, including liquid biopsy firm Guardant, which announced its own 10,000-person registrational trial to support a bid to FDA last year, and CellMax Life, which has said it also intends to move forward with a multi-site prospective trial of its cell-capture, methylation, and mutation method in several thousand individuals.
Other firms like Grail and Thrive Earlier Detection are taking similar tacks, but for assays designed to screen for multiple cancers at one time.
Otte said that Freenome believes that an advantage of developing a test for the frontline cancer space is that assays must go through regulatory review if they want to be reliably reimbursed, meaning there will be equal, and hopefully contemporaneous review of the various tests now being developed and trialed in prospective cohorts. Based on how tests perform relative to one another they may be approved with certain restrictions in their intended use relative to others.
"FDA knows how to navigate multiple tests with different degrees of performance," Otte said, citing the example of Epigenomics and its Epi proColon test for colorectal cancer screening. "[That test] is FDA approved, but the label it got says basically that you have to say no to every other test first."
"I imagine something very similar happening" with some of the new generation of tests, he added.