NEW YORK – Caris Life Sciences is hoping that a new publication helps make the case for broader adoption of a chemotherapy personalization algorithm it launched this spring called MI FOLFOXai.
The predictor, which uses data from Caris' 592-gene tumor profiling panel, is designed to identify colorectal cancer patients who are likely to benefit from first-line treatment with FOLFOX in combination with the anti-VEGF agent bevacizumab (Genentech's Avastin), as opposed to other chemotherapy regimens available in this setting, such as FOLFIRI or FOLFIRINOX, which combines both.
MI FOLFOXai is somewhat unique in the precision oncology arsenal. Most of the hallmark genomic and molecular tools that have been developed have been in the context of targeted therapies: companion or complementary diagnostics that identify patients eligible for a particular drug or drug class.
Other prominent precision oncology assays help predict patient prognosis or overall likelihood of responding to chemotherapy, but rarely offer a way to help physicians choose which specific chemo regimen to use.
Absent methods to predict individual patients' responses, FOLFOX and FOLFIRI have been viewed as largely equivalent in their first-line efficacy for advanced colorectal cancer. And while the combination of the two, FOLFIRINOX, has shown itself to offer improved progression-free survival, the double hit is associated with toxicity and side effects that many patients cannot tolerate.
All metastatic CRC patients generally receive both treatments either combined or in sequence with decisions about which to use first resting on side-effect profiles or other clinical concerns.
Caris launched MI FOLFOXai commercially several months ago, and President and Chief Scientific Officer David Spetzler said this week that the new publication in Clinical Cancer Research will now be key in driving adoption.
So far, those utilizing the algorithm have been mainly physicians that know the company and the data behind this predictor well. "Now that we have the publication we can actually provide [the larger community] with that information and they know that it's been well scrutinized and thoroughly reviewed and now they can take it seriously," Spetzler said. "I think we'll find out what the uptake is and how many people are beginning to use it [in coming months] and we'll see what happens," he added.
In the study, Caris researchers reported that they initially trained the MI FOLFOXai algorithm using a real-world evidence dataset that included 63 patients recorded to have had increased benefit from FOLFOX and 42 who had decreased benefit. A fivefold cross validation showed that a 67-gene signature could consistently separate increased versus decreased benefit patients.
To further validate the predictor, the team applied it in another independent group of real-world samples: 412 patients who had received first-line FOLFOX/BV and 55 patients who had received first-line FOLFIRI. For 50 patients, the model did not predict either increased or decreased benefit, but among the rest, there was a significant different in outcome. Patients predicted to have greater benefit by MI FOLFOXai survived a median of 42 months, while those predicted to have little benefit had an average overall survival of only 24.5 months.
The signature also showed good specificity when applied to the 55 patients in the cohort treated with first-line FOLFIRI. These patients had inverted survival curves based on their predicted FOLFOX benefit. In other words, patients predicted to have a higher FOLFOX benefit had worse outcomes on FOLFIRI (18.7 months median overall survival) compared to those predicted to have lower FOLFOX benefit (34.4 months median OS).
Finally, investigators performed a second, blinded validation using 271 samples from a trial called TRIBE2. Authors reported that MI FOLFOXai yielded a call for all but 35 individuals. The algorithm was predictive of OS in both oxaliplatin-containing arms of the trial — for FOLFOX and FOLFOXIRI.
"Currently the standard of care is that … most patients either get either FOLFOX or FOLFIRI first, and then they'll get the other one second," Spetzler said, reflecting on the utility implied by the study results.
"We're not saying don't give one or the other, we're just saying the order that they should be given … and I think it's remarkable to think about the life extension that simply sequencing these two chemotherapy regimens differently can give patients [based on what we saw in the study,]" he added.
When it launched MI FOLFOXai this May, Caris didn't detail any plans to potentially extend the predictor to other cancer types where similar dilemmas exist for decision-making regarding platinum-based chemotherapy options.
But in the newly published study the authors reported that they have also investigated the signature's potential to predict benefit from oxaliplatin-containing regimens in advanced esophageal cancers and pancreatic ductal adenocarcinoma, again using real-world data.
"We were wondering why this wouldn't this work elsewhere," said Jim Abraham, Caris' senior vice president and chief data officer. "What could there be about colorectal cancer that would make it only work there?"
According to Abraham, the data so far for pancreatic and esophageal cancer is only preliminary. "We're not making these claims of clinical validation ... because we don't have blinded samples but we're going to be doing that [and] all the data suggest so far that this is a pan-tumor platinum predictor," he said.
In addition to pancreatic cancer, the company is also looking actively at the ovarian cancer space, and Spetzler argued that there could be potential utility across any tumor type where similar drugs are used.
"Depending upon the particular clinical setting, there may or may not be alternative standard-of-care options," he said. "But even in the absence of that, if a physician knows that this patient is unlikely to respond well to a platinum-based therapy, while they'll still probably have to give it to them just to stay compliant with guidelines, they may increase the rate or frequency of seeing them, paying more attention to potential side effects to move them on to second-line faster."
For colorectal cancer, a randomized prospective trial would be an ideal follow up to cement that MI FOLFOXai actually improves patient outcomes. But Spetzler said that ethical issues preclude that as a next step.
"We've had lots of conversations with various groups … and here's the fundamental problem … you'd have to have patients where [our test] said they should get FOLFOX first and intentionally give half of those people FOLFIRI first with the fundamental knowledge and belief [based on our current data] that that they are going to live 17 months shorter because of that decision. You're actually doing harm to the patient as part of the trial. And that is just completely unethical," he said.
Instead, Caris has created a prospective registry that is tracking patients over time to accumulate more data, which will hopefully be sufficient to convince clinicians that the use of this predictor offers real benefit.
One hurdle the company doesn't have to overcome is reimbursement, since MI FOLFOXai is not billed separately from the Caris' comprehensive tumor profiling panel used to calculate it.
"There's no additional cost to this … [because] it's coming off of already approved testing. So we're not changing our reimbursement strategy around this at all — just trying to maximize the amount of clinical utility that physicians and patients get out of our offering," Spetzler said.
However, for patients to get access they have to actually get sequenced, something that continues to be a challenge for the field despite progress in guideline endorsement of comprehensive genomic profiling, increased regulatory clearance of next-generation sequencing assays, and the resulting institution of public and private payor coverage policies.
Looking to the future, Spetzler and Abraham said that the company believes that it will be able to use its MI platform to develop additional precision oncology tools.
"We are doing whole-exome and whole-transcriptome sequencing on every patient and the thinking is that we're going to find more of these," Spetzler argued.
"We think we can move way beyond therapy selection," he added, highlighting research the firm has conducted looking at distinguishing patients who will develop specific types of outcomes like brain metastases, as well as possibilities across other areas of drug decision-making outside of the FOLFOX/FOLFIRI niche.
"The process of intervening in the growth of a tumor, whether it's through targeted therapy, IO, chemotherapy, it's all about modification of a very complex system. And so, the more we understand about the current state of that system, the better we're going to tailor the choice of appropriate intervention."
"We see this methodology working not just for chemo, but for IO, for IO-chemo combinations... There's no therapy out there that is 100 percent effective in the patient population in which it's given," he said.
Finally, Spetzler added, Caris also envisions potential utility for new molecular signatures used serially, or at multiple timepoints during treatment. For that, the company would have to be able to translate signatures generated by its molecular intelligence platform to liquid biopsy, which is something it is also working on now.