NEW YORK (GenomeWeb) – Time-to-next treatment can act as a surrogate endpoint to predict cancer patient outcomes and help establish the clinical utility of tests that match tumor biomarkers to drugs, according to a Caris Life Sciences study.
As Caris Chief Scientific Officer David Spetzler noted, electronic medical records don't always capture survival data, but they do generally note when treatments occur, as that's how physicians bill. As a change to treatment could reflect a change in patient status, Spetzler and his colleagues reasoned that it could be a surrogate marker for overall survival.
In a poster presented at the American Society of Clinical Oncology meeting this week, the researchers reported that the time-to-next treatment measure largely reflected overall survival. In addition, they found that patients whose treatments were matched to their molecular profiling results — like from a service offered by Caris — had longer overall survival times and underwent fewer lines of treatment.
"We could see … a year increase in overall survival, and those patients in the matched group were actually getting one fewer lines of therapy," Spetzler told GenomeWeb. "So not only were they living longer, they were also less heavily treated, which of course is pretty important from a quality-of-life perspective."
The traditional model of determining the clinical utility of biomarker-drug associated relies on a study with four different arms. Such an approach, Spetzler said, isn't always feasible because it leads to hypersegmentation of the patient population, the need for potentially high numbers of patients to enroll in the study, and could bring up ethical concerns if patients would be assigned to receive a compound that is known to likely be of little benefit.
Instead, Spetzler said Caris has chosen to zoom out. Instead of examining individual biomarker-drug associations, company researchers are look more in general at how patients who receive a drug matched to their molecular profile fare as compared to patients who don't receive a matched drug.
"We lose some of our resolution, we can see the forest, [but] we can't pick out the individual trees," Spetzler said. "But it allows us to begin to define what that underlying utility of broad-based profiling is."
In this retrospective study, Spetzler and his colleagues examined some 1,180 patients with solid tumors who had been referred to Caris for testing between 2009 and 2015. Panomic testing — encompassing immunohistochemistry analysis, fluorescent in situ hybridization, next-generation sequencing, RNA-seq, and more — was performed on the patients as part of their clinical care. Caris then made recommendations to the clinicians on whether some 60 drugs were likely to benefit the patient or not.
The researchers retrospectively categorized the patients as being matched or unmatched based on whether they received drugs tailored to their tumor.
The patients who received matched treatments, they reported, had longer overall survivals — living more than a year longer than unmatched patients. They also tended to receive one less round of drug treatment, on average.
But, according to Spetzler, such survival data is difficult to come by. He said they spent seven years and millions of dollars to gather that data, and to ramp that up to an even larger cohort would quickly become infeasible economically.
Instead, he noted that time-to-next treatment data could be more easily extracted from EMRs.
"If you want to acquire outcomes data on patients in a relatively cheap and efficient manner, then time-to-next treatment becomes a very reasonable endpoint to go for," he said. He added, though, that it has its disadvantages as an endpoint as changes in treatment could reflect changing insurance coverage, toxicity, or patient preference, not just a progression event.
Still, in their cohort, time-to-next treatment — which they defined as the interval between start of first treatment after tissue collection and next line of treatment — for matched patients was longer than for unmatched patients. The median time-to-next treatment was 15 percent longer in the matched versus unmatched cohorts, 248 days or 215 days, respectively.
Further, when they folded in an additional cohort of 3,702 patients from IntrinsiQ Specialty Solutions who also underwent panomic tumor profiling, they also observed significant changes in the time-to-next treatment values between the matched and unmatched patients.
With time-to-next treatment as a surrogate marker for overall survival, Spetzler said Caris plans to acquire more retrospective data for hypothesis generation as well as embark on prospective studies to confirm their findings.
At the same time, he said the company plans to take advantage of the MolDx program developed by Elaine Jeter at Palmetto and seek coverage with ongoing data determination status as it validates the utility of its offerings.