NEW YORK (GenomeWeb) – Building on a joint US Food and Drug Administration and American Association for Cancer Research workshop that took place last year, a group of drugmakers, test developers, and academics gathered at the American Association for Cancer Research annual meeting earlier this month to discuss concerns and challenges in establishing analytical and clinical validity for liquid biopsy tests.
Although most liquid biopsy tests are currently offered as LDTs, and as such need not be approved or cleared by the FDA, the diagnostics field is still paying close attention to the agency's expectations and potential future regulation. In addition, it is facing concerns from clinicians regarding the analytical and clinical validity of the tests.
"The number of symposia and presentations at this meeting speaks to the variety of technologies being applied and the number of sponsors looking to grow into this field," said Memorial Sloan Kettering Cancer Center's Howard Scher, who leads an academic team working with Epic Sciences, and now Genomic Health, to advance a CTC-based test in prostate cancer.
"Unfortunately, there is a certain degree of performance that should be mandatory and what we are seeing now is that very few of these assays have really gone through the rigor that should be required for clinical use," added Scher, who is MSKCC's chief of genitourinary oncology
Since Roche is so far the only diagnostics firm with an FDA-approved liquid biopsy test, it has served as an example in both last year's workshop and the panel at this month's AACR meeting.
Roche received approval for its PCR-based Cobas EGFR Mutation Test v2 as a companion diagnostic for Genentech's Tarceva (erlotinib) last June, and in September, the FDA expanded the labeling for the test, adding an indication as a companion diagnostic for non-small cell lung cancer patients considering treatment with AstraZeneca's Tagrisso (osimertinib).
During the panel session at AACR this month, representatives from Roche, AstraZeneca, and the FDA highlighted some of the differences between Roche's first and second assay approvals and how they reflect the broader challenges of appropriate analytical validation in this space.
In the case of Roche's test — and liquid biopsy assays in the lung cancer space more broadly — the existence of tissue-positive liquid biopsy-negative, or tissue-negative liquid biopsy-positive patient cases poses related but different challenges.
In developing its test, Roche retrospectively compared blood-based mutation status to tissue results in patients treated with Tarceva or Tagrisso, but the concordance it was able to achieve was far from perfect.
For the company's first approval alongside Tarceva, the number of patients who were tissue-negative but liquid biopsy-positive was relatively low. However, a substantial portion of patients were blood-negative but tissue-positive.
The fact that these individuals, who are expected to respond to the drug based on their tissue status, would be missed by blood testing alone led to the reflex indication as specified by the FDA: If the results of the liquid biopsy test are negative, physicians are recommended to seek out tissue testing, and in blood-negative/tissue-positive cases, tissue should trump liquid biopsy in terms of whether a patient is eligible for treatment.
In the case of Tagrisso, the concordance data was a bit different. There were again patients positive for T790M resistance mutations in tissue but not in plasma. But in addition, there was also a significant number of tissue-negative blood-positive individuals.
FDA Medical Officer Abraham Tzou explained during the session that patients are known to respond or not respond to Targrisso based on their tissue mutation status, and that's how the pivotal approval trials for this and similar targeted therapies were designed.
There isn't prospective data available speaking to how patients with a mutation found in their blood but not tissue sample respond to treatment.
As a result, the FDA approval of the Roche test in this second indication specified that the test is only for patients in whom tumor tissue cannot be obtained. It's also conditional on Roche and AstraZeneca collecting and reporting post-market data on the outcomes of patients treated based on blood mutation status.
Some liquid biopsy test developers have pressed that blood-positive tissue-negative patients may respond just as well to therapy as tissue-positive patients, and it should be the actual patient outcomes that determine validity, not necessarily concordance with tissue.
The FDA does not appear to disagree, but so far there isn't comprehensive data to support this conclusion, and there are some results that hint at the opposite, at least in the context of EGFR drugs.
"The issue is, there can be both biological heterogeneity and clinical heterogeneity," Tzou explained. In other words, plasma-positive tissue-negative cases might not be false-positives, but they also might not be responders, or responders as robust as tissue-positive patients.
Researchers have begun to share data from prospective studies treating patients with targeted therapies based on blood-based mutation status alone. For example, Inivata shared results from a small study with collaborators from the Institut Gustave Roussy in France, showing that patients tested using the company's liquid biopsy assay benefitted from genomically targeted treatment just as much as subjects who received tissue-based tests in prior clinical trials.
But the development of this type of outcomes evidence is still in its infancy.
In terms of complexity, Roche's PCR-based test also pales in comparison to many NGS-based liquid biopsy assays that are now being used clinically as LDTs.
Tzou highlighted a number of other considerations that can confound efforts to demonstrate analytical and clinical validity that could pose challenges for broader, more complicated tests.
For example, he said, for tests that address many analytes, it is important to recognize that validation on a single target or class of targets doesn't necessarily apply to other targets.
For example, data that speaks to the sensitivity of detection of SNVs in plasma doesn't always translate to other types of variants, like copy number changes or structural alterations.
"If I have a class of students and I have one student that does very well and gets all As, if I just test the A student, am I confident the rest of the class is learning? No." Tzou said.
"If it's not complicated enough that different variants behave differently, the same variant can also behave differently," he added.
Consider one molecule of interest in 100 molecules of DNA versus 10 targets in 1000 molecules. The relative frequency is the same, but the target in the first case will be harder to detect because the absolute amount is lower.
"The takeaway here is, when the FDA for these molecular diagnostics has used the term 'limit of detection,' it is in the context of what a consistent level of detection is — for example at 95 percent or higher — not just the capacity or potential LOD in some cases," Tzou said.
Aside from the question of FDA regulation, many of these same issues make it difficult for clinicians to assess the validity of commercial LDTs, or comparing them to one another.
Tests are described by different companies using different terminologies. Sensitivity, for instance, can be described either per sample or per base pair, and the two are not necessarily directly comparable.
Companies also varyingly describe LOD in terms of a threshold that can be consistently measured, which is what Zhou said the FDA usually means when it talks about limits of detection, or as a more absolute technological capacity that may not always be reached in every case.
Girish Putcha, who is director of laboratory science at Medicare contractor Palmetto GBA but was speaking as an independent observer, highlighted at the previous AACR/FDA meeting that the way these types of metrics are reported is not always consistent from test to test, and is not always or often verified by outside observers.
Putcha has said previously that both strategies for looking at accuracy have downsides. It is best, he suggested, to report both, as Palmetto's MolDx recommends in the analytical performance specifications it has developed for ctDNA testing.