Posted 26 February 2016
By Michael Mezher
The US Food and Drug Administration (FDA) is seeking input on how to regulate next-generation sequencing (NGS) tests that can scan a tumor's genes to guide cancer treatments.
At a public workshop at FDA's campus in Silver Spring, MD on Thursday, experts in oncology and NGS discussed some of the challenges and considerations for establishing standards for validating what's known as NGS-based oncology panels or assays.
In recent years, increased understanding of genetics has shown that genetic variations can impact both the progression and likelihood of a disease, as well as how well a patient may respond to a particular treatment.
With this knowledge, drugmakers have worked to develop highly targeted therapies that rely on companion diagnostics to identify a gene, or a set of genes that could impact a treatment's efficacy for a particular patient. So far, FDA has approved more than 20 such tests as companion diagnostics to cancer treatments.
A dozen miles south of FDA on Thursday, President Barack Obama touched on the importance of genetic variations in cancer at a White House summit on precision medicine.
"Areas like cancer show that each cancer may be unique – even if it's the same organ. All these insights promise the possibility of being able to cure diseases that up until now we couldn't figure out," Obama said.
Unlike traditional diagnostics, which typically detect the presence of one, or several, biomarkers or genes, NGS-based tests can screen for "numerous biomarkers [and] relevant mutations simultaneously."
Because of this, NGS-based tests hold great promise for oncology, but "introduce a challenge to the current companion diagnostics paradigm," said Reena Philips, director of the Center for Devices and Radiological Health (CDRH) Division of Molecular Genetics and Pathology.
Since 2014, FDA has been working to develop a regulatory framework for NGS diagnostics based on validated analytical standards "that would ensure that NGS tests produce accurate and reliable results."
The goal of FDA's workshop, according to a discussion paper distributed by the agency, is to get "input on strategies for establishing performance characteristics for NGS-based oncology panels," as well as specific insights on how to validate "rare variants across tumor types, claims for follow-on companion diagnostic, and post-approval assay modifications."
Pre-Analytical Challenges and Quality Control Metrics
One of the areas FDA hoped to get input on is how different pre-analytical factors, such as tumor type/location and how a sample is prepared and can impact the performance of a diagnostic.
For instance, should a NGS-based diagnostic require validation for different tissue types or specimen acquisition methods? Or are there "suitable pre-analytical tests … that may be used instead" to validate a NGS-based test?
John Pfeifer, professor of pathology and immunology at Washington University School of Medicine in St. Louis, emphasized that the choice of method used to prepare tissue samples is much less relevant to a diagnostic result than factors in the tissue itself.
Because tumors display a high degree of heterogeneity, labs may get a different result depending on the tumor sample.
"The importance of that is two labs can look at the same tissue and differ in terms of what mutations they find in important driver genes, not because of differences in quality of [their] assay, simply because they're looking at different parts of the tumor," Pfeifer said. "There is more variability in your assay in terms of coverage based on [guanine-cytosine] GC-content between different places in individual genes than there is between formalin fixation or fresh [frozen] tissue," he added, referring to two common methods of preparing tissue samples.
Responding to a question on whether NGS-based diagnostics should be validated for each tissue type (lung, liver, pancreas, etc.), Dara Aisner, a molecular genetic pathologist at the University of Colorado, asked, "To what degree is it necessary to validate tissue origin separately? Is lung really that different from skin, that different from liver? … Pretty much every sample that we look at, with some exceptions, consists of core constituents … and, from my perspective, lung and skin and liver are largely equivalent."
Aisner suggested that the emphasis should be placed on chemicals tumors produce, such as mucins or melanin, which can interfere with diagnostic results. She also cautioned that too many standards for pre-analytical methods for samples could limit access to NGS-based diagnostics and interfere with the diagnostic practices.
"The mitigating factor is that there are mechanisms to evaluate nucleic acid integrity … to me the solution here is to establish metrics that look at the resulting product, i.e. the nucleic acid integrity or the NGS data," Aisner Said.
Rajyalakshmi Luthra, director of the molecular diagnostics laboratory at the MD Anderson Cancer Center at the University of Texas, agreed, adding that setting too strict of limits could cause labs to reject samples for testing when they are below a particular threshold. "When we strictly adhered to the [DNA] concentration cutoff … in about 10% of cases, we wouldn't even test them, but when we reduced the [threshold] of DNA concentration, our success rate went from 85% to 95%."
However, Pfeifer stressed that some standards or thresholds are needed. "From a guy who's involved in a lab that does this, we need to strongly consider where there are minimal thresholds … somehow we have to figure out a way that we can get some clarity, some reproducibility," he said.
"Laboratories need to be able to demonstrate that they can extract nucleic acids of high quality … The question is, how is it sustainable to look for all these variances? Where do you draw the line?" Pfeifer asked. More importantly, he added, is that overprescribing pre-analytical methods "doesn't do any good if … your bioinformatics pipeline isn't appropriately tuned."
While a pan-cancer NGS-based diagnostic remains elusive, there are companies pursuing the technology. In September, FDA sent a letter to Pathway Genomics challenging its claim that the test can detect up to 10 types of cancer, and in January, Illumina announced it was launching a new company to develop a blood test that could detect multiple types of cancer "at the earliest stages."
At FDA's workshop on Thursday, the agency was interested in practical issues that could help it create standards to support so-called "pan-cancer" diagnostic claims.
"What tumor types have been most difficult for you to get NGS data from, and are there tumor types that should be excluded from pan-cancer claims unless the manufacturer specifically produces data from that tumor type?" Aaron Schetter, a scientific reviewer in CDRH's Office of In Vitro Diagnostics asked panelists.
Luthra responded that her group has had reproducible results with NGS-based tests when they had resection specimens from large tumors, but had less success with smaller tumors.
"It's not tumor type, it is the specimen, the DNA … and how it is fixed, that is more critical than tumor type," she said.
Labeling and Approval
Another issue, Pfeifer raised, is that it is important for NGS-based test designers to specify what types of mutations their tests can actually find. "It is absolutely unclear to the ordering physician the differences in the range of mutation types that that test is designed to detect," he said.
Schetter replied that FDA expects "to have labeling such that those sort of limitations [are] present."
In FDA's discussion paper, the agency proposes two tables that could appear on the label for an NGS-based cancer test. One that specifies the companion diagnostic indications the test is approved for, and another that lists other genetic variations the test can detect that have not been validated for guiding targeted therapy "in patients who have already been considered for all appropriate therapies."
Donna Roscoe, who heads the molecular genetics branch of CDRH's Office of In Vitro Diagnostics and Radiological Health, emphasized the importance of gathering data to support a diagnostic's intended use, saying, "We have manufacturers coming to us wanting to market their onco[logy] panels, [saying] 'what do you need from us?' and we say 'what's your intended use?'"
She added that one thing FDA will look for is a defined protocol that covers each step from the tissue sample to the results to demonstrate a diagnostic's expected performance.
During the final discussion of the day, FDA questioned panelists on what evidence NGS-based tests should rely on for clinical indications.
Abraham Tzou, a medical officer at CDRH, asked: "What sort of performance measures … would provide evidence of safety and effectiveness for a follow-on companion diagnostics claim and what types of study design considerations would be important?"
"I'm the target user of these tests that sees patients," said Dane Dickson, a medical oncologist and CEO of the Molecular Evidence Development Consortium. Dickson expressed skepticism about approving NGS-based tests without demonstrating that the tests improve patient outcomes.
"Do we need trials? Absolutely … the endpoints of the trials should be clinically based," he said.
Apostolia-Maria Tsimberidou, an associate professor at the MD Anderson Cancer Center, agreed, saying "I think it is important to … make sure there is clinical relevance. We need to standardize next-gen sequencing and establish certain rules … before we use them to make clinical decisions."