By Clinical Informatics News Staff 

August 22, 2014 | This week, at the Next Generation Dx Summit in Washington, DC, two FDA representatives addressed the agency’s thought process when reviewing assays based on next-generation sequencing (NGS), especially in the form of companion diagnostics paired with specific drugs. Jennifer Dickey, an FDA reviewer in the Office of In Vitro Diagnostics, and Živana Težak, the OIVD’s associate director for science and technology, shared their insights into winning strategies for submitting NGS assays for premarket approval. Their comments offer a valuable perspective on the future prospects for NGS testing in clinical settings, as the industry wrestles with best practices for gaining regulatory approval for this cutting-edge technology. (As is customary for FDA representatives, Dickey and Težak stressed that the views they expressed are their own and not necessarily official FDA positions.)

The FDA has not yet cleared any NGS-based companion diagnostics, but there are encouraging signs that the climate is becoming more favorable to these tests. Companion diagnostics as a class are now a regular part of trial design, particularly in oncology, and the FDA has approved 12 CDx tests to capture actionable cancer mutations. Meanwhile, last November marked the first approval of an NGS sequencing platform, the MiSeqDx, with related assays for cystic fibrosis mutations. As Težak pointed out, this approval came only a little over two years after the FDA held its public meeting on the subject of “Ultra High Throughput Sequencing for Clinical Diagnostic Applications — Approaches to Assess Analytical Validity,” to take comments on appropriate standards for evaluating NGS diagnostics.

Advice for Submissions 

The major message both FDA officials stressed was that the agency is eager to work with companies to expand access to NGS testing in the clinic. While there are significant challenges to validating these tests to the FDA’s standards — in particular, they can cover such a wide range of genetic variants that it is effectively impossible to verify that each one is called and interpreted correctly — a combination of careful labeling, and creative references, can overcome these obstacles. “The intended use, and the indications for use, are the key to how we review,” said Dickey, so being specific about what a test can and cannot capture, and how this should impact clinical practice, can go a long way toward satisfying the FDA.

The two approved MiSeqDx cystic fibrosis assays provide a useful case study. For both assays, the FDA was willing to use the public CFTR2 database, rather than published literature or clinical studies, as the core evidence that variants flagged by these tests can be considered diagnostic of cystic fibrosis. Težak clarified that the agency will continue to consult similar databases, provided they meet three major conditions: that they use standard nomenclature to make it easy to find and contribute information; that they state the clinical impact of each variant with supporting evidence for the claim; and that the history of annotations in the database is tracked so that sources of information can be identified. She also encouraged the community to share data with public databases to make future NGS clearances easier, specifically citing ClinVar as a promising project. (Though “not there yet” by the FDA’s standards.)

The narrower of the two MiSeq cystic fibrosis tests captures 139 specifically enumerated variants from the CFTR2 database. In this kind of case, said Dickey, because it is not practical to validate the assay’s sensitivity to all these variants in clinical samples, “the accuracy of extremely rare variants can be established with contrived samples, such as plasmid blends or cell lines.” It is important, however, that trials establish analytical validity in each type of variant claimed, which could include single-nucleotide variants, insertions and deletions, and larger structural or copy number variants.

Meanwhile, the broader cystic fibrosis assay, which outputs the complete sequence of a patient’s CFTR gene, demonstrates how the FDA will look at comprehensive sequencing. Where a test claims accuracy across a whole genomic region, said Dickey, it will be necessary to use real clinical samples. However, “accuracy will be considered both for the variant calls, and for the wildtype calls across the region,” meaning every sample can contribute to the body of evidence backing an assay. A sequencing technique in which the FDA has a baseline level of confidence — usually Sanger sequencing — will be used as the basis for comparison.

Companies looking to market broad sequencing tests that span genomic regions need to give extra consideration to their intended use claims. These tests place an extra burden on lab technicians and clinicians interpreting the results, because they are likely to turn up benign variants and cannot state the clinical significance of every result. Dickey suggested that best use cases for such tests might be in atypical disease presentations, or where other tests have failed to make a diagnosis, and that labeling should reflect that.

Companion Diagnostics 

There are extra wrinkles if an NGS test is being considered as a companion diagnostic alongside a drug with a molecular target linked to a specific gene. Not only does this very direct link to treatment raise extra specters of risk to the patient from misdiagnosis, it also demands a high level of coordination with clinical trials for the related drug. Every FDA-cleared companion diagnostic so far has been approved at the same time as the drug it is paired with, and while Dickey said that the agency could consider exceptions to this policy, it is the preferred route for approval.

Dickey cautioned that developers of co-diagnostics must have regular contact with their pharmaceutical partners to anticipate changes in the clinical trial schedule, and that if the trial design changes — say, a test that was used to stratify patients becomes the basis for enrollment in a new trial — an experimental assay may well find itself in a higher risk class in the FDA’s eyes. Test producers should be constantly wary of the potential need to file for an investigational device exemption.

She further stressed that any changes to a co-diagnostic between the beginning and end of a Phase III trial will absolutely mandate new analytical data to ensure the altered test would not have skewed the trial’s results through differences in patient enrollment. It is vital that the same patient samples be banked and reanalyzed: “You need the clinical samples from the trial to bridge between the two assays.”

Earlier this month, the FDA published its final guidance on “In Vitro Companion Diagnostic Devices,” which can be consulted for more information.

While the FDA is increasingly open to creative regulatory measures to evaluate NGS testing, no test maker should ever trust that the agency will accept evidence such as public databases, contrived samples, or representative subsets of tested variants, without first consulting with regulators. “It’s absolutely essential to get buy-in from the FDA before you try some of these alternative approaches,” said Dickey.

In related news, Illumina announced yesterday that it plans to release a broad oncology gene panel for the MiSeqDx, as part of a partnership with leading pharmaceutical companies to support patient stratification in cancer drug trials.