Complex software algorithms are the real reason FDA is after Laboratory Developed Tests
Laboratory developed tests (LDTs), are a category of In Vitro Diagnostics (IVDs) that currently enjoy FDA enforcement discretion, meaning FDA does not enforce regulatory requirements such as registration, listing, pre-market authorization or post-market surveillance and controls. Both IVDs and their LDT subset function similarly at their core: they use body fluids or cells/tissues (specimens) to detect or quantify a desired level of biomarker, often with the goal of detecting a disease or condition or to monitor health status to cure, prevent or treat disease states. IVDs and their LDT subset are themselves, according to the FDA, a subcategory of medical devices, “an instrument, apparatus, implement, machine, contrivance, implant or other similar product.”
FDA also regulates certain software as a medical device (SaMD) and many types of softwares that are incorporated into medical devices. Clinical Decision Support (CDS) is a broad term, used many different ways, but in general is a software algorithm that provides knowledge and person-specific information, and is intelligently filtered or presented to enhance health care. FDA regulates certain CDS functions as a subset of SaMD, but also has identified four criteria to identify when CDS does not even meet the definition of a medical device, and thus is not regulated by FDA. In general, software used in IVD applications cannot escape the “medical device” definition because FDA has defined the non-device CDS to exclude any software functions that acquire, process or analyze a signal from an IVD. Software functions that assess or interpret clinical implications or clinical relevance are medical devices, including software that processes or analyzes the genetic sequences from a Next Generation Sequencing (NGS) analyzer or an electrochemical or photometric response generated by an assay to obtain clinical test results.
LDTs have been described by FDA as “a type of in vitro diagnostic that is designed, manufactured and used within a single laboratory,” and these tests have been a hot topic for both congress and FDA in recent history. FDA takes the position that it has had the authority to regulate LDTs as medical devices since 1976, when congress clarified that IVDs are medical devices, but that it has purposefully exercised enforcement discretion over LDTs and took action only in limited circumstances, such as during declared public health emergencies.
The technology behind LDTs has advanced significantly since 1976, with high-complexity and high-risk tests critical to oncology, rare disease and infectious disease management more often than not being an LDT that has not undergone FDA review and authorization. Testing categories that rely heavily on genetic analysis or very large data outputs from the laboratory assays require rapid adaptation and agility to stay current with scientific advancements, such as application of machine learning (ML) and artificial intelligence (AI) tools. When FDA adopted a policy of enforcement discretion for LDTs, these tests were very localized, typically serving the patients living near the laboratory that developed the test, often academic medical institutions. LDTs are increasingly critical to the development of personalized medicine and may be used by treating physicians to make critical treatment decisions such as surgical intervention or selection of a drug therapy. It is widely acknowledged that LDTs are no longer always simple, localized tests, particularly as transportation solutions have evolved to allow same day or overnight shipping of specimens to a laboratory. Now a laboratory with an LDT can receive and test specimens from all over the country.
Complex Software Functions make Existing LDT Enforcement Discretion Policies Impractical
One of the most notable features of the rapidly evolved diagnostic testing industry is the need for and use of extremely complex software algorithms to interpret, digest, analyze and provide clinical interpretations. For example, Next Generation Sequencing, a technique used for DNA and RNA sequencing and variant/mutation detection, is essentially impossible to complete without complex software that digests large data sets and creates patient-specific results that require extensive bioinformatic processing (data quality monitoring, development of analysis protocols and data management). Bioinformatics software tools perform text mining (for comparative research and clinical protocol analysis), gene prediction, gene annotation, expression analysis, gene regulation analysis, identify gene variation and conduct pharmacogenetics analysis. Bioinformatics software can be obtained through open-source resources, developed in house, or purchased from hundreds of commercial vendors. It is rare that a precision medicine or NGS based LDT would be developed without the inclusion of a bioinformatic software tool as part of the test.
Notably, FDA has clearly articulated a concern with the software aspect of LDTs as far back as the 2014 Draft Guidance titled Framework for Regulatory Oversight of Laboratory Developed Tests (LDTs). FDA used software, along with high-tech instrumentation, as a defining characteristic that evidences a changed landscape and the need for reconsideration of FDA’s previous enforcement discretion policies. FDA specifically noted that the “attributes of modern LDTs, which were not attributes of the types of LDTs offered in 1976, create potential increased risk for patients in the absence of appropriate oversight.” A key attribute of concern is high complexity software, such as automated interpretation, use of non-transparent algorithms and complex software to generate clinical results. In fact, back in 2014, FDA considered the use of software for interpretation (as opposed to interpretation by qualified laboratory professionals) as a key indicator that the test was not actually an LDT. FDA was concerned then, and has only grown more concerned, that appropriate software validations are not performed, nor are they evaluated by other laboratory regulatory frameworks (such as the CLIA regulations).
Another concern with complex software is the fact that, to qualify for FDA enforcement discretion, an LDT must be “designed, manufactured and used within a single laboratory”.
Yet, many bioinformatics tools used by laboratories in developing their LDTs are not entirely home-grown. And, even if the software is literally coded and developed within the single laboratory, many of the databases and resources needed to perform the analysis and generate clinical results will reside outside the laboratory as general purpose or publicly accessible databases. A notable example FDA provides of a test that would not meet the definition of an LDT is “a laboratory contracts with a third party manufacturer to produce a key component…used in its device.” If one is truthful about the key, central role most bioinformatics software plays in laboratory tests, it becomes easy to see how FDA would consider anything other than a completely in-house developed software, within only in-house databases, as exceeding the current LDT enforcement discretion policy. To emphasize this point, legislative proposals have defined a “component or part” as including software intended by the developer to be included as an aspect of an in vitro clinical test, not accounting for where or who produced the software.
FDA has issued Warning Letters to laboratories marketing LDTs, particularly when FDA identifies an area of concern that warrants enforcement of FDA regulatory requirements, rather than enforcement discretion. In 2019, FDA issued a Warning Letter to Inova Genomics Laboratory for marketing genetic tests for predicting medication response and patient receptivity to drugs. As described in the Warning Letter, “FDA is concerned that the clinical validity of [Inova’s] MediMap Tests has not been established for their intended uses. Specifically, we are unaware of data establishing the relationships between the genotypes assessed by your tests and your assertions regarding drug response for multiple drugs.” These pharmacogenetic tests actually used a third-party software analysis tool to interpret the assay results and present the clinical result.
FDA Takes Matters into Its Own Hands
FDA believes continuing a policy of enforcement discretion results in an absence of appropriate oversight for LDTs, which may lead to increased risk for patients. Despite FDA’s insistence that it has always had authority to regulate LDTs as medical devices, many believed that congressional action was the best path forward to explicitly clarify FDA’s ability to regulate LDTs. However, various attempts at legislative proposals, including the most recent Verifying Accurate Leading-Edge IVCT Development Act (VALID Act), to clarify FDA’s authority, have been unsuccessful. The stakeholders potentially impacted by the proposed regulatory framework for in vitro clinical tests are wide-ranging and include test kit manufacturers, reference laboratories, or hospitals and academic medical centers with laboratories developing tests for patients — and many remained unconvinced that the risks posed by certain tests warrant FDA oversight, making it difficult to gain the consensus needed to pass the VALID Act.
FDA has recently taken two important and independent steps in its reconsideration and reshaping of the LDT enforcement discretion policy, despite the lack of congressional action. In late June of this year, the Office of Management and Budget (OMB)’s Unified Agenda included a Notice of Proposed Rulemaking (NPRM) for August 2023 stating, “This proposed rule would propose to amend the Food and Drug Administration’s regulations to make explicit that laboratory developed tests (LDTs) are devices under the Federal Food, Drug, and Cosmetic Act.”
Then, on June 20, 2023, FDA announced the launch of a pilot program for selected CDER-regulated oncology drug products with biomarkers. The details of the voluntary pilot program are outlined in the guidance document, Oncology Drug Products Used with Certain In Vitro Diagnostic Tests: Pilot Program, which states that the pilot is “intended to improve oncology patient care by providing transparency regarding minimum performance necessary for in vitro diagnostic tests used with oncology drug products enrolled into the pilot program.”
FDA’s announcement of the pilot explains that currently, in vitro companion diagnostic tests provide information for the safe and effective use of a corresponding treatment, e.g., a test may be used to identify patients who may or may not benefit from certain cancer treatments. In rare circumstances, FDA may approve a life-saving treatment requiring the use of an in vitro companion diagnostic even if a corresponding in vitro companion diagnostic has not yet received marketing authorization, and in these rare cases, LDTs are being used for patient treatment decisions. FDA has become increasingly concerned that these LDTs are less accurate and less reliable compared to FDA-authorized tests, and this could negatively impact treatment decisions.
Under the pilot program, FDA will request performance information for the tests that were used to enroll patients into clinical trials that support drug approval (the clinical trial assays or CTAs) from the drug manufacturers. FDA- then will post “the minimum performance characteristics recommended for similar tests that may be used to select patients for treatment with the approved drug” to its website. Laboratories then can use this information to guide the development of their LDTs. The pilot program began on June 20, 2023, and the initial phase of the program is anticipated to last up to one year, during which FDA will consider up to nine drug product sponsors to be accepted into the pilot program. This pilot program provides a way for FDA to influence the performance standards used in LDT development and also for FDA to gather data on LDT performance more broadly.
What’s Next?
While FDA taking matters into its own hands by announcing that it intends to pursue notice and comment rulemaking to regulate LDTs is a significant step, this may be a long and complex path for FDA. FDA’s decision to pursue rulemaking likely will be subject to legal and public policy challenges. Potential legal challenges include jurisdictional issues, i.e., that LDTs are laboratory services and not medical devices, and that LDTs operating within a single lab do not enter interstate commerce; arguments that development and use of LDTs should be considered the practice of medicine; and arguments that LDTs already are regulated by CLIA and several states (e.g., NYSDOH) and therefore FDA regulation is unnecessary. Potential public policy challenges include arguments that many critical tests are available only as LDTs; FDA regulation will impede and slow important innovation and patient access; and that other solutions to address FDA’s concerns are available and superior to FDA regulation and oversight of LDTs.
However, all of these arguments against FDA regulation of LDTs may be futile if the real driver behind the FDA regulation framework is the long accepted and fairly well-defined regulation of software as a medical device. If the driving concern for FDA in seeking to exert stronger controls over LDTs really is the complex software, then tests incorporating complicated bioinformatics and software analysis tools may be the first to experience FDA’s enforcement under a policy that sets the longstanding enforcement discretion aside.