Building Trust for Precision Medicine
by Barbara Greenspan, Legal Counsel for Health and Life Sciences and Alice Borrelli, Global Director for Healthcare Policy
Precision Medicine offers the medical community new tools for patient care based upon the many factors that influence a patient’s health, including genomics, environment, clinical history, and current health status. Big data analytics which can compare and contrast an individual’s complex health factors to those of millions of other patients are providing new insights to clinicians, from automating journal searches for recommended pharmaceutical pilots to discovering new and unlikely pairings for diagnosis and treatment. In order for this to happen, the big data algorithms and the clinical algorithms used in these systems must be in constant “learning” mode and must adapt and improve based on the learnings. There is significant concern that traditional medical device regulatory oversight by the FDA is unsuited for application to big data analytic systems that are driven by continuous improvement. If an alternative oversight approach is not established, the traditional regulatory structure will prevent these systems from reaching their full potential to improve the public health through more insightful clinical care.
Intel has been at the forefront of designing hardware and software that optimize today’s most advanced health technologies. For example, Intel is partnering with the Broad Institute to develop the open-source Genome Analysis Toolkit and collaborated with Dell and the Translational Genomics Research Institute (TGen) on a first-of-its kind clinical trial for children suffering from Neuroblastoma, a deadly fast growing cancer where timing is crucial for patients. A positive outcome of the TGen, Dell and Intel collaboration is the reduction of sequencing time from 12 days to 5 and data analysis from several days to six hours. Together with Oregon Health Sciences University Intel created the Collaborative Cancer Cloud (CCC) – which was recently extended with participation from Dana Farber Cancer Institute and Ontario Institute of Cancer Research. It is a precision medicine analytics platform that allows institutions to securely share patient genomic, imaging and clinical data for potentially lifesaving discoveries. It will enable large amounts of data from sites all around the world to be analyzed in a distributed way, while preserving the privacy and security of patient data at each site.
A cutting-edge health technology that stands to benefit from today’s advances in big data analytics are Clinical Decision Support (CDS) systems. CDS systems take information about a patient and, using methods and data sources that vary in complexity, produce a patient-specific, actionable result such as a diagnosis or the identification of relevant or recommended clinical trials or therapies. FDA’s policy on CDS products today remains unclear. The extent of FDA’s authority over these products is currently under congressional review, and several pieces of proposed legislation would attempt to update and streamline the regulation of such products intended for use in clinical settings.
As a technology leader that is working on-site with research teams in the world’s leading medical centers, Intel offers the following recommendations for the modernization of regulations affecting CDS software. Given the variety of CDS products and services that are possible and the variety of intended use cases for CDS systems, there is no single risk profile that is applicable generally to CDS. Intel’s recommendations emphasize and respect patient safety above all else, while also removing unnecessary regulatory obstacles that stifle innovation.
1. The FDA should regulate CDS systems when their intended use presents a high risk to the patient.
Intel agrees with the conclusions of the Food and Drug Administration Safety and Innovation Act (FDASIA) report to Congress that the FDA should focus its attention and oversight on medical device health IT functionality, such as remote display or notification of real-time alarms from bedside monitors, and robotic surgical planning and control.
For the wide range of evolving CDS products and services Intel agrees that the best approach is one that:
• Assesses risk to patient safety based on the complexity of a CDS and its intended use.
• Leverages standards and best practices in the software industry.
• Employs industry-ratified tools, testing and quality certifications of software conformity to quality standards.
• Selectively encourages non-punitive reporting of adverse events.
2. Regulation of CDS should recognize that the risks to patient safety attributable to different functional components vary, as do the controls necessary to mitigate these risks.
To optimize the regulatory burden, regulation of each functional component of a CDS system (technology platform, machine algorithms, data quality, clinical algorithms, etc.) should be calibrated to the risks to patient safety and efficacy attributable to that function.
For example, FDA should separately examine analytic validity and clinical validity since the risks attributable to a defect in the analytical validity of a software product can be sufficiently mitigated through compliance with appropriate industry consensus standards for software quality that ensure mathematically accurate calculations at a high degree of confidence. Likewise, because the clinical validity of an algorithm can typically be demonstrated through references to appropriate peer-reviewed literature, clinical practice guidelines, data registries, etc., FDA should not ordinarily require manufacturers to conduct new clinical trials to demonstrate clinical safety and efficacy.
3. Regulation of CDS should leverage public-private partnerships and the use of consensus standards to avoid unnecessary delays.
Data science and software development are fast-moving and iterative, dramatically more so than traditional medical device development. Regulation of CDS systems should recognize this fact and leverage public-private partnerships and consensus standards to make review of new CDS products and services and modifications to existing systems as efficient as possible. For example:
• Review should, either in part or in whole, be conducted by accredited third parties.
• Review should be subject to reasonable and enforceable time limits.
• FDA should designate an appropriate ANSI-approved international standards body, NIST or other appropriate third party qualified to assess software quality to work with industry to establish this standards-based approach to certification of analytic validity.
• Manufacturers should be able to demonstrate clinical validity by referring to peer-reviewed scientific literature, clinical practice guidelines, data registries, certified genomic databases, and other sources of information accepted in the medical profession.
• Industry should have available conformity assessment tools, such as product testing, certification and accreditation to give assurance that certain products, services, systems, or organizations meet specified standards or fulfill certain requirements. The goal is to give confidence to clinicians adopting these new tools and to distinguish high quality products.
4. Alternative approaches to adverse event reporting should be considered that reflect the risk presented by a CDS system’s components, functionalities and intended use.
Adverse events should be reported in a non-punitive, environment. Intel agrees with the FDASIA report that a Health IT Safety Center or a similar public private entity should be established to receive life threatening adverse event reporting. The Patient Safety and Quality Improvement Act of 2005 should be a model to consider.