Regulatory agencies have taken a keen interest in the development of digital devices to ensure that data acquired in clinical trials are collected systematically and rigorously, and with a focus on providing real benefits to patients while not compromising privacy or introducing other ethical problems, said Husseini Manji. Thus, he said, regulators need to be at the table early in the evolution of the technologies to avoid downstream roadblocks, and clinicians need to be engaged in the process to ensure that these devices will be successfully integrated into clinical practice.
Vaibhav Narayan, vice president of Research Information Technology at Janssen Pharmaceuticals, noted that although mobile computing devices and biosensors are evolving more rapidly than drugs and biologics evolve, the regulatory pathways for digital biomarkers and other diagnostic and monitoring solutions have the same need for accuracy, validity, utility, and value. What has changed, he said, is the speed at which the technology itself changes. Thus, when these technologies are used to provide digital readouts in clinical trials, in the time it takes to design a study and agree on a protocol, available sensors may have changed from what was originally envisioned in terms of the application programming interface, data standards, and algorithms for managing data, he explained. Moreover, the ability to take measurements remotely in large populations may result in data- and evidence-generation studies that look very different from classic clinical trials that have been done in the past. As a result, iterative designs and more agile methodologies are needed, as well as new regulatory pathways that enable rigorous evaluation of digital solutions and that are responsive to the different time scales at which these solutions evolve, he said.
Although digital technologies may help increase the efficiency of clinical trials in terms of recruiting and consenting participants as well as in maximizing adherence and retention, they also introduce several potential regulatory challenges, said Narayan. For example, given the variety of devices available, he advocated a device-agnostic paradigm where many different devices can be used to generate similar types of data. However, this approach introduces other challenges in terms of maintaining data quality, standardization, and interpretation, he said.
REGULATORY PATHWAYS FOR MEDICAL DEVICES
From a regulatory perspective, digital health technologies may be defined as medical devices if they are intended for use in the diagnosis, cure,
mitigation, treatment, or prevention of disease, or if they impact the structure or function of the human body through a non-chemical action,1 according to Carlos Peña, director of the Division of Neurological and Physical Medicine Devices at the Food and Drug Administration’s (FDA’s) Center for Devices and Radiological Health. This includes devices that take measurements to provide users with information about their conditions, he said. Medical devices can be classified into one of three categories with different levels of regulatory control based, in part, on the individual risk/benefit profile of the device, said Peña. Clinical data are typically required only for the highest risk (Class III) devices such as deep brain stimulators, he said. Class II devices may follow the 510(k) submission pathway where clinical data is typically not needed, although there are cases where clinical data was submitted. For devices that are not comparable to anything on the marketplace and that present low to moderate risk, a de novo pathway is also available.
Peña strongly urged device developers to take advantage of FDA’s free presubmission process, which allows them to obtain input from FDA on data that will be needed for regulatory approval. Presubmission allows regulators and developers together to map out important issues before a study is conducted so that everything is transparent moving forward. He said these front-end discussions are particularly important for digital health technologies because there may be an incomplete understanding about what the technology actually does. The presubmission process takes about 75 days to complete, said Peña. That process, as well as other efforts to increase interactions with sponsors and provide more transparent guidance on the regulators’ expectations for different types of products, has helped support and in part expedited the review of investigational device exemption studies from an average of more than 400 days in 2011 to about 30 days in 2017, he said.
Even if the product is not subject to regulatory approval, if it speaks to people about their condition Peña suggested it may be helpful for developers to consult with the agency to ensure that the right information about the device is conveyed to users. Narayan added that literally thousands of apps are available that make claims related to behavioral health with no supporting data, including some of which give egregiously bad advice. The abundance of these apps makes it challenging for developers to create and promote rigorously designed products in the behavioral health field. Peña responded that a digital health software precertification
___________________
1 The definition of a medical device is specified in Section 201(h) of the Federal Food, Drug, and Cosmetic Act (21 U.S.C. 321).
pilot program, designed to expedite the approval of these technologies and reduce the burden for both sponsors and FDA, may represent part of the solution to the challenge of the pace of digital health, because it may encourage sponsors to come to FDA to help get their product to market. This program is in its formative stages, he said, but should soon be moving to more defined criteria.
Peña noted that the 21st Century Cures Act, signed into law in 2016 to accelerate medical product development and encourage innovation, established a “Breakthrough Devices” program and codified several FDA policies related to digital health. For example, it amended the definition of a “device” to exclude certain software functions, including those intended to provide recommendations to health care professionals for clinical decisions in which the user can independently review the basis of the recommendation. FDA has increased its focus on patient use and preferences in other ways as well, said Peña, in part by partnering with patients and advocacy organizations. For example, they recently held a meeting on the use of real-world data and have published guidance on using real-world evidence in premarket submissions (FDA, 2017b) to ensure that these data are of sufficient quality.
Peña added that federal agencies, including the National Institutes of Health (NIH) and FDA, have also been coordinating efforts to advance device development by including in their funding announcements a requirement for presubmissions.
REGULATORY USE OF DIGITAL DATA IN CLINICAL TRIALS
Digital technologies offer potential benefits for patient-centric drug development through electronic data transmission from patients at home or remote locations and from the ability to capture clinically meaningful measurements continuously in real-life situations, said Jacqueline Corrigan-Curay, director of the Office of Medical Policy at FDA’s Center for Drug Evaluation and Research (CDER). Other clinical trial efficiencies facilitated by digital technologies include enhanced recruitment and retention through the use of electronic informed consent, audiovisual presentations, and virtual visits; improved tracking of adherence to protocols; and increased use of patient-reported outcomes.
For example, CDER has developed a mobile app called FDA My Studies to gather real-time contextual data from research study participants
about medication use and other health issues. The app has been tested in a small pilot study as a tool for recruitment and consent in a cohort preselected from electronic health records. Other apps have been developed to track adherence to medication protocols, said Corrigan-Curay. The simplest of these apps employs an electronic diary, but there are also apps that pair with an ingestible marker or approved drug delivery device such as an autoinjector or inhaler. Designed primarily for clinical practice, Corrigan-Curay said these devices may also be incorporated into clinical trials.
Corrigan-Curay classified these technologies into two groups: interactive technologies that rely on active submission of data, and electronic monitoring technologies that gather data automatically from sensors. In either case, devices would need to be evaluated in terms of reliability, reproducibility, sensitivity, specificity, and whether they provide clinically meaningful data, she said. The devices must also be suitable for the intended population, she added. Studies suggest that age and socioeconomic status are not necessarily barriers to the use of mobile devices and online health portals (Irizarry et al., 2017; Ramirez et al., 2016; Rothenhaus, 2015), but certain physical limitations or sensory impairments may be relevant. Whether passive data are fed back to participants remains an open question because of the possibility that they could influence trial results by introducing bias, said Corrigan-Curay. She said FDA has published a draft guidance on the use of electronic records and electronic signatures in clinical research, which addresses issues related to data security, privacy, and traceability (FDA, 2017a). While data security is key to prevent malicious or inappropriate access to data, she noted that clinical investigators and regulators need to have access.
Corrigan-Curay co-chairs the Clinical Trials Transformation Initiative (CTTI), which has developed a roadmap for using mobile technologies in clinical trials. These recommendations were unveiled at an event with FDA on July 16, 2018.2 Another organization working to advance the use of digital tools in clinical trials is the Critical Path for Alzheimer’s Disease. Originally named the Coalition Against Major Diseases (CAMD), the organization has worked over the past 10 years to establish data standards and clinical trial simulation tools to support AD clinical trials, said Stephen Arnerić. They established the first integrated database of anonymized information from AD trials and worked with the Clinical Data Interchange
___________________
2 To learn more about the CTTI recommendations, see https://www.ctti-clinicaltrials.org/projects/mobile-technologies (accessed August 8, 2018).
Standards Consortium (CDISC) to develop data standards for clinical outcome measures, genetic, and biomarker data (Neville et al., 2017). CDISC standards are now required for clinical trial data submitted to FDA (FDA, 2014). CAMD also developed the first clinical trial simulation tool to be endorsed by regulatory agencies as a tool to model disease progression, placebo and treatment effects, and patient dropouts in clinical trials (Romero et al., 2014, 2015).
The next frontier, according to Arnerić, is for these tools to incorporate biometric and other digital assessments and to use these measures to better model what happens in the presymptomatic stages of disease by understanding real-time changes in function. For example, Arnerić suggested that it may be possible to detect subtle changes in cognition by tracking medication adherence. Computer use and walking speed also predict changes in cognitive function, said Arnerić, and could lead to a dramatic reduction in the sample sizes needed for prevention trials in cognitively intact individuals, resulting in substantial cost savings and a reduction in exposure to potentially unsafe drugs (Dodge et al., 2015). Real-world data can be collected with many different devices across multiple domains (mental and physical function, social engagement, etc.), ideally with data-agnostic platforms. But to transform these data into meaningful real-world evidence will require careful data standardization to bring information into a consistent format, aggregation with other types of contextual information or metadata, and quantitative modeling, said Arnerić (Piwek et al., 2016). Eventually it would also be valuable to link these data to other types of observational data and health records, although these data would introduce additional noise. Data sharing through an integrated research platform such as the Global Alzheimer’s Association Interactive Network (GAAIN)3 is key to making this big data vision a reality, said Arnerić.
A challenge that regulators will have to deal with is how to interpret findings when the objective data obtained through passive monitoring differs from how a person feels or what the physician thinks is happening, said Corrigan-Curay. She said this points to the importance of transparency and dialogue with user communities to make sure users trust the technology and do not believe it is replacing their own experience. This can be a particular problem with technologies that purport to objectively measure symptoms such as fatigue and pain, said Narayan. He said contextual in-
___________________
3 For more information about GAAIN, see http://gaain.org (accessed July 13, 2018).
formation incorporated into the algorithms can sometimes provide answers that align more closely with patient perception. Peña added that these issues are important topics that can be discussed in a presubmission.
NAVIGATING THE REGULATORY WEB
Structuring clinical trials and other clinical research to comply with an evolving web of laws and regulations presents a substantial challenge for research, especially for conducting collaborative research, according to Kristen Rosati. The challenge is particularly acute because these regulations are not harmonized across countries, states, or even across agencies within a country and are constantly in flux, she said. For example, drug and device developers in the United States must comply with the Health Insurance Portability and Accountability Act (HIPAA), the Common Rule (rules governing human subjects research across all federal agencies), FDA regulations, NIH policies, reimbursement policies from Medicare and various state Medicaid programs, and various state health information confidentiality laws. Developers of products for international markets must also comply with the European Union General Data Protection Regulation (GDPR), which became effective in May 2018.
Meanwhile, Rosati said potentially competing policies have emerged across the globe aimed at increasing the ease and utility of research and the ability of individuals to control their own data and biospecimens in both clinical care and research. This includes more individual control over deidentified data because of the potential for reidentification.
Another area of regulatory confusion arises because of the scope of various regulations, said Rosati. For example, in the United States, although HIPAA applies to the health care industry, which is the source of electronic clinical data, mobile medical device companies may fall outside the scope of HIPAA. In 2013, HIPAA was amended to give individuals the right to access their own data and also direct their health care providers to share their electronic data with a third party. This has allowed for the flow of data from mobile medical devices to third parties using the data for clinical care or clinical research, said Rosati. The GDPR, however, has a much broader scope and thus places more restrictions on confidentiality and the flow of information among different entities, said Rosati, although she added that many aspects of the GDPR remain poorly defined and understood.
The Common Rule has also been subject to change in recent years. A new rule was supposed to take effect in January 2018, but was delayed until July 2018 and then delayed again until January 2019, said Rosati. The status quo at this time is that researchers must comply with the pre2018 Common Rule, she said, but to future-proof their activities may want to follow the new rule, including one part of the rule that mandates use of a single Institutional Review Board for collaborative multisite research studies. This part was not supposed to take effect until January 2019. Another reason for following the new, but not yet implemented, Common Rule is that other elements related to the information included in informed consent documents and informational confidentiality may be more closely aligned with the GDPR and possibly other state health informational confidentiality laws, said Rosati.
The new Common Rule also has a new HIPAA exemption for secondary research using data or information derived from biospecimens (although not the biospecimens themselves) collected for clinical care or research repositories if the research is otherwise regulated by HIPAA, said Rosati. The new Common Rule also embraces the concept of broad consent, which although poorly defined has raised concerns in the research community as to how it will be implemented and whether it is more onerous than regular informed consent.
Rosati suggested that it might be beneficial for the United States to preempt all the different laws and regulations and replace them with a GDPR, as long it does not end up exactly like the GDPR, which she said poses real barriers to research. But as the laws exist now, anyone contemplating a collaborative research project will have to think carefully about whether their partners will be able to share data and implement appropriate governance activities in a manner that complies with the laws that oversee them. One measure that could help, she said, would be to build in consent to use data for future research, making the description of that research as broad as needed to cover anticipated studies, and also to build in consent to contact individuals about future research activities or future consent.
