Bringing Together Technology and Health Care Expertise

Launched in 2015, Livongo is a participant-centered digital health business that seeks to address the confusion, complexity, and cost of interacting with the health care system for those living with chronic diseases. Livongo initially focused on diabetes monitoring and prevention, but is now expanding its programs to address cardiovascular disease. Livongo has developed a platform, Applied Health Signals,¹ which aggregates real-world patient health data from multiple sources, interprets these data using medical literature and clinical insights, applies data to solutions, and iterates until desired patient outcomes are achieved.

In addition to the data collected from digital health technologies, Livongo aggregates data on people's health behavior, physical activity, and medical and pharmacy claims, with the aim of helping people lead healthy lives and spend less time worrying about their conditions. Livongo leverages a diverse team of data scientists, behavioral health specialists, certified diabetes educators, and physicians to interpret and apply these data in a way that is clinically meaningful to the patient and their treating physician.

Challenges of Managing Chronic Conditions Remotely

Bollyky described three challenges associated with remote management of chronic conditions: attribution of data, creation of meaningful data, and coordination of care. The first challenge involves ensuring that aggregate measurements are coming from the person being observed. As an example, Bollyky illustrated how a digital weight scale used by multiple people can create noisy data and make it difficult to determine which measurements should be attributed to the participant. This can be rectified by various approaches, such as by introducing capabilities in the digital weight scale that allow the participant to indicate when he or she is using it, or by collecting self-reported weight at registration from the participant and leveraging data science techniques to filter out obvious outlier data.

The second challenge Bollyky emphasized is the creation and provision of meaningful health signals to the participant in the context of the disease being monitored. Using Livongo's glucose meter as an example, Bollyky discussed how this digital health technology can provide participants value in terms of instant feedback on their blood glucose levels, measurement trends, and tailored recommendations based on those trends. For people living with diabetes, a very meaningful measurement is their glycated hemoglobin A1c (HbA1c) level.² However, Livongo currently does not have the capability to take this measurement. As a result, it uses data science techniques to derive HbA1c values from collected blood glucose data and validates the reliability of estimated HbA1c values based on the medical literature (Ford et al., 2018). However, reliable estimates require that observed data have a high enough frequency to confer statistical power.

The third challenge Bollyky discussed is in regard to coordinating care for people with chronic conditions. While Livongo does not currently serve as a health care provider or write prescriptions for its members, it does have algorithms to identify when someone may not be on the right medication or the right dose of medication. Bollyky noted that Livongo's goal is not to replace providers, but rather to support them. By giving its members health information, Livongo empowers them to talk to their providers. In addition to engaging providers, Livongo has ongoing relationships with pharmacy benefit managers and health plan managers, which enable more streamlined care coordination for participants.

Leveraging real-world evidence can offer health benefits in part, Bollyky observed, due to the Hawthorne Effect—the alteration of behavior by the subjects of a study due to their awareness of being observed (McCambridge et al., 2014)—which can pose complications in the research setting. Though using digital tools to support clinical trials may provide convenience, allow for more frequent assessments, and offer cost savings, it will be important to take into account how monitoring may impact the outcomes being measured.

All of Us

Joshua Denny centered his presentation on cohort research and how it can facilitate clinical trials, using All of Us,³ U.K. Biobank,⁴ and Project Baseline⁵ as examples. The All of Us Research Program was launched by NIH nationally in May 2018 with the goal of enrolling at least 1 million diverse participants.⁶ Enrollment, he explained, occurs through two routes: (1) health care provider organizations, and (2) direct enrollment of volunteers. Currently, more than 200 sites enroll participants through health care providers or directly at numerous consumer health sites, such as Walgreens clinics, blood banks, Quest, EMSI,⁷ and QTC/Leidos.⁸ To ensure that similar information is collected for all individuals, All of Us uses a common interface for recruitment, a common process for consent (which includes sharing EHR data), and common modules for health surveys. A key part of All of Us, said Denny, is that participants will have access to their data and may be recontacted over time.

All of Us will aggregate data from a variety of sources, including EHRs, collected specimens, claims data, and data provided directly by participants. All of Us is piloting the Sync for Science mechanism,⁹ a national collaboration among EHR vendors, to receive medical records data directly from participants. To harmonize these disparate types of data, All of Us uses the observational medical outcomes partnership common data model (OHDSI, 2019).

Denny noted that All of Us will begin collecting genome-wide association study (GWAS) data on participants and eventually plans to add whole-genome sequencing data. The program also plans to pilot Fitbits, Apple Watches, and other wearables in addition to working to link participants' geographical locations with data.

U.K. Biobank

U.K. Biobank, an open-access, prospective study, recruited more than 500,000 men and women ages 40 to 69 from 2006 to 2010. Participants consent for a wide range of research and for long-term follow-up. U.K. Biobank takes extensive baseline assessments of its participants and links to them longitudinally through their health record.

U.K. Biobank currently has GWAS data on all participants and the results from a standard panel of biochemical assays, such as lipids and metabolites. Whole-exome sequencing for every participant is under way, as well as whole-genome sequencing and metabolomics assays. There are plans for proteomic analyses, too. For a subset of participants, U.K. Biobank is collecting whole-body imaging data.

Project Baseline

Project Baseline is a large-scale project being conducted by Verily, Duke University, Stanford University, and Google that will collect real-world and clinical data, as well as biospecimens and survey results from some 10,000 individuals. Project Baseline will also heavily leverage digital tools by providing participants with home sensors and wearables, Denny emphasized.

The Role of Cohorts in Supporting Clinical Research

The combination of data that emerges from cohorts, such as EHRs, survey data, and wearable data, can be leveraged to conduct a wide range of clinical research activities, including

• Phenotyping and genotyping research to identify new disease targets or enable pharmacogenomics discovery;
• Phenome-wide investigations to test for indications and adverse drug events; and
• Artificial intelligence research and clustering approaches to identify disease subtypes.

To illustrate the potential of cohorts in informing drug discovery, Denny provided examples of studies that used EHR and genomic data to identify new classes of drugs to treat patients with rheumatoid arthritis, validate adverse effects for an enzyme-metabolized drug, and identify a novel class of cholesterol-lowering drugs (see Box 3-1).

BOX 3-1

Investigation of Genetics Data to Identify Drug Targets and Drug Response.

Large, diverse observational study cohorts, Denny continued, stand to accelerate clinical trials by generating basic discoveries, enabling direct and targeted recruitment of diverse populations and facilitating their assessment through existing cohort technology platforms, and enabling more intelligent trial design. The cohort landscape, Denny noted, is expanding and includes groups such as the China Kadoorie Biobank,¹⁰ the Electronic Medical Records and Genomics (eMERGE) Network,¹¹ Estonian Genome Center,¹² Kaiser Permanente's Division of Research,¹³ Million Veterans Program,¹⁴ PCORnet,¹⁵ and 23andMe.¹⁶ These emerging platforms and their capabilities for continuing contact will provide opportunities for investigators to discover novel therapeutic compounds.

Direct-to-Participant Trials

Authors

Affiliations

Cummings discussed what he termed D2P trials, which he defined as having no physical clinical sites, and thus no geographic limits on recruitment. Unlike the term “virtual” clinical trial, Cummings expressed that the term D2P trial more aptly captures the importance of building relationships with participants. Using examples of three D2P trials, Cummings provided lessons learned about the importance of incorporating the participant perspective and simplicity in trial design.

REMOTE

REMOTE was an Internet-based trial, launched by Pfizer Inc., designed to mimic a site-based trial for overactive bladder to test the efficacy of Detrol. REMOTE used similar identification verification and consent as in the KALM trial, was run from a single center, and used Web-based recruitment.

However, replicating a site-based trial was complex, largely attributed to protocol requirements and regulations that increased the burden of enrollment. For example, the protocol involved more than 90 interactions with participants to enroll and be randomized. Additionally, after signing eConsent, study staff were required to call the participant and read the full consent over the phone. Although REMOTE had received U.S. Food and Drug Administration (FDA) approval to ship the investigational product directly to participants' homes, prescribing laws limited recruitment to nine states, some of which required physical examinations by a physician prior to dispensation. Furthermore, during the run-in period, participants were asked to carry a plastic container to measure urine volumes and enter this information into a digital health technology. If the participant made an entry error, there was no opportunity to correct it and that participant could be excluded from the trial.

Recruitment was also challenging, noted Cummings. While nearly 21,000 women viewed the online introduction to the trial, only 1,159 were deemed eligible to participate. Of those eligible, only 1.6 percent of those women, or 19 women, made it through enrollment and randomization (see Figure 3-1). Although the perception is that REMOTE failed because of recruitment issues, Cummings noted that if the protocol were simplified so that 25 percent of the interested and eligible women could have easily enrolled, the trial would have achieved its goal of having 283 women participate.

FIGURE 3-1

Attrition of interested and eligible participants at each step of the REMOTE trial protocol. NOTE: ID = identification SOURCE: As presented by Steven Cummings, November 29, 2018.

NIH's Health Care Systems Research Collaboratory

Authors

Affiliations

The NIH Health Care Systems Research Collaboratory²⁰ (The Collaboratory), explained Weber, aims “to strengthen the national capacity to implement cost-effective, large-scale research studies that engage health care delivery organizations as research partners.” In leveraging data already passively collected as part of the EHR, it will be important, Weber noted, to determine when it is possible to partner with health care systems to answer the questions on improving health. The Collaboratory refers to its approach as embedded pragmatic trials because they are set in the health care system based on how patients see their health care providers and receive care in the health care systems. The Collaboratory has completed nine trials involving sites across the country, initiated six trials in the spring and summer of 2018, and will fund an additional set of trials in September 2019.

The Collaboratory Coordinating Center assists the trials via working groups²¹ and releases lessons learned into The Collaboratory's knowledge repository.²² Materials posted include how to design and conduct trials embedded in health systems and disseminate their results for diffusion into learning health systems, noted Weber. Using these lessons learned, The Collaboratory is able to help projects troubleshoot real issues as they arise, such as transitioning from International Statistical Classification of Diseases and Related Health Problems (ICD)-9 codes to ICD-10 codes in the middle of trials.

Decentralized Clinical Trials

Authors

Affiliations

While the public has expressed an interest in participating in clinical trials, recruitment and participation rates are low due to a lack of awareness of trials, the necessity to travel long distances to study sites, and the duration and number of clinical visits required. Emerging digital health technologies provide an opportunity to design decentralized clinical trials, which she called a disruptive approach to organizing the trial around the patient. In particular, Hawkins noted five opportunities that decentralized trials provide:

1.

Increased flexibility such that the burden of participation is reduced for both the patient and clinical trial sites;

2.

Increased participation of diverse patient populations by expanding access to those who may not reside near traditional academic centers, particularly for patients with rare diseases;

3.

Increased frequency of data collection and use of continuous data flows to more accurately and rapidly detect signals;

4.

Improved patient recruitment and retention; and

5.

Improved long-term follow-up to increase understanding of drug safety profiles and home-based dispensing.

Through work on decentralized clinical trials, Hawkins and her colleagues have identified a number of operational, ethical, regulatory, and management challenges.