The COVID-19 Pandemic and Decentralized Clinical Trials

The COVID-19 pandemic significantly disrupted clinical care and research across the US and the world [Reference Van Norman6]. Although telehealth technology and home health care were established in clinical practice prior to the pandemic, COVID-19 forced rapid-cycle adoption of crisis-related policies and procedures establishing pandemic standards as the new normal [Reference Fu, Gerber and Beg13]. Organizations had to be nimble in adopting alternatives to the traditional top-down approach to clinical trials, with emphasis shifting to avoiding having clinical trial participants come to healthcare facilities when their care did not require it and to accommodate research teams working remotely [Reference Whitehead and Mehrotra15]. Virtual interactions between clinicians and patients to provide continuity of care while maintaining social distancing, coupled with improving smartphone health applications and connectable medical devices, hastened the shift to decentralized trial models. The COVID-19 pandemic was a call to action to expand the use of decentralized and virtual interfaces while testing the feasibility of hybrid decentralized trial models incorporating traditional in-person and virtual modalities, fostering wider and more generalizable research [Reference Daudelin, Brewer, Cabrera, Dulko and Selker16].

Early in the COVID-19 pandemic, a collaboration between Tufts Medical Center and Tufts University brought together a multidisciplinary team to conduct an NIH National Center for Advancing Translational Sciences (NCATS)-supported decentralized clinical trial. The study recruited outpatients with mild to moderate COVID-19 to determine whether a widely available repurposed therapeutic, niclosamide, reduced SARS-CoV-2 oropharyngeal (OP) and fecal shedding and duration of COVID-19 symptoms [Reference Cairns, Dulko and Griffiths17]. Beyond testing the efficacy of the drug, overarching goals in the conduct of this decentralized trial included: (1) avoiding having patients visit a healthcare facility enhancing convienence of participation,  (2) reducing outpatient visits to limit infectious exposure of other patients and healthcare providers, (3) providing extensive decentralized patient screening prior to enrollment to ensure that all enrolled persons met inclusion and exclusion criteria (including avoidance of enrolling persons already severely ill with COVID infection), and (4) conducting in-home participant monitoring post enrollment via a Health Insurance Portability and Accountability Act of 1996 (HIPPA)-compliant telehealth platform to evaluate for progression to severe COVID.

Decentralized Recruitment

Participants were identified from records of outpatients who came to Tufts Medical Center seeking COVID-19 testing. The Tufts Medical Center Institutional Review Board (IRB) approved an omnibus information sheet, in multiple languages, for display at the Tufts Medical Center outpatient laboratory stating that patients requesting COVID-19 testing might be contacted by a research study team. Each Tufts Medical Center and affiliate site displayed a study-specific poster with QR code directing potentially COVID positive patients to study team contact information in the emergency department (ED), primary care, and infectious disease clinic of each Tufts site. In additon, any patient with mild to moderate COVID-19 symptoms presenting to the EDs or clinics who did not require supplemental oxygen or hospitalization received a brief overview of the study from the local clinical team with contact information for the study team at the time of testing. Patients were advised that following confirmation of a positive COVID-19 test they could proactively engage with the study team if interested in participation.

During trial activation, we simultaneously built a corresponding Research Electronic Data Capture (REDCap) database as this electronic-database capture was readily available within the Tufts CTSI infrastructure. A structured data query and a secure automated data delivery process were developed by the Tufts CTSI Informatics Team that created a data file of outpatients testing positive for SARs-CoV-2. These data were electronically provided to the study team twice daily, at 7am and 3pm, via REDCap. The data files were housed in a secure directory with access limited to members of the study team, password protected. The Tufts CTSI Central Screener quickly identified and broadly screened each potential participant positive for SARS-Cov-2, referring them directly to the study team through an automated e-mail notification in REDCap.

A Successfully Deployed Decentralized Clinical Trial

Upon enrollment and consent, each study participant received a Clinical-Trial-in-a-Box – including a thermometer, an oximeter, and oropharyngeal and fecal sampling materials with materials to return self-collected specimens, delivered directly to their home (Fig. 1). The boxes were barcoded, scanned, and delivered by a HIPPA and Occupational Safety and Health Administration (OSHA)-trained and certified medical courier. Strict chain of custody protocols from the moment of pick up to drop off were maintained by the courier service. Our Tufts R-SPECT platform was used to conduct all study visits including e-consenting, monitoring of symptoms, assessment of adverse events, and coordination of participant self-collection and return of oropharyngeal and fecal specimens. The blinded study drug (or placebo) was dispensed, sealed, and labeled by the Tufts Research Pharmacist with direct hand off to the study coordinator, maintaining Investigational Drug Accountability Record Forms. The study coordinator retrieved the investigational product from the research pharmacy for inclusion in the study box. A certified partner courier (HIPAA compliant and OSHA compliant) was notified of a ready pickup and received the study box via direct hand off from the study coordinator for immediate delivery to each participant. A proof of delivery photo was provided in real-time to the study team by the courier.

Figure 1. Decentralized clinical trial box prototype.

The trial opened to accrual on October 1, 2020, and the last participant was enrolled on April 20, 2021. In all, 73 participants were enrolled and randomized [Reference Cairns, Dulko and Griffiths17]. The study team completed 728 telehealth visits with participants. Over 650 SARS-CoV-2 test samples were self-collected by study participants: 316 respiratory samples and 348 fecal samples.

Of the 67 eligible trial participants invited to take part in a post-study mixed methods evaluation of the telehealth platform, 46% (n = 31) completed a post-trial survey to determine participant and research team perspectives of the decentralized trial [Reference Daudelin, Brewer, Cabrera, Dulko and Selker16]. Key findings included:

• Compliance with planned telehealth visits ranged from 84% to 100% over the course of the study.

• Fecal and respiratory (oropharyngeal) samples were self-collected with a mean of 94.6% of planned fecal specimens and 96% of respiratory specimens successfully collected and processed across the duration of the study.

• The proportion of participants who dropped out of the study early was sustained below the target threshold of 10% (Fig. 2).

• Ninety-seven percent (n = 30) of post-trial survey respondents had not previously participated in clinical research; 77% (n = 24) of whom said they would consider taking part in another clinical research study in the future for a condition other than COVID-19.

• Of those who reported they would consider participating in a future study, 96% (23/24) indicated they would be more likely to participate if the visits were not in-person.

• The greatest barrier identified (n = 3) involved technology issues with the telehealth visits, including poor internet connection or cell phone coverage. However, those who identified this challenge said when one platform was not working, the researcher and participant would switch to a second platform, causing minimal disruption.

Figure 2. Proportion of patients who droped out prior to end of study.

We recognize that the 46% of participants who completed the post-trial survey are a subset of all participants and there may be response bias. However, this information does provide some useful data regarding the feasibility of remote patient self-collection of specimens.

Among all patients in the Intent-to-Treat sample, there were 4 African American individuals (6.0%), 5 Asian individuals (7.5%), 7 Hispanic individuals (10.4%), 53 White individuals (79.1%), 1 individual with multiracial or multiethnic background (self-reported Asian and White; 1.5%), and 4 individuals with other race or ethnicity (6.0%). All participants were from Massachusetts. While state-to-state licensing accommodation was an option during the pandemic, we did not need to identify physicians with medical licenses in multiple states.

The team acknowledges the relative lack of ethnic diversity in our trial and that input from diverse participants is paramount to ensuring that trials are responsive to all patient needs while also meeting research requirements including gaining ethnically diverse perspectives of interfaces and data collection tools. Our future work includes collaboration with the Tufts CTSI Community and Stakeholder Engagement team to design studies that foster gaining this important perspective. In the next phase of development, we will engage the Tufts CTSI Stakeholder Expert Panel, a group of local community members, in providing feedback on the R-SPECT platform and its processes in order to ensure that it meets the needs of diverse patients across the geographic spectrum. With the median age of 31 years in our initial COVID-19 decentralized trial, our study demographic may indicate that younger people were more comfortable with decentralized technology, telehealth visits, or self-sampling. Further study is needed with older adults as well as with special populations [Reference Daudelin, Brewer, Cabrera, Dulko and Selker16].

An Academic-Engineering-Manufacturing Collaboration

While the Tufts R-SPECT platform was first deployed during the pandemic, we believe it has broader applicability for the conduct of clinical research. Collaborating with the Massachusetts Institute of Technology (MIT) Master of Engineering in Manufacturing Program, along with clinicians and researchers at Maine Medical Center, we are engineering and testing our decentralized clinical trial platform with enhanced workflow processes and a more user-friendly interface for use in a range of settings. Key considerations informing platform design include: (1) interfacing with wearables, smartphones, and other devices; (2) capturing electronic patient-reported outcomes; (3) CTMS integration; (4) supporting use of diverse learning products (print, video, animations, and images); and (5) managing and tracking study inventory.

Our v2.0 Clinical-Trial-in-a-Box contains home-delivered trial participation materials, including consumer wearables with data access agreements and adaptable learning products (print, video, animations, and images). Logistics and tracking managing inventory not shipped, items shipped, delivery, and replacement management are integrated within the CTMS. Through this collaboration, the platform became more capable of outreach for decentralized trials to populations of patients that often do not participate in clinical trials, e.g., rural communities where a clinical visit can be an hour or more away.

The overarching objectives for the Tufts CTSI-MIT ongoing collaboration are to: (1) develop a platform to run a fully decentralized trial from enrollment to participation, (2) integrate reporting systems and metrics at the study level, (3) customize home delivery of trial materials, and (4) collect patient-reported and Social Determinants of Health (SDoH) data. This last goal was a primary objective, as including participants with a range of SDoH is key to gaining a deeper understanding associated with barriers to trial participation, specifically in under-resourced populations. Based on our initial work, patients who may have faced challenges to involvement in centralized trials appear to be able and willing to consider enrollment in a decentralized trial. We also seek to foster enhanced trial access for other populations currently underrepresented in traditional trials, e.g., the elderly, those living in remote locations, and racially and ethnically minoritized groups.

Decentralized Clinical Trial Data Acquisition

The privacy of data acquisition is a key consideration, as is blinding, as investigators or sponsors should not see certain data. Digital health technology must ensure privacy and security to prevent unauthorized access to the technology and the data it collects [8]. Privacy-Preserving Record Linkage (PPRL) enables the de-identified linking of individual-level data across time and data sources to mitigate privacy concerns through encryption of personally identifiable information [Reference Mirel, Resnick, Aram and Cox18]. In traditional clinical trials in which data is collected onsite by investigators and research study staff, there often is no need for PPRL. In a decentralized-based platform, particularly where home and consumer-grade medical devices (i.e. smart watches, smart scales, smart thermometers, electrocardiographic devices such as the KardiaMobile 6L, etc.) are used to acquire trial measures, data may first be uploaded to the medical/consumer device cloud server under different user identifiers which often have the same data elements as are captured within a traditional clinical trial; e-mail address, first/last name, date of birth, and addresses. The externally staged data need to be linked via PPRL methods to the core patient medical histories and data from the study “home” research site or clinical information system. In a decentralized clinical trial platform, we do not make the assumption that all data involved are easily and readily accessible. In its design and implementation, we envision providing tools and services to support PPRL when needed. The data acquired from disparate devices, systems, and methods produce vendor proprietary data formats that can be normalized and aggregated into a central data collection center, e.g., at the Tufts Research Data Warehouse, for management, quality assessment, and analysis (Fig. 3).

Figure 3. Decentralized clinical trial platform data acquisition process/systems.

In our ongoing Long-COVID study, decentralized trial participants are assigned study-specific e-mail addresses that provide a level of identity protection and enable linkage to other study systems such as previous care data from the EHR or within a CTMS. With assigned e-mail and SMS Text messaging, study participant identity is protected when handling responses to patient-reported surveys, a component of our PPRL approach.

The assignment of study-focused e-mail addresses has been applied in a number of MIT-run studies involving Oura® rings and other wearable devices. Using study e-mails anonymizes patients’ participation and data in terms of its relationship with wearable device vendors and data acquisition/staging solutions. We do not expect study participants to actually use these study-assigned e-mails. However, there could be a protocol in which the study participants might use the assigned e-mails to access wearable device portals and resources that are completely independent of their personal e-mail accounts. Building on previous work, global unique identifiers are generated for each study participant using a hash function on the data source identifier and other known descriptors of the study participant [Reference Olden, Holle, Heid and Stark19].

A shared Common Data Model (CDM) and standardized language are capabilities critical to the overall successful execution of decentralized studies. As such, Tufts CTSI has implemented the Observational Health Science and Informatics (OHDSI) Common Data Model to represent EHR patient care and outcomes data. The Observational Medical Outcomes Partnership (OMOP) is a public-private partnership, chaired by the FDA, administered by the Foundation for the National Institutes of Health, and funded by a consortium of pharmaceutical corporations in collaboration with academic and clinical researchers. The OMOP established a research program seeking to advance the science of therapeutics’ safety surveillance applying observational healthcare data [Reference Stang, Ryan and Racoosin20]. The OMOP experiments demonstrated feasibility of establishing a common data model and standardized vocabularies that could accommodate different data types from various care settings represented by different source vocabularies in a manner that could facilitate cross-institutional collaboration and computationally efficient analytics. This supports the ability of connected systems used in the conduct of clinical research to effectively and securely exchange information, integrated into the data platform. This concept is important since many digital device technologies are developed by consumer health care companies and may not be intended to interface with a clinical data collection tool. The OMOP provides a standardized format that fosters patient data collection in a single space, facilitating accessibility and analysis.

To further enable data integration from decentralized trials to existing patient care, data are mapped to standard OMOP concepts and transfered to an OMOP study database. Then SMS text messaging is used to send and retrieve survey data, with survey questions and responses mapped to standard concepts, managed within an OMOP study database. Having study data encoded and managed within an OMOP database leverages standard tools and support from the international community of OHDSI investigators. Recognizing the technical challenges of conducting research across disparate observational databases, the Tufts team adopted the OMOP Common Data Model as a mechanism to standardize the structure, content, and semantics of observational data, making it possible to input statistical analysis code that could be readily adopted and re-used at every data site [Reference Overhage, Ryan, Reich, Hartzema and Stang21]. In our experience, the generalized and patient-centered design of the OMOP Common Data Model makes it a viable database schema to manage clinical trial data. The underlying value proposition is that OMOP Common Data Model supports standardizing data at study initiation, saving time and effort at study closure where data must be cleaned, normalized, etc., allowing for formation of study tracking, reporting, and dissemination based on established standards.

Building Blocks for a Decentralized Platform

With each unique decentralized trial applying digital health technology, a catalog of standards-based data acquisition solutions and relationships must be built with various vendors. Each trial-specific device, as well as in-hospital or clinic-based devices, is a Lego®-like piece requiring data mapping, patient-centered documentation, and a study deployment process. Each newly designed study can draw from this warehouse of Lego®-like pieces to build and customize a unique solution for conducting their specific trial. The use of the term Lego®-like is descriptive and intended to convey our approach to designing the software and process infrastructure as reusable components rather than “one-off” structures and solutions that are discarded at the end of each clinical trial. While every study will require different instruments and solutions, our informatics approach is based on decades of designing software and processes using Object Oriented Design/Programming (OOD/P) principles and Design Patterns. We are applying those well-tested principles from software development to the process of clinical trial platform construction. The Tufts CTSI Decentralized Trial Team provides researchers consultation, production assembly of components, and database building support for this work. In the ongoing Tufts CTSI-MIT, Long-COVID collaboration, engineering students work in small teams to assess and develop decentralized clinical trial components such as packaging, distribution, inventory management, and other components that continually evolving digital health technology specifics, patient-centric solutions, and logistics in accounting of and managing devices and processes (Fig. 4).

Figure 4. Decentralized clinical trial manufacturing process.

Ongoing Decentralized Trial Work at Tufts

Building on the successful implementation of our Tufts R-SPECT platform during the pandemic and our Long-COVID work, we are expanding the reach of an enhanced platform as a demonstration project within an N-of-1 trial design, first focusing on the chronic conditions of rheumatoid arthritis and asthma. N-of-1 trials are single-patient trials in which study participants are given FDA- approved treatments in a randomized order and a double-blinded fashion. Structured data collection is applied to determine the most effective treatment for that individual. N-of-1 trials are particularly amenable to our R-SPECT platform which enables seamless integration with consumer health devices, bi-weekly telehealth interaction with patients, collection of patient-reported data via mobile platforms and onsite structured clinical assessments. Selker and colleagues published “A Useful and Sustainable Role for N-of-1 Trials in the Healthcare Ecosystem” which presented a comprehensive review of opportunities and approaches for N-of-1 trials [Reference Selker, Cohen and D'Agostino22]. In this review, we acknowledge that standardized creation of N-of-1 trial software platforms and services has sustainability challenges. For rare and ultrarare diseases, and for precision treatments based on individual patient characteristics and responses, N-of-1 trials may provide a paradigm for the collection of data, homogenizing N-of-1 protocols and decentralized platforms applied within geographically diverse sites. While it is well known that the impact of decentralized trial methods varies, we propose that in conditions where the patient population is small and visits to a central research site may be very costly and burdensome to participants, there is an opportunity to evaluate further opportunities and challenges. Our research collaborators at the Tufts Center for the Study of Drug Development (CSDD) are evaluating the financial aspects of N-of-1 trial by conducting a longitudinal study with specific decentralized trial approaches and technologies, facilitating analysis by participant demographics, trial phase, trial complexity, and other key trial operations metrics [Reference DiMasi, Smith and Oakley-Girvan30].

We anticipate facilitating clinician engagement and patient participation by reducing the number of in-person visits via an adaptive “Clinical-Trial-in-a-Box.” This will allow integration of condition-specific electronic Patient-Reported Outcomes collection. To support participants in rural settings, we plan to collaborate with pharmacy chains that perform relevant laboratory tests. For participants without internet service, decentralized collection of self-reported data could be handled by Web-enabled platforms at local pharmacies and community centers.

We will use SMS text messaging to send and retrieve survey data, and questions and answers will be mapped to standard concepts and managed within an OMOP study database, as done now in our Tufts Splenic Tumor Assessment Tools (T-STAT) and Long-COVID-19 study. Having study data encoded and managed within an OMOP database takes advantage of standard tools and support from the international community of OHDSI investigators. As an open-science standard, it is also sharable and can be aggregated with other OMOP implementations and institutions.

Challenges to Decentralized Trials

As decentralized trials often employ wearable digital device technology, participants have different user interfaces. Lack of consistency across applications in terms of icons, text, fonts, color, and reactions to user actions, challenges seamless user experiences. As a consequence, requests for user support increase the need for responding study personnel who require training on various devices and applications. Although having only two predominant mobile phone operating systems (Android and Apple) reduces user experience variability to some extent, the service organization still must train research staff and have operational skills in multiple wearable devices and resources available to support the delivery of a decentralized trial, increasing trial costs.

In addition to user experience variability, the process of acquiring digital technology data is handled on an individual vendor basis, as there are no standards covering data integration and dissemination from wearable devices. AliveCor, for example, provides healthcare organizations with a central portal (KardiaPro) to manage all patients using their KardiaMobile devices to acquire and analyze ECG recordings. Other wearables such as Oura® and Garmin® have taken more highly customized approaches to data access and use agreements in enabling research teams to acquire patient data from their cloud systems. However, at present, none of these solutions subscribe to a common standard to represent the data. Open Authorization (OAuth) 2.0 is a common and modern standard for accessing resources from a third party and is being used to authenticate and execute daily downloads from AliveCor within our Tufts CTSI Long COVID-19 study. However, in our experience, not all third-party collaborators may wish to use this method for a particular study. For example, Oura® is another third-party wearable collaborator who currently prefers to use a secure batch data transfer method. It is important to have an infrastructure design that can accommodate and adapt to current and emerging standards.

Opportunities and Future Direction

Development and innovation in technology accompanied by the disruption brought on by the COVID-19 pandemic, has resulted in further evolution of clinical trial conduct post-pandemic. The structures and processes for delivering decentralized clinical trials are not yet flawless, but there is a growing consensus that this approach may help to support data sharing and collaboration across settings. The real-time acquisition and transmission of health-related data using home-based, monitoring devices and mobile applications to assess vital signs, symptoms, adverse events, and other research outcome measures are key drivers of future decentralized trial research.

Decentralized clinical trials may allow recruitment and retention of participants regardless of where they reside or work and offer a solution to the cost and time burden of traveling to trial sites for multiple study visits. Because of this flexibility, decentralized trials may foster diversity in enrollment. Evidence continues to evolve assessing exactly how decentralized clinical trials perform relative to otherwise similar clinical trials that didn’t apply decentralized methods.

Social Determinants of Health, including socioeconomic status, race, and ethnicity create barriers to basic medical care and prevent minoritized groups from participating in clinical trials- decentralized or other. As individuals from racial and ethnic minority groups as well as rural populations may disproportionately live in areas without clinical research facilities, decentralized trials may open access to research.

The use of wearables, such as activity trackers, may herald more inclusive participation if made available to a diverse population. Decentralized clinical trials may reduce costs associated with recruitment and retention compared to conventional in-person trial designs and reduce costs for participants as well. However, because of digital divides, there is a risk that decentralized trials actually lead to less diversity and poorer representation of the individuals with the target condition. Individuals who are Black and Hispanic, older, have lower educational attainment, have lower household income, and live in rural areas are least likely to use the internet or have broadband internet access [Reference Dahne and Hawk31]. Despite this, some recent decentralized trials have had success in increasing participation of minoritized groups [Reference Tan, Wu and Day25]. Given the importance of community engagement in decentralized clinical trials, we plan to apply our increased understanding of community-specific health needs and local structural and SDoH to promote increased study participant diversity and health disparities research. Integration of SDoH and community engagement will be key to our understanding of the impact that decentralized trial implementation could have on recruitment and retention, particularly of less often studied populations [Reference Wilkins, Schindler and Morris23–Reference Tan, Wu and Day25].

We acknowledge the small sample size of our decentralized niclosamide trial and this experience does not allow us to understand the range of challenges within the conduct of decentralized clinical trials. We continue to develop and implement our decentralized clinical trial platform as integral to our Long- COVID-study. Two additional remote trial platform pilot studies are currently underway at Tufts, a veterinary Tufts Splenic Tumor Assessment Tools (T-STAT) [Reference Burgess, Price and King32] in which SMS Text-based clinician and client remote outcomes survey is used and a pilot collaboration with MIT Lincoln Labs testing usability and data acquisition of Long COVID-19-related physiological parameters. Our work with N-of-1 trials presents an opportunity to evaluate a hybrid model to assess patient-reported symptoms and health-related outcomes to guide patient-centric management of chronic conditions employing the Tufts R-SPECT platform.