Measuring the added benefits of decentralized trial elements (SHASTA): impact on enrollment, demographics, and data completeness through a single sponsor

Abstract

This paper explores the impact of decentralized clinical trial (DCT) elements through a single Sponsor’s perspective, measuring outcomes such as enrollment rate, baseline demographics, and data quality. Analyzing data from six studies – including four interventional and two observational – provides insights into the efficacy and challenges of implementing DCTs. While decentralized approaches offer enhanced data completeness due to direct access to medical records and ancillary care, they also present significant data collection burden. For example, obtaining and abstracting relevant information from medical records is a substantial effort. The study highlights that while DCTs are promising, they are not a universal solution. Having more tangible metrics around benefits and limitations will help researchers design more fit-for-purpose approaches. Moreover, the skills and infrastructure developed for DCTs can pave the way for future decentralized healthcare. As patients become increasingly digitally savvy, the integration of technology into clinical trials is expected to enhance convenience and improve informed health decisions.

Introduction

Advances in medicine depend largely on prospective clinical trials. While these trials provide the current gold standard for answering critical safety and efficacy questions, they do so at a high cost, time, and burden to patients and caregivers. Concerns around trust and commitment to participate have also historically translated into slow recruitment and limited involvement from the broader community. As such, evidence generated through these trials may not adequately reflect the drug’s safety and effectiveness in a real-world setting across all populations.

In oncology, it has been estimated that only 2–8% of adult participants enroll in clinical trials due to rigid eligibility criteria and other barriers, such as access to healthcare services^1–3. The clinical trial model for any therapeutic area has traditionally been location-centric, demanding the participant’s physical presence at a trial site for consent, sample collection, clinical assessments, vital signs, investigational medicinal products, and adverse events⁴. One study showed that approximately 40–50% of U.S. patients with advanced-stage cancers reside at least a one-hour drive from the nearest clinical trial site recruiting for their specific condition⁵. This statistic is further exacerbated by a steady decline of sites per clinical trial over the past 15 years⁶. Given this pervasive disconnect between patient homes and trial locations, there is an urgent opportunity for sponsors to adopt more flexible trial designs to reduce the burden of participation, support inclusion of broad populations, and ultimately provide more meaningful evidence about the safety and efficacy of critical medicines.

Trials with decentralized elements can address some of these geographical constraints experienced in traditional trials. Previously known as virtual or digital clinical trials, decentralized clinical trials (DCTs) are not limited to technology. A DCT is a clinical trial where trial-related activities occur at locations other than traditional clinical trial sites⁷. DCT elements include the use of local healthcare facilities, local healthcare providers (HCPs), local clinical laboratory facilities, visits to participants’ homes, and direct distribution of Investigational Product (IP) to trial participants at their choice of location.

Some DCT elements are more established and some elements are novel. For example, electronic consent (eConsent), electronic clinical outcome assessments (eCOA), home nursing, and the use of remote participant monitoring tools such as thermometers, pulse oximetry, blood pressure, and weight scales are routinely considered during clinical trial design. Approaches such as direct distribution of the IP or supplies to trial participants, leveraging local HCPs, setting up a network of locations under the oversight of the investigator (e.g., “central hub”), and more advanced remote monitoring tools such as continuous physiological measurements are being explored. Together, these existing and novel DCT elements and the mindset shift of creating studies to be inherently flexible are the foundation for the strategy and execution of sustainable clinical trials.

Prior to the COVID-19 pandemic, DCTs were largely viewed as a “nice-to-have” without sufficient evidence about benefits and return-on-investment. During the pandemic shutdown, participants and healthcare providers sheltered-in-place and stopped non-essential in-person visits to avoid spreading COVID-19. Many health systems and healthcare providers, including the Centers for Medicaid and Medicare Services supported telehealth reimbursement and permitted participants to seek remote visits with clinicians from different states. This benefitted patients particularly in rural areas, according to an analysis conducted between 2017 and 2020 that evaluated Medicare Beneficiaries⁸. Through telehealth-enabled routine care, the health system became more familiar with enabling technology to deliver continuity of care. Similar trends were seen in the clinical trial setting as well. Surveys conducted by the Food and Drug Administration (FDA) and Clinical Trials Transformation Initiative similarly identified that almost all (95%) sponsors incorporated at least one DCT element during the pandemic in oncology trials intended for regulatory submission. These results provide early signals that oncology trials with DCT elements can generate data suitable for FDA review and marketing approval⁹.

Despite these pandemic success stories, industry skepticism about DCTs remains. In fact, some surveys report that only 1% of the total number of new trials since 2022 contain DCT solutions¹⁰, demonstrating a hesitation to fully embrace and scale a more decentralized approach. As with most innovative designs, initial hype can set unrealistic expectations about the impact and lead to a period of disillusionment impeding further investment. Leading industry experts classify the current sentiment of DCTs to be in the trough and advocate for more data from actual use cases to move the industry forward into acceptance and practice. Unified metrics that clearly define value and measurement of decentralized elements are critical to compare execution across the industry¹¹. Pre-competitive efforts in metrics development are underway to capture key performance indicators for DCTs that will help the industry standardize how we measure the success and failures of these new advancements in clinical research^12,13.

Given the skepticism and varied evidence collection across the industry, we explore the impact of DCT elements through a single Sponsor, focusing on enrollment rate, baseline demographics, and data quality. We examine data from four interventional and two observational studies, with two of these being randomized controlled trials. These six studies have variable start times before and after the COVID-19 pandemic, which complicates the analyses of our results. As such, we focus our attention not only on the interpretation of the combined evidence, but also on observations gathered throughout the studies. Given the heterogeneity across studies, we cannot pool results for statistical interpretation but rather report trends.

We hypothesize that DCT elements can bring quantifiable improvements in clinical trial efficiency despite increased complexity during setup. We posed the following research questions:

1. What is the improvement in clinical trial enrollment efficiency compared to historical benchmarks, defined as the number of participants enrolled per site per month?

2. What is the improvement in enrollment of the demographic minority population compared to historical benchmarks, specifically African American, Asian, and Hispanic participants?

3. How do we measure data quality obtained in a DCT vs. brick-and-mortar approach?

4. What have been some notable clinical operational learnings that can be applied across DCT studies?

Materials and methods

We selected studies that contained one or more of the following DCT elements: eConsent, telehealth, remote participant monitoring, home health care, direct-to-participant shipping, local healthcare providers, central hub, or utilized eCOA as a primary endpoint. The full definition of each DCT element can be found in Appendix 1. We focused our analysis on North American studies to ensure the findings are directly applicable to the U.S. healthcare delivery system and clinical practice patterns. As a U.S.-based Medical Affairs team, our primary objective was to evaluate evidence generated within the specific regulatory and reimbursement landscape in which these interventions are utilized.

The enrollment rate is defined as the total number of participants enrolled divided by the number of months from First Participant In to Last Participant In. Enrollment rate efficiency is the enrollment divided by the number of sites. For example, an enrollment rate of 6 participants across 2 sites per 1 month would be 3 participants per site per month.

We identified 12 recent studies that included DCT elements in the protocol [Figure 1]. We then excluded two studies using eCOA outside of the primary endpoint requirement. We excluded 1 study because it is part of a larger study included in the analysis and participant data would have been double-counted. Studies need to have enrolled at least 3 participants for enrollment analysis and at least 10 participants for diversity analyses to protect patient privacy.

Fig. 1.

Studies included in the analysis.

We performed a pooled analysis using weighted averages, intentionally omitting a controlled comparator arm. This non-randomized approach was selected to capture a diverse amalgamation of trials incorporating DCT elements, thereby providing a more representative view of real-world implementation. Furthermore, because these studies spanned heterogeneous therapeutic areas and designs, a singular active comparator was neither available nor appropriate for this portfolio-wide assessment.

Historical benchmark data was utilized to provide a backdrop of comparison for the recruitment and demographic analyses. SHASTA oncology benchmarks were derived from large-scale meta-analyses encompassing diverse cancer types to reflect broad industry standards for oncology recruitment¹⁴. SHASTA non-oncology benchmarks were anchored to longitudinal acute stroke data¹⁵. This specific baseline was selected because the TIMELESS trial, a late-window stroke study, represents the largest cohort within the SHASTA non-oncology enrollment efficiency analysis, making stroke-specific metrics the most relevant comparator for assessing the impact of DCT elements in this subgroup.

Results

6 studies are included in the SHASTA analysis. The most common DCT elements used were eConsent (n = 5), telehealth (n = 4), a central model or remote coordinating center (n = 2), and eCOA as a primary endpoint (n = 2) (Table 1).

Table 1.

Brief Summary of Studies Included in the Enrollment Analysis.

Study/n/Study Type	Objective/Therapeutic Area	DCT Element(s)
TIMELESS n = 448 Ph III	Determine the efficacy and safety of tenecteplase compared with placebo if given between 4.5–24 h window; acute ischemic stroke	● Consent given by another individual: participants enrolled via a caregiver en route to the hospital
POTENTIAL n = 4 Observational	Determine the feasibility and reproducibility of a DCT; metastatic prostate cancer or metastatic non-small cell lung cancer	● eConsent ● Telehealth ● Remote participant monitoring ● Direct-to-participant ● Central hub *
Expanded Access Program (EAP) n = 164 EAP	Enable continuity of care during the COVID-19 pandemic; breast cancer	● eConsent ● Telehealth ● Direct-to-participant ● Home health care
XTEND CIU n = 206 Ph IV	Evaluate the efficacy and safety of subcutaneous omalizumab as an add-on therapy through 48 weeks for treatment of H1 antihistamine refractory chronic idiopathic urticaria	● eCOA as primary endpoint
CATORI n = 61 Observational	Estimate the proportion of a American Indian/Alaskan Native participants who are seen by a specialist for advanced carefollowing a referral from PCP to specialist; oncology, ophthalmology, neurology	● eConsent ● Telehealth ● eCOA as primary endpoint ● Central hub †
WeSMA n = 387 Ph IV	Evaluate the long-term safety of risdiplam in participants with spinal muscular atrophy (SMA)	● eConsent ● Telehealth ● Remote coordinating center ‡

The table above describes the objectives and various DCT element(s) in each study design.

^* POTENTIAL Central hub activities (B&M site): conduct participant screening, participant enrollment via eConsent, direct-to-participant shipment, routine checkups via telehealth, telehealth appointments, participant satisfaction questionnaires; review wearable data.

^† CATORI Central hub activities (vendor): conduct participant enrollment via eConsent, direct-to-participant recruitment, telehealth appointments, participant surveys, and centralized data collection.

^‡ WeSMA remote coordinating center activities (vendor): conduct medical record review, safety data review, and centralized data collection.

Learnings

Research question 1

What is the improvement in clinical trial enrollment efficiency compared to historical benchmarks, defined as the number of participants enrolled per site per month?

We found that supplementing traditional trial approaches with DCT elements is associated with higher enrollment rate efficiencies compared to historical rates, measured by participants enrolled per site per month [0.46 vs. 0.40 for oncology and 1.015 vs. 0.26 for non-oncology, Fig. 2]. We analyzed oncology and non-oncology trials separately to account for their respective differences in enrollment dynamics, screening complexities, and patient populations inherent to these fields. The availability of distinct benchmark data helped to anchor our analyses as well. Oncology benchmarks were derived from large-scale meta-analyses across diverse cancer types, while non-oncology benchmarks were anchored to acute stroke data. The latter was selected because the TIMELESS trial represents the largest study within the SHASTA non-oncology enrollment efficiency analysis^14,15. This division ensures that our findings more accurately capture interstudy variability and reflects standard approaches in clinical trial recruitment analyses.

Fig. 2.

Combined Enrollment Rate Histogram. Note: This analysis encompasses a wide variety of disease states, ranging from rare diseases, which present challenges in identifying sufficient populations, to more prevalent conditions such as stroke. Furthermore, the COVID-19 pandemic impacted some studies, and therefore enrollment timelines may be skewed. This may influence the enrollment rates. Caution should be exercised when broadly extrapolating these findings.

WeSMA and CATORI are the only studies designed with both brick-and-mortar (B&M) and virtual sites. Participants had the option to enroll in either experience. The enrollment timelines for WeSMA and CATORI are different for the B&M and DCT elements; therefore, separate enrollment rates were calculated first and then a weighted average was calculated.

Research question 2

What is the improvement in enrollment of the demographic minority population compared to historical benchmarks, specifically African American, Asian, and Hispanic participants?

We compared each study against distinct diversity benchmarks given that some diseases disproportionately affect certain races and ethnicities. If a disease benchmark was unknown, the standard clinical trial benchmark was used from the 2020 FDA Drug Trials Snapshots Summary Report: 75% Whites, 8% Black or African American, 6% Asians, and 11% Hispanics¹⁶. For acute ischemic stroke, previous studies indicate that clinical trials enrolled 76% White, 19% Black or African American, 2% Asian, and 11% Hispanic ethnicity¹⁷. For SMA, previous neuroscience studies indicate that trials typically enrolled 85.6% Whites, 7.1% Asian, and 1.6% Black or African American¹⁸ (Fig. 3).

Fig. 3.

Diversity analyses compared to historical benchmarks. Note: some studies may have prioritized enrolling broader populations due to the nature of their scientific rationale or the necessity to address evidence generation gaps for specific groups. It is important to note that the presence of DCT elements does not directly cause broader patient population enrollment. Each study has unique objectives, and some may have aimed specifically to recruit minority populations. Therefore, the observed increase in their enrollment should not be attributed solely to the inclusion of DCT elements. Note: We excluded AI/AN and unknown patients from the graphs above due to insufficient data across studies for a reasonable comparison.

Research question 3

How do we measure data quality obtained in a DCT/hybrid vs. standard brick-and-mortar (B&M) approach?

We selected one study in particular to answer this research question because of its unique setup as a hybrid interventional study with both the B&M and DCT arm.

Data quality measurements are based on characteristics such as accuracy, completeness, consistency, validity, uniqueness, and timeliness. Higher quality data can be achieved by the standardization, inspection, and evaluation of the data-generating processes and tools. Herein we define data quality as the accuracy and completeness of data maintained over time and across sources sufficient to provide analyzable data supporting study endpoints. We also define database completeness as ≥ 80% of critical data fields, based on previous work that demonstrated that a data completeness of ≥ 80% is an appropriate target for DCT/hybrid studies¹⁹. Data accuracy is monitored and tracked in ongoing studies by query rates for select data related to endpoints. In the SHASTA analysis, we focus on data completeness (Table 2).

Table 2.

Data completeness for select endpoints in an ongoing study¹.

Data Element	B&M Sites	Virtual Site	Overall
Baseline
Mandatory ethnicity	331/332 (99.7%)	52/55 (94.5%)	383/387 (99.0%)
Mandatory disease type	328/332 (98.8%)	51/55 (92.7%)	379/387 (97.9%)
Optional COA #1	105/195 (53.8%)	12/43 (27.9%)	117/238 (49.2%)
Optional COA #2	104/195 (53.3%)	9/43 (20.9%)	113/238 (47.5%)
6-Month
Optional COA #1 2	53/195 (27.2%)	2/43 (4.7%)	55/238 (23.1%)
Optional COA #2 2	48/195 (24.6%)	3/43 (7.0%)	51/238 (21.4%)

1. This study is a real-world data collection study with B&M and virtual participation options. Virtual sites conduct electronic medical review for data collection.

2. Includes all enrolled patients who have completed 6 months in the study and are past the eligibility window for the 6-month assessment. Some subjects may have completed other baseline and post-baseline efficacy assessments not reported here. These COA represent the assessments with the highest completion rates.

For required data collection, virtual sites are comparable to the B&M sites. However, for the optional COA assessments which are more burdensome to collect, B&M sites are more likely to generate more complete data due to having direct access to the medical records, the patients, and ancillary care records. The data is considered more burdensome to collect due to the level of effort needed to obtain the medical record then locate and abstract the relevant information. For data collection of COAs such as physical therapy assessment, B&M sites may have greater access to complete assessments with individual scores while virtual sites may only have access to total score if assessments are available in the medical record at all.

The overall data completeness for mandatory baseline data fields is 97.9–99.0%. We selected ethnicity and disease type data because we felt these would be the most translatable data elements to other hybrid studies. The overall data completeness for the selected baseline efficacy assessments is 47.5–49.2%. Of these patients, 21.4–23.1% have completed the 6-month same selected baseline efficacy assessments. The data is considered optional because data is collected per standard of care which varies from site-to-site and patient-by-patient, and some patients may not have a 6-month visit. This results in lower data completeness rates for optional assessments, such as COA #1 and COA #2, and compounds the issue of low data completeness.

However, as this is an ongoing study and a snapshot of ongoing data completeness, we expect the completion rate to change as the study progresses. Our intent is to provide a glimpse into data completeness in an ongoing hybrid study. Caution should be exercised when extrapolating learnings.

Research question 4

What have been some notable clinical operational learnings that can be applied across DCT studies?

There have been positive learnings and areas of growth. Starting with the positives, we identified that a remote coordinating center can maximize enrollment beyond the more limited proximity of a traditional B&M site. These centers differ from study to study; B&M sites may choose to manage participants virtually and in other studies, a vendor may manage participants virtually. For WeSMA, all virtual patients enrolled via a brochure distributed through a central pharmacy. This impacted the enrollment rate projections, which was rapid during the times of two mass brochure distributions.

Through this central model, participants also experienced stability during the study regardless of the location of the PI. In CATORI, the PI relocated to 3 different states and continued to enroll and oversee participants in the study due to the PI’s possession of appropriate licenses and certifications across state lines. A consistent touchpoint is critical for building and maintaining trust for participants, especially in populations that may be newer and less trusting of clinical research. Future DCTs conducted in the US should consider the PI’s medical licensures across different states. Broad cross-state coverage should be considered when designing and selecting the PIs for a study to enable flexibility.

When provided with both options, some participants selected the DCT experience over the B&M approach. This is seen in the manner of enrollment across studies and the manner of engagement in POTENTIAL, WeSMA, and CATORI. Some participants select for eConsent and some participants select for telehealth vs. in-person visits, particularly in observational/low-interventional studies.

These positive experiences are not without challenges. It is essential to proactively plan and consider incorporating DCT from the outset, as adding a virtual arm later in the process poses significant challenges. In addition, standard setup processes required additional clarifications for vendors providing CRO services versus virtual sites collecting study data. Vetting technology platforms to meet sponsor standards for participant privacy and security added another layer of complexity. There were also uncertainties about how participants would perceive DCT experiences and concerns about the virtual staff’s ability to build rapport compared to B&M staff. Maintaining high participant engagement is challenging, especially in studies with fewer face-to-face interactions.

Vendors need to have direct experience and technical expertise relevant to the study. For example, in one study, a vendor’s lack of experience in reviewing medical records for a rare disease led to incomplete data. Vendors must have their study staff understand and be trained in specific disease terms and have a proven track record in medical record retrieval and data handling for similar conditions.

DCT elements are relatively new, posing unknown and potentially riskier operational processes. Investigators need alignment on logistics, particularly concerning home health nursing and maintaining patient oversight and accountability. An internal audit found mobile nursing posed a higher regulatory risk due to additional oversight channels, suggesting the need for a robust risk management plan.

Discussion

Using appropriate methods, we combined data from 1,069 participants enrolled in six clinical trials across oncology and non-oncology areas, all conducted by the same sponsor. Our analysis suggests that deploying select DCT elements—eConsent, telehealth, central hubs, and eCOA as a primary endpoint—is associated with early signs in improving enrollment rate efficiency. It remains uncertain whether DCT elements enhance the enrollment of demographic minorities, though slight improvements were observed among Asian, Black and Hispanic populations in select studies. These results align with findings from the Tufts Center for the Study of Drug Development PACT, where representation is higher among select demographic subgroups²⁰.

To compare DCT data quality with industry standards, it must be evaluated against a standard registrational trial. However, we can still gently observe that for one ongoing study that includes both B&M and the DCT site, data completeness is comparable for mandatory data that are considered available and accessible. For optional data that are more difficult to collect, such as COA #1 and COA #2, B&M sites are more likely to produce more meaningful data than virtual sites. As mentioned previously regarding the individual and summary scores for collecting standard of care COA: when there are multiple data sources providing similar or same data, teams should be aware of possible variance to ensure data continuity and consistency. Also as previously noted, this is a snapshot of an ongoing study and we expect data completeness to change as the study progresses. At this stage, it is crucial to build expertise while learning to adjust to new environments and ensuring both the technology and support can achieve the study goals. It is vital to be flexible when designing the data management strategy for hybrid studies with real world evidence-type medical records collection. We now know how to better design a study with these elements to mitigate some of the data quality and completeness issues. Also, with increasing access to real world data and ability to link, we can explore how to fill in gaps while preserving the trending benefits (speed, diversity) of DCT approaches. While we cannot yet conclude that data quality is comparable in B&M vs. DCT, continuous improvement should help achieve comparable quality over time. Regulators emphasize addressing data flow complexity and potential variability in DCTs by mitigating data quality risks from the outset. Insights from SHASTA trials are instrumental in developing systematic approaches to preserve data integrity and reliability in DCTs.

Quantifiable improvements in DCTs come with added operational complexities, such as creating new workflows and logistical plans. This demands strong clinical research knowledge, data science familiarity, and health technology literacy. A systematic and collaborative approach is essential to properly implement the DCT model. When technology and remote services are central to the design, it is important to include data strategy leaders and data engineers to ensure data flows and system integrations are sound. It is no longer just about designing a scientific study with a good operational plan, but also a sound technical strategy and plan. Our approach has several strengths and limitations. To our knowledge, this is the first peer-reviewed report to aggregate and share specific trial efficiency metrics for DCT elements from a single sponsor’s portfolio, using weighted averages. We separated oncology and non-oncology trials to account for therapeutic differences. Limitations include variations in study budget, regulatory intent, and the impact of the COVID-19 pandemic. For example, the enrollment rate for TIMELESS may have been skewed by COVID-19, and results from CATORI and POTENTIAL may not be generalizable due to specific participant demographics and site-to-sponsor partnerships. Additionally, we don’t yet have enough data to evaluate the impact of remote patient monitoring, home health care, direct-to-participant shipment, and local HCP use on trial efficiency.

While DCT components are not new, our federated approach is innovative. By examining the bundle of DCT services and technology holistically, we can assess the ROI for the entire suite. Unified metrics and publications will help dispel myths about DCTs and modernize clinical research. The DCT strategy is a fit-for-purpose approach, not a panacea. Realistically assessing its contributions is essential. The infrastructure and skills developed for DCTs provide a foundation for future decentralized healthcare delivery. As patients become more digitally savvy and data-literate, they will expect to use digital tools and telehealth services in their clinical trial journey for more informed and convenient health decision-making. These approaches can reduce healthcare system pressure by requiring in-person care only when necessary. Lessons from DCTs will help us build expertise, take action towards implementation, and continue to evolve decentralized, patient-centered care approaches that optimize access and evidence needs.

Appendix 1

Definitions of DCT elements

• eConsent: Electronic form that may include multimedia components such as images, audio, videos, diagrams and a digital signature to aid the collection of the informed consent of a participant²¹.

• Telehealth: The use of electronic information and telecommunications technologies to support and promote long-distance clinical health care⁷.

• Remote participant monitoring: A type of telehealth in which healthcare providers monitor participants outside the traditional care setting using digital medical devices, such as weight scales, blood pressure monitors, pulse oximeters, and blood glucose meters. The data collected from these devices are then electronically transferred to providers for care management²².

• Home health care: Encompasses a wide range of health care services that are given to a participant in their home. A variety of providers may be involved, including but not limited to home health nurses, phlebotomists, doctors, among others. This care is typically provided during home health visits²¹.

• Direct-to-participant: Direct shipment of clinical supplies and investigational medicinal products (IMP) to the participant’s residence or other agreed upon location²¹.

• Local healthcare providers: Healthcare providers contracted to provide trial-related services that are part of routine clinical practice (e.g., performing physical examinations, reading radiographs, obtaining vital signs) and where a detailed knowledge of the protocol, IMP, and the investigator’s brochure is not necessary. These individuals are not listed on Form FDA 1572 as sub-investigators, however, are included in a task log⁷.

• Central Model or Remote Coordinating Center: Protocol-trained staff who can oversee and conduct the research of participants who enroll in the trial. They can identify, enroll, retain, and manage participants during the duration of the study^23,24.

• eCOA as a primary endpoint: An electronic measurement based on a report that comes directly from the participant (i.e., study subject) about the status of a participant’s health condition without amendment or interpretation of the participant’s response by a clinician or anyone else. Data collected contributes to the primary objective of the study²⁵.

Author contributions

AY and EM wrote the main manuscript text. SA and BT were key contributors to specific sections in the text. JD, RS, and RG added language that connected the different expertise together. KB provided oversight and general positioning of the manuscript and the conclusions. All authors reviewed the manuscript.

Data availability

Data is provided within the manuscript files.

Declarations

Competing interests

The authors declare no competing interests.