Decentralized Clinical Trials in the Era of Real‐World Evidence: A Critical Assessment of Recent Experiences

ABSTRACT

Since the first decentralized clinical trial (DCT) was conducted in 2011, there has been an increased usage of DCT due to its benefits of patient‐centricity and generalizability of findings. This trend was further expedited by the global COVID‐19 pandemic. We identified 23 case studies across various therapeutic areas and grouped them into different categories according to their purposes—by necessity, for operational benefits, to address unique research questions, to validate innovative digital endpoints, or to validate decentralization as a clinical research platform. We leveraged the estimand framework from ICH E9(R1) including its five attributes (population, treatment, variable, intercurrent event, and summary measure) to critically assess their design and conduct. Common trends, opportunities, and challenges were reported along with recommendations for future DCT. Of note, intercurrent events and associated handling strategies are largely not present when reporting DCT. This is an area that can impact study conclusions and require more dedicated efforts when designing new DCTs.

1. Introduction

In recent years, there has been a significant increase in the utilization of decentralized elements in clinical trials (hence decentralized clinical trials or DCTs), partially driven by health technology advancements and the need for more flexible and patient‐centric approaches to clinical research. DCTs aim to overcome some barriers associated with traditionally site‐based trials by leveraging digital health technologies (DHTs), including telemedicine, wearables, and direct‐to‐patient delivery of investigational products (IPs). The COVID‐19 pandemic has further accelerated the adoption of DCTs, as it highlighted the critical importance of maintaining patient safety and continuity of clinical studies during the time of restricted access to healthcare facilities. Studies have shown that DCTs offer numerous benefits, such as faster enrollment in the targeted population, improved participant engagement and retention, reduced burdens on participants and sponsors, and enhanced data collection and analytic capabilities. In the meantime, there are also challenges in the design, conduct, data analysis, and result interpretation of DCTs, which require careful pre‐planning, collaboration, training of involved staff, building end‐to‐end infrastructure, and statistical considerations to address unique challenges associated with DCTs.

As part of output from the American Statistical Association Biopharmaceutical Section Real‐World Evidence Scientific Working Group (ASA BIOP RWE SWG), the objective of this manuscript is to discuss DCT case studies through the lens of a companion paper [1]. This focused literature search utilized keywords of “decentralized,” “remote,” “virtual,” “direct‐to‐patient,” “digital,” “site less,” “pragmatic,” or “hybrid” from Google Scholar and www.clinicaltrials.gov over the period of Jan 2010–March 2024. This research is based on a convenience sample and focuses on randomized clinical trial evaluating medication (including vaccine) efficacy and safety profile. Studies with this objective were included in the assessment regardless of whether they were successfully completed. For studies that do not have such objective, a subset of studies was kept, specifically two pragmatic trials, three observational studies, and four studies with DHTs as intervention. This paper categorizes these case studies based on the nature, extent, and purpose of decentralization and discusses unique features, insights, and lessons learnt from these cases. The case studies shed light on different aspects of DCT design and execution, including best practices and limitations.

2. What Is a Decentralized Clinical Trial?

Decentralization is a modern approach to conduct clinical studies where some or all trial‐related activities take place outside of traditional clinical trial (TCT) sites. In the companion paper Chen et al. [1], figure 1 illustrates the relationship of DCT, TCT, and pragmatic clinical trials, while figure 2 describes fit‐for‐purpose data collection and clinical assessment in DCT. TCT sites are typically academic medical centers and have the advantages of a large volume of patients, well‐trained research staff, and established infrastructure for conducting clinical studies. Participation in TCTs is mostly limited to patients who are geographically close or can regularly make the trip to the selected sites. Therefore, a TCT may not be able to accrue a representative sample of the target patient populations, and its findings may have generalizability issues [2, 3]. Unlike TCTs that rely on traditional study sites, DCTs that leverage DHTs to enable remote participation of patients from the comfort of their own homes or other convenient locations may yield more generalizability. However, not all DCTs achieve this goal; active planning and seamless execution of all aspects are required.

There is a broad range of decentralized elements. A DCT can be fully decentralized if all trial‐related activities take place outside traditional trial sites or hybrid if some of such activities involve in‐person visits to the traditional sites. This innovative approach allows for greater accessibility and flexibility by removing participation barriers and offering patient‐centric solutions. DCTs can incorporate various elements, such as remote recruitment and consent, virtual visits, remote monitoring, and use of DHTs to track participants' health outcomes and to collect trial‐related data. Considering DCT's wider adoption, the US Food and Drug Administration (FDA) has issued guidance regarding the implementation [4].

3. Review of the DCTs Case Studies

DCTs reported in the literature thus far have included both fully decentralized and hybrid trials. Table 1 summarizes 23 studies with respect to study status, number of patients enrolled, and study start/completion dates. The first recorded DCT in literature is the REMOTE trial initiated in 2011, and the majority of identified DCTs had a start date between 2020 and 2022 coinciding with the global COVID‐19 pandemic.

TABLE 1.

List of decentralized clinical trials identified as of March 2024.

Case number	Study	Study status	Enrollment	Study start	Study completion	Fully DCT or hybrid
Decentralization by necessity
1	NCT04368728	Completed	43,548 (actual)	2020‐04	2023‐02 (actual)	Hybrid
2	NCT04470427	Completed	30,415 (actual)	2020‐07	2022‐12 (actual)	Hybrid
3	NCT04308668	Completed	1312 (actual)	2020‐03	2020‐05 (actual)	Fully
4	ISRCTN17149988	Completed	439 (actual)	2017‐03	2021‐10 (actual)	Hybrid
5	NCT04923464 (observational)	Completed	51 (actual)	2021‐06	2021‐12 (actual)	Fully
Decentralization for operational benefits
6	NCT05633602	Ongoing	700 (estimated)	2020‐04	2025	Hybrid
7	NCT04644315	Terminated	1 (actual)	2021‐05	2022‐05 (actual)	Hybrid
8	NCT05432466	Recruiting	150 (estimated)	2022‐11	2025‐11 (estimated)	Fully
9	NCT01694667	Completed	57 (actual)	2021‐09	2013‐05 (actual)	Fully
10	NCT02832063	Completed	372 (actual)	2016‐08	2017‐07 (actual)	Fully
Decentralization to address specific/new research questions
11	NCT05340309	Recruiting	37 (estimated)	2022‐12	2026‐12 (estimated)	Hybrid
12	NCT03924414	Recruiting	3500 (estimated)	2019‐11	2026‐10 (estimated)	Fully
13	NCT04584645	Completed	49,138 (actual)	2020‐09	2021‐04 (actual)	Fully
14	NCT02697916	Completed	15,076 (actual)	2020‐04	2020‐06	Fully
15	NCT04262206	Ongoing	20,000 (estimated)	2020‐09	2026‐30	Hybrid
Decentralization for validation at endpoint level
16	NCT03538262 (observational)	Completed	226 (actual)	2018‐10	2022‐03 (actual)	Fully
17	NCT04770285	Completed	386 (actual)	2021‐02	2022‐10 (actual)	Hybrid
18	NCT04701177 (observational)	Enrollment by invitation	100,000 (estimated)	2021‐03	2026‐11 (estimated)	Fully
Decentralization for validation DCT as a platform
19	NCT01302938	Terminated	18 (actual)	2011‐03	2012‐08 (actual)	Fully
20	NCT04471623	Completed	102 (actual)	2020‐08	2021‐03 (actual)	Fully
21	NCT04862143	Terminated	2 (actual)	2022‐03	2022‐09 (actual)	Hybrid
22	NCT04091087	Completed	66 (actual)	2020‐06	2021‐10 (actual)	Fully
23	NCT05780151	Ongoing	600 (estimated)	2023‐07	2025‐11 (estimated)	Hybrid

The reasons for decentralization may vary; they are summarized in Figure 1.

FIGURE 1.

Decentralized clinical trial types and primary reason for decentralization.

For each of the case studies, we summarize the primary reason for decentralization as one case study can potentially be classified into more than one category. One example is COVID‐19 vaccine trials conducted during pandemic that gain operational efficiency as well. However, without DCT, it is not practical when travel restrictions are in place. Therefore, they were grouped into by necessity. This is different from another case study where a smart watch was employed to continuously monitor atrial fibrillation episodes. That study can still be conducted at traditional trial sites, but DCT substantially reduced the cost and improved data collection.

As the intercurrent events and associated handling strategies were largely not reported in these case studies, four of the five elements of the estimand framework (namely population, treatment, variable [endpoint], and summary measure) were reported in the study summary, with intercurrent events specifically addressed in the Section 4.

3.1. Decentralization by Necessity

Compared with TCTs, DCTs offer flexibility across different aspects of study conduct. This flexibility is essential for continuing clinical studies, as evidenced by their resilience during the recent global COVID‐19 pandemic, when travel restrictions and social distancing became the norm. Many studies credited decentralization as the reason for avoiding disruption during this challenging circumstance, and it was considered a necessity by other studies for the collection of innovative endpoints via digital technology [5].

The most noticeable DCTs in this category are the two large COVID‐19 vaccine trials conducted in 2020:

Case 1

NCT04368728 is a Phase 1/2/3 placebo controlled, randomized, observer‐blind, dose‐finding study to evaluate the safety, tolerability, immunogenicity and efficacy of SARS‐COV‐2 RNA vaccine candidate against COVID‐19 in 43,548 healthy individuals [6]. This was a hybrid study that enabled the activation of 150 global sites in a few weeks after the study started in late July 2020 and its readout secured FDA emergency authorization in early December 2020. Vaccinations were administrated on‐site including observation of acute reactions by local healthcare workers. Electronic diaries (eDiary) were employed to record local or systemic reactions including solicited, specific adverse events that were prompted by eDiary and unsolicited adverse events reported by the participants. About 90% of medical monitoring was conducted remotely.

Case 2

NCT04470427 is a Phase 3, randomized, stratified, observer‐blind, placebo‐controlled study to evaluate the efficacy, safety, and immunogenicity of mRNA‐1273 SARS‐CoV‐2 vaccine in 30,420 health individuals aged 18 years and older [7]. Similarly, this was a hybrid study with trial activities mostly taking place at home, but local health care workers took post infusion assessments. For the primary endpoint of syptomatic COVID‐19 infection, it is centrally assessed by a blinded, independent committee.

The scale and the speed from study initiation to completion of these two studies were exceptional. Without decentralization, this would not have been possible in 2020 during the global pandemic. The close medical monitoring, timely adverse event collection via eDairy, centrally adjudicated efficacy endpoint ensures the quality of trial conduct is not adversely impacted.

There are also DCTs assessing treatments for COVID‐19 and other conditions:

Case 3

NCT0308668 is a prospective, randomized, double‐blind, placebo‐controlled study assessing hydroxychloroquine in non‐hospitalized adults with early COVID‐19 [8]. It was a fully decentralized trial with web‐based enrolment, recruitment via social media or traditional media, digital signature to consent, on‐line follow‐up survey based on multiple methods (proxy may be contacted). Internet searches for obituaries were performed to ascertain vital status if lost to follow‐up.

Case 4

ISRCTN17149988 is a double‐blinded, randomized, placebo‐controlled trial to evaluate the efficacy and safety of enterosgel (polymethylsiloxane polyhydrate) for the treatment of irritable bowel syndrome with diarrhea (IBS‐D) [9]. The trial used a hybrid design with decentralized components such as (1) general practitioners and “virtual sites” as trial facilities; (2) trial‐related patient activity was reduced to a short daily diary available online with daily text reminders. Paper diary was provided as an option for patients who were not able or willing to use eDiary; and (3) clinical visits were kept to three sessions. An independent Data Monitoring Committee was set up to monitor both efficacy, safety outcomes and can make recommendations regarding study conduct.

Case 5

NCT04923464 is a decentralized study to evaluate physical activity and cough frequency using wearable technology in cystic fibrosis. It was an observational, single‐arm, prospective cohort, phase 4 study. The purpose of the trial was to evaluate the performance of wearable technology in collecting physical activity, cough frequency, sleep quality, the compliance of using the wearable device and address COVID restrictions. These specific research questions requires DCT. All visits were home based and were conducted via telemedicine video conference via a mobile app.

For more details on the estimand attributes of DCTs in this category, refer to Table 2.

TABLE 2.

Estimand attributes of case studies where decentralization was implemented by necessity.

Study	Population	Treatment	Variable(s)	Summary measure
NCT04368728	Healthy population	30‐μg doses of BNT162b2 or Placebo	Confirmed Covid‐19 with onset at least 7 days after the second dose (21 days apart) in participants who had been without serologic or virologic evidence of SARS‐CoV‐2 infection up to 7 days after the second dose	Vaccine efficacy was estimated by 100 × (1 − IRR), where IRR is the calculated ratio of confirmed cases of Covid‐19 illness per 1000 person‐years of follow‐up in the active vaccine group to the corresponding illness rate in the placebo group
NCT04470427	Healthy population	mRNA‐1273 or Placebo	First occurrence of symptomatic Covid‐19 illness starting 14 days after second dose (28 days apart)	Percentage reduction in the hazard ratio estimated from a stratified Cox model
NCT0308668	Symptomatic and asymptomatic participants who have severe acute respiratory syndrome coronavirus 2 (SARS‐CoV‐2)	Hydroxychloroquine or Placebo	(1) Active COVID‐19 disease at Day 14 among those who were asymptomatic at baseline (2) Change in disease severity over 14 days among those who are symptomatic at baseline	(1) Incidence of COVID‐19 (2) Median (interquartile range) of severity score
ISRCTN17149988	IBS‐D patients aged 16–75 years old	Polymethylsiloxane polyhydrate or placebo	Composite abdominal pain (≥ 30% decrease in the weekly score) and stool consistency (50% reduction in days per week with at least one stool of BSFS type 6 or 7) score during at least 4 weeks of the treatment period	Proportion
NCT04923464	Adult CF participants who are currently on a stable regimen of commercially available Elexacaftor (ELX)/Tezacaftor (TEZ)/Ivacaftor (IVA)	ELX/TEZ/IVA	Compliance with actigraphy wearable device as measured by time a participant is wearing the actigraphy wearable device	Percentage of time

3.2. Decentralization for Operational Benefits

Flexible enrollment and patient convenience are key factors of DCT adoption. This covers rare diseases, specific segments within a disease where patients are hard to reach. It is also an attractive option in conditions where patients have limitations in mobility. From a sponsor's perspective, DCT can reduce resource commitment or shorten the study duration.

Case 6

NCT05633602 (Pragmatica Lung Study) is a phase 3 randomized clinical trial of a two‐drug combination to treat patients with advanced non‐small cell lung cancer (NSCLC) with a planned sample size of 700 [10]. This is one of the first NCI‐supported clinical trials (joint effort with FDA) to use a trial design that simplifies data collection and mainly focuses on collecting overall survival. The trial had fewer and simpler eligibility criteria than conventional trials, while still ensuring the safety of patients. As the safety profile in experimental arm (ramucirumab, pembrolizumab) have been well studied and characterized, a more frequent safety monitoring like those typically implemented in the conventional randomized trials may not be needed. The focus on overall survival (OS) is clinically relevant and there is a low risk of data quality issue due to the subjectivity of the endpoint.

Case 7

NCT04644315 (ALpha‐T) is an open‐label, single‐arm decentralized home‐based interventional phase 2 study evaluating the efficacy and safety of alectinib in patients with locally advanced or metastatic ALK‐positive solid tumors other than lung cancer [11]. The trial allowed enrollment regardless of geographical locations, allowed home‐based assessment by a mobile nurse (through face‐to‐face), with investigator support via telemedicine. Tumor assessments were performed at local radiology facility. The study is terminated early due to recruitment difficulty for this extremely rare condition.

Case 8

NCT05432466 (DiSCOVER) is a prospective, randomized, double‐blinded, placebo‐controlled, phase 3 study comparing the efficacy of celiprolol with that of placebo in patients with COL3A1‐ positive Vascular Ehlers‐Danlos Syndrome (vEDS). The study was fully decentralized with no specific study sites established and patients could participate in the study from home. This is an on‐going study and decentralization provides convenience for patients with the rare connective tissue disorder of vEDS to participate in.

Case 9

NCT01694667 is an internet‐based, randomized, double‐blind, placebo‐controlled phase 2 trial assessing the efficacy and safety of omega‐3 fatty acids versus placebo for treating hyperactivity in children with an autism spectrum disorder [12]. This was a fully decentralized trial where all procedures, including screening, informed consent, and collection of outcome measures took place over the internet which led to a marked reduction in time and cost. Omega‐3 fatty acids have been found to have a favorable safety profile. A procedure was embedded in the trial conduct platform such that whenever an adverse event was entered, principal investigators will be notified. They would call the patient to obtain necessary information and ensure patients' best interests are being protected.

Case 10

NCT02832063 is a randomized, double‐blind, DCT evaluating the safety, tolerability, and efficacy of B244 compared to placebo in the treatment of mild to moderate acne vulgaris [13]. Telemedicine and photographic technology adopted in this trial has long been used in dermatology to bridge the access gap and to open patient–physician interactions. The feasibility of patient‐mediated photography was assessed by comparing photos taken by trained staff to those taken by patients. The feasibility assessment showed strong inter‐rater reliability and supported the usage of photographs taken by patients.

For more details on the estimand attributes of DCTs in this category, refer to Table 3.

TABLE 3.

Estimand attributes of case studies where decentralization was implemented by choice for operational benefits.

Study	Population	Treatment	Variable(s)	Summary measure
NCT05633602 (Pragmatica Lung Study)	Adults ages 18 and older with stage 4 or recurrent NSCLC who were previously treated with immune checkpoint inhibitors and chemotherapy	ramucirumab + pembrolizumab or standard treatment	Primary endpoint: overall survival	Hazard ratio
NCT04644315 (ALpha‐T)	Adult Anaplastic Lymphoma Kinase (ALK)‐positive locally advanced or metastatic solid tumors other than lung cancer.	600 mg oral alectinib twice daily	Confirmed Objective Response Rate (ORR) as Determined by the Investigator per RECIST v1.1 (from 28 days after initial response up to 5 years)	Proportion of patients who achieved complete response and partial response per the time frame specified in the protocol
NCT05432466 (DiSCOVER)	Patients (15–64 years old) genetically confirmed as COL3A1‐positive Vascular Ehlers‐Danlos Syndrome (vEDS)	ACER‐002 (celiprolol) 200 mg or placebo BID	Time to first occurrence of a vEDS‐related clinical event requiring medical attention	Hazard ratio (when 46 events are accumulated)
NCT01694667	Children ages five through eight with an autism spectrum disorder (ASD; hyperactivity)	Omega‐3 fatty acids or placebo	Standard questionnaires of Aberrant Behavior Checklist (ABC), Social Responsiveness Scale (SRS), clinical Global Impressions Scale (CGI)	Change from baseline
NCT02832063	Male and female age 18 or older with clinical diagnosis of mild to moderate facial acne vulgaris	B244 or placebo	Inflammatory and non‐inflammatory lesion; Investigator Global Assessment (IGA) success	Count of lesions; proportion of IGA Success

3.3. Decentralization to Address Unique Scientific Questions

The patient‐centric approach of DCTs enables us to address unique scientific questions that may not be supported by TCTs. Examples include the effectiveness of home‐based care and the development of self‐injection formulation.

Case 11

NCT05340309 is an open‐label, single arm interventional phase 2 trial assessing the efficacy and safety of subcutaneous atezolizumab in the treatment of NSCLC patients [14]. This is a hybrid study to determine if a telemedicine‐based approach that gives a subcutaneous injection version of atezolizumab at home is safe and feasible. The study adopted remote consenting, included technology training and in‐office administration of subcutaneous atezolizumab at the beginning followed by home administration and telemedicine. It combined data reported by patients and from remote data capture smartphone applications.

Case 12

NCT03924414 (TOPAZ trial) is a randomized, placebo‐controlled phase 4 trial of a single infusion of zoledronic acid for the prevention of fractures in patients with Parkinson's disease or parkinsonism with at least 2 years of follow‐up [15]. This fully decentralized trial takes advantage of online consent technology, the capacity to confirm diagnosis using telemedicine and the availability of research nurses to provide screening and parenteral study therapy in homes. TOPAZ aimed to determine whether treatment that improves bone mineral density (BMD) and reduces fractures in people with osteoporosis also reduces fracture risk in people with parkinsonism. It tests a novel approach to treatment that any person with parkinsonism age 65 or older without contraindications to treatment would receive treatment without referral for evaluations or BMD testing. As a fully decentralized trial, TOPAZ was able to accommodate the following: (1) with a broad reach (US only) of a frail population (elderly PD population at higher risk of fractures), opening participation to almost any person with neurodegenerative parkinsonism, unlimited by their location; (2) Optimizing evaluation of treatment effect on clinical fractures (since neurologists may not be familiar with that aspect); (3) Reducing barriers to the clinical implementation of study results (no BMD testing needed).

DCTs can accommodate various trial designs such as pragmatic clinical trials (PCTs). Cases 14 and 15 are two large, practice‐changing PCTs where decentralized features generate evidence of effectiveness mimicking real‐world settings.

Case 13

NCT04584645 (CardioFlu: The Digital Flu Intervention for People with Cardiovascular Conditions) is a 6‐month prospective, digital, pragmatic randomized controlled trial to evaluate the effectiveness of an influenza digital intervention for people with cardiovascular (CV) conditions [16]. This fully decentralized trial is perhaps the first large‐scale RCT using digital interventions for increasing vaccination rates in people with CV conditions following similar efforts in general population or patients with diabetes. Considering the minimal‐risk nature of digital intervention and study, adverse events was not collected. This pragmatic trial has demonstrated the potential effectiveness of general messaging and incentives via a health‐related app to increase vaccine uptake in the general population, suggesting this kind of intervention could be effective in people with CV conditions.

Case 14

NCT02697916 (ADAPTABLE) is an open‐label pragmatic clinical trial in patients who are at high risk for atherosclerotic cardiovascular disease (ASCVD) with a sample size of 15,076 [17]. The purpose of the study was to identify the optimal dose of aspirin for secondary prevention in patients with ASCVD. The trial used existing electronic health records (EHRs), as well as a web‐based patient portal to collect patient‐reported outcomes (PROs), and available patient encounter data to supplement the EHR. Patients were directed to the electronic patient portal for the eConsent as well as an abbreviated eligibility confirmation and randomization. ADAPTABLE intended to engage patients, their healthcare providers, and trial investigators in using the infrastructure developed by PCORnet. Patients were followed either by e‐mail or phone call for any health care encounter, for example, hospitalization, adherence to trial medication and use of concomitant medication. A detailed plan was put in place to ascertain, reconcile, confirm endpoints before study initiation. Previously validated coding algorithms were distributed to participating health systems for identification of non‐fatal MI, stroke and bleeding. They were not verified by manual review of source documents. Instead, a random sample of 25 patients per clinical research network were selected at periodic intervals and adjudicated as in traditional clinical trial for validation purposes. All‐cause mortality was used as EHR or claims data do not routinely record cause of death. Data linkage was implemented to enhance the endpoint identification, and events outside the participating network are confirmed with queries to Medicare and/or via communication with patients' families or friends. In the case of reconciliation issue, qualified experts will review the medical records. All these efforts are to improve the quality of outcome ascertainment and consistency across different health systems.

Case 15

NCT04262206 (PREVENTABLE) is a multi‐center, randomized, parallel group, placebo‐controlled superiority study for reducing the primary composite of death, dementia, persistent disability of atorvastatin in older adults without cardiovascular disease (CVD) or dementia [18]. It plans to enroll 20,000 patients. Participants are identified using computable phenotypes derived from the EHR and local referrals from the community where they will receive either atorvastatin 40 mg or placebo. Participants will undergo baseline cognitive testing, physical function and a blinded lipid panel if feasible. Cognitive testing and disability screening will be conducted annually. Multiple data sources will be queried for cardiovascular events, dementia, and disability; survival is site‐reported and supplemented by a National Death Index search. Using in‐person or remote contact (therefore a hybrid DCT), the site confirms eligibility and enrolls those who consent. The Geriatrics Outcomes Assessment Center at Wake Forest University School of Medicine is responsible for baseline and annual phone‐based assessments of cognitive and physical function. Only a subset of study participants (approximately 2000) will return for a repeat lipid panel at 3 months. Adverse event data is collected through the EHR, rather than site reported. DSMB was established to monitor safety based on aggregated safety results.

For more details on the estimand attributes of DCTs in this category, refer to Table 4.

TABLE 4.

Estimand attributes of case studies where decentralization was implemented to address specific/new scientific questions.

Study	Population	Treatment	Variable(s)	Summary measure
NCT05340309	Non‐small cell lung cancer (NSCLC) patients who are eligible for treatment with atezolizumab for approved indications	Subcutaneous atezolizumab at a dose of 1875 mg every 3 weeks (Q3W)	Adverse events (AEs) up to 30 days after last dose; Successful completion of home drug administration visits up to 2 years	Incidence of AEs; mean number of successfully completed home visits per patient and proportion of patients who successfully completed all home administration
NCT03924414 (TOPAZ trial)	Men and women aged 60 years and older with Parkinson's Disease or parkinsonism	A single intravenous infusion of zoledronic acid (5 mg) infused over 45 min or placebo	All clinical fractures	Proportion of patients
NCT04584645 (CardioFlu)	US residents, 18 years or older, with CV conditions	Cardiovascular disorders digital intervention or without digital intervention	Self‐reported influenza vaccination status as vaccinated	Ratio of vaccination rates
NCT02697916 (ADAPTABLE)	Patients who are at high risk for Atherosclerotic cardiovascular disease (ASCVD); 96% previous received aspirin and among which, 85.3% already took 81 mg/day	Aspirin dose of 81 mg/day versus 325 mg/day	Composite endpoint of all‐cause death, hospitalization for MI, or hospitalization for stroke; safety endpoint of hospitalization for major bleeding with an associated blood product transfusion	Hazard ratio for time to the first occurrence of a composite endpoint
NCT04262206 (PREVENTABLE)	Community‐dwelling older adults (≥ 75 years old) without cardiovascular disease (CVD) or dementia (emphasizing the inclusion of minority populations and older adults with multimorbidity)	Atorvastatin 40 mg daily or placebo	Survival free of new dementia or persisting disability; composite of cardiovascular death, hospitalization for unstable angina or myocardial infarction, heart failure, stroke, or coronary revascularization; composite of mild cognitive impairment or dementia; LDL‐C reduction	Hazard ratio estimated from stratified Cox model

3.4. Decentralization for Validation of Endpoints

Some decentralized trials may serve as pilot or validation studies for specific endpoints collected remotely or assessing DCT as a platform to compare the results of onsite versus remote assessments.

Case 16

NCT03538262 (AT‐HOME PD) is an observational, longitudinal prospective cohort with a 2‐year follow‐up including participants from two former Phase 3 Parkinson's disease (PD) trials, STEADY‐PD III and SURE‐PD3 [19]. All study procedures, including screening, informed consent, and collection of outcome measures were conducted over the internet. One of the aims of this fully decentralized trial is to establish the concordance between in‐person and video‐based clinical assessments. To maintain patients' safety, a workflow on how to handle medical issues during video visits was implemented and medical services across locations were identified for medical emergencies.

Case 17

NCT04770285 (MIRAI) is a multi‐center, randomized, controlled phase 3 trial to evaluate the effectiveness and safety of two digital therapeutics in adult subjects diagnosed with major depressive disorder (MDD) who are on antidepressant therapy (ADT) monotherapy for the treatment of depression [20]. Trial visits were conducted remotely either by video visit or telephone. The screening visit may be performed in person at the discretion of the investigator. Eligible subjects were randomized to one of two digital therapeutics within a mobile application that will reside on the subject's personal smartphone. COVID‐19 constraint and convenience of DCT for digital therapeutics were also cited for choice of design.

Case 18

NCT04701177 (ANANEOS) is a digitally enhanced observational study remotely assessing the symptoms of preclinical stages in Alzheimer's disease and movement disorders using remote measurement technologies (RMT). This study leverages the uniformly collected data from a patient registry and hypothesizes that RMTs will allow better detection of impairment in functional components of Activities of Daily Living (ADL).

3.5. Decentralization for Validation of DCT Platform

Due to a lack of in‐depth understanding of its operational characteristics, some DCTs are designed and conducted to assess the performance of DCTs versus the TCTs.

Case 19

NCT01302938 (REMOTE) is the first reported DCT in literature. It is an exploratory, randomized, double‐blind, placebo‐controlled, single‐center, phase 4 trial to evaluate the efficacy and safety of tolterodine ER in subjects with Overactive Bladder. An entirely web‐based approach was implemented to streamline and improve the convenience of clinical trial participation with the objective of replicating previous clinic‐based trial findings. The study was terminated early due to inadequate enrollment.

Case 20

NCT04471623 (DeTAP) is designed to validate a combined DHT approach to fully decentralize an intervention to administer, promote, and track oral anticoagulant therapy in patients with atrial fibrillation [21]. The DHTs used include DeTAP App (data collection, televisit function, information, and reminders), bluetooth‐connected 6‐lead home (electrocardiogram) EKG device, bluetooth‐connected blood pressure (BP) cuff. Participants completed study surveys weekly to capture adherence and adverse symptoms.

Case 21

NCT04862143 (TELEPIK) is an open‐label, single arm, multi‐center, phase 2 interventional pilot trial. The primary objective is to assess participant satisfaction with the decentralized clinical trial (DCT) experience in breast cancer patients with HR‐positive/HER2‐negative Advanced Breast Cancer with a PIK3CA Mutation. Training on using telemedicine platforms and other monitoring devices was offered at study beginning in‐office. An end of trial visits was also in‐person at clinical sites with remote visits in between. Imaging for tumor assessment during remote participation would be performed at local facilities and transferred to investigators. This pilot study has since been terminated due to low enrollment.

Case 22

NCT04091087 is a randomized, double blind, vehicle controlled, fully decentralized phase 2b trial evaluating efficacy and safety of crisaborole in adults with stasis dermatitis [22]. This small study (N = 65) had three in‐home visits, six telemedicine interactions and successfully piloted an innovative siteless design. Efficacy evaluations were performed by medically qualified home visit practitioners (HVPs) at patients' homes and centrally by a board‐certified dermatologist with expertise in the disease. Similarly, safety assessments were performed by HVPs at patients' home and remotely by central readers. Efficacy according to success and improvement per Investigator's Global Assessment score and lesional percentage body surface area reached statistical significance based on central reader but not in‐person assessments.

Case 23

More recently with the increased usage of DCT, a pan‐European pilot study NCT05780151 (called RADIAL, Remote and Decentralized Innovative Approaches to Clinical Trials) is conducted to compare the scientific and operational quality of hybrid and fully decentralized clinical trial models with conventional brick‐and‐mortar settings. Results will help define legal, regulatory, ethical, and operational challenges and corresponding solutions, as well as be incorporated into recommendations for technologies that underpin successful DCTs [23]. Operational metrics include recruitment speed and diversity of enrolled populations. Performance indicators (KPIs) will be evaluated across the three operational models (conventional, hybrid, and fully remote). Tables 5 and 6 summarize the estimand attributes for validation of endpoints and DCT as a platform respectively.

TABLE 5.

Estimand attributes of case studies where decentralization was implemented for validation at endpoint level.

Study	Population	Treatment	Variable(s)	Summary measure
NCT03538262 (AT‐HOME PD)	All individuals with early idiopathic Parkinson Disease enrolled in the STEADY‐PD3 (NCT02168842) and SURE‐PD3 (NCT02642393) studies	None	Tele‐visit Modified MDS‐UPDRS Parts 1–3 (Total Score); Tele‐visit MDS‐UPDRS Part 2 (Score); Smartphone Tapping (Score); Fox Insight MDS‐UPDRS Part 2 (Score)	Mean change from baseline
NCT04770285 (MIRAI)	Subjects ages 22–64 with a primary diagnosis of major depressive disorder (MDD) who are on an antidepressant therapy (ADT)	One of two digital therapeutics within a mobile application that will reside on the subject's personal iPhone or smartphone	Montgomery‐Asberg Depression Rating Scale (MADRS) total score at Week 6	Change from baseline
NCT04701177 (ANANEOS)	Pre‐symptomatic Alzheimer's disease and movement disorders Healthy volunteers with negative AD biomarkers as control, subjects with preclinical AD and MCI due to AD dementia	Diagnostic tests using remote measurement technologies (RMT)	Detection of impairments in functional components of ADLs that occur below the threshold (Neurological Progression Index) of clinical scale detection or disability questionnaires	Change in Diagnostic Area Under the Receiver Operating Characteristic Curve (ROC‐AUC)

TABLE 6.

Estimand attributes of case studies where decentralization was implemented for validation of DCT as a platform.

Study	Population	Treatment	Variable(s)	Summary measure
NCT01302938 (REMOTE)	US females aged ≥ 21 years with self‐reported overactive bladder (OAB) symptoms for ≥ 3 months	Tolterodine ER 4 mg or placebo	micturitions/24 h (primary endpoint) after 12 weeks of treatment	Mean change from baseline
NCT04471623 (DeTAP)	Patient with atrial fibrillation	Single arm cohort of combined DHT approach	OAC adherence (OACA), and completion of televisits, surveys, and ECG and BP measurements	Change in pre‐ versus end‐of‐study OAC adherence
NCT04862143 (TELEPIK)	Men and pre and postmenopausal women with breast cancer with HR‐positive/HER2‐negative Advanced Breast Cancer With a PIK3CA Mutation	Alpelisib 300 mg daily in combination with fulvestrant 500 mg administered intramuscularly on Cycle 1, Day 1 and Cycle 1, Day 15, and on Day 1 of each cycle thereafter until Cycle 12	Participant satisfaction using the Trial Feedback Questionnaire (TFQ)	Change from baseline by visit
NCT04091087	Patients with stasis dermatitis without active skin ulceration	Risaborole ointment 2% twice daily for 6 weeks or placebo	Total Sign Score (TSS) at Week 6: in‐person assessment	Change from baseline
NCT05780151 (RADIAL)	Type 2 diabetes who have reached the stage of requiring basal insulin	Three arms of 300 patients at their homes, 150 patients in a hybrid setup (part home, part trial site) and 150 patients at a trial site	Operational metrics such as recruitment speed and improvement of enrollment of target population. Performance indicators (KPIs) will be evaluated across the three operational models (conventional, hybrid, and fully remote)	Mean of continuous variable, proportion for categorical outcome

4. Discussion

4.1. Common Trends and Insights From the Case Studies

DCTs have gained wider adoption reflecting a patient‐centric philosophy. They have been conducted for different purposes: by necessity due to extrinsic constraints, operational benefits, addressing unique research questions, and validating innovative endpoints and platforms. Their conduct was made possible by the evolution of DHTs. One such technology platform is NORA (Network Oriented Research Assistant), which has been used successfully in many FDA‐registered DCTs to assess skin and mucosal disease in the home‐based setting. NORA is a combination technology that includes the functionality of a telemedicine platform, an EHR, an electronic data capture (EDC), and a mobile data collection tool. NORA was recently used to compare digital photography with face‐to‐face acne scoring and has been used in other therapeutic areas as well. Some other examples that are routinely used include the collection of patient reported outcomes (PRO) via a web‐based platform or app (ePRO), an electronic diary that captures how patients feel, function, and respond (eDiary), DHT that enables quantification of innovative endpoints such as sleep quality, and telemedicine. They have also been credited with saving costs; for example, in the large pragmatic REACT‐AF study of over 5000 patients receiving treatments for atrial fibrillation [24], the decentralized components (including smartwatch to sense atrial fibrillation) are expected to reduce costs by 50% while also enhancing data collection. This approach will hopefully improve patient outcomes and advance the field of medicine by using more real‐world data.

Related to their complexity, the adoption of DHTs, EHRs, and telemedicine poses unique challenges for DCT conduct. First, different patient populations may have various levels of familiarity and openness in working with these technologies (vs. site‐based traditional RCTs). This can lead to reluctance to participate in a DCT, a nonrepresentative patient sample, or informative data missingness. Enabling equal access to technology, (as needed) providing the paper Case Report Form option for patients who are unwilling to use DHTs or when technology breaks down, a detailed training plan, and timely technical support during study conduct are some of the options to address this challenge. Second, employing diverse data collection mechanisms can increase variability in outcome (versus central lab, central adjudication) or lead to a higher proportion of data missingness when the technology fails. This was evident from Case 22 NCT04091087, where analyses of at‐home assessments tend to underestimate treatment effects and do not achieve statistical significance, while central rater results do. That is, they lead to different conclusions for one key efficacy outcome. Larger variability from at‐home assessment and different qualifications of at‐home assessors and central readers contribute to this phenomenon. Incorporating such factors in the power analysis at the study design stage and carefully ascertaining, reconciling, and validating the outcomes are essential for a high‐quality DCT (see Case 14 NCT02697916 as an example for validating outcome). Third, a pilot study or run‐in phase to test out the technology platform, identify issues and implement enhancements, continuously monitor data quality and enable workflow to immediately address abnormalities, and have an analysis strategy prespecified to foresee different scenarios are some of the best practices to mitigate this risk (see Chen et al. [1] for more comprehensive statistical considerations).

Regardless of remote or in‐person, DHT or traditional data collection, the requirements for data quality of both efficacy and safety remain the same. In the reviewed cases, many studies required on‐site in‐person evaluation for critical outcomes. Evaluations, tests conducted at local healthcare facilities need to meet specific criteria, for example, approved list of tests for confirming Covid infection; therefore, balancing convenience of local clinical assessment with qualification requirements. Specifically, regulatory guidance requires inclusion of a safety monitoring plan for adverse events capture and procedures to address them. Participants under remote monitoring need to always have the option to go to a clinical site or local HCPs if there is a safety concern.

The DCT model will unlock access to underserved populations (e.g., geographically dispersed, physically immobile) who usually do not have access to traditional clinical sites and thus better reflect the target population of clinical trials. In a recent review of 13 DCTs, 11 reported improvements in recruitment; 7 reported positive retention outcomes; 6 indicated a trend toward heterogeneity of population in terms of race and geographic location [25]. Another recent review paper of DCTs in the Trial Innovation Network [26] reported that while some decentralized components have worked well, some other elements such as remote recruitment and virtual patient monitoring require further improvement; for example, the first reported DCT in literature, the REMOTE study, was terminated early due to inadequate patient enrollment. Building trust with potential participants, promoting channels that can best reach potential participants (e.g., social media of patients with specific conditions), and building a global physician network that includes broad local health care professionals are some of the strategies that can improve DCT study conduct and avoid those observed in the literature with early termination due to inadequate participation.

We also included three case studies that are observational (Cases 5, 16 and 18). They are an important part of clinical studies but do not fall into the clinical trial category. These studies by design incorporate decentralized components as participants receive routine care and will not be asked to visit traditional clinical trial sites. The findings from observational studies can be more generalizable but require careful design and analysis strategies to minimize bias and account for confounding. The causal inference, including propensity score‐based methods, augmented inverse‐probability‐treatment weighting, and targeted maximum likelihood estimation, plays a central role in comparative effectiveness research.

Another category that usually adopts decentralized components is pragmatic trial. The PRECIS‐2 tool defines nine domains such as recruitment, setting, delivery of care, follow‐up, and primary outcome that allow different levels of pragmaticality over TCTs [27]. These studies bridge the TCTs via employment of randomization and real‐world studies by mimicking routine care. The ADAPTABLE study enrolling more than 15,000 patients is such an example of PCTs to compare two doses of aspirin for the secondary prevention of CV disease. It incorporated a range of innovative methods centering around patients and substantially lowered the cost via simplifying the patient recruitment and follow‐up. The ADAPTABLE trial reports no significant difference between the two aspirin doses for the composite CV outcome and hospitalization for major bleeding during the 26.2 months of mean follow‐up. One contributor to this finding is the potentially informative switching: 41.6% of patients randomized to the 325 mg dose switched to the 81 mg daily dose, and the majority of these switches happened before the first follow‐up between 1 and 3 weeks after randomization [28]. The intention‐to‐treat analysis does not address this intercurrent event of treatment cross‐over. Additional analyses, including aspirin dose as a time‐dependent covariate, show that 81 versus 325 mg was associated with a higher risk of all‐cause mortality and hospitalization for myocardial infarction or stroke (hazard ratio, 1.25; 95% confidence interval, 1.10–1.43). This study highlights the challenges and importance of work during the stage of designing a DCT. In addition to the scientific rigorousness, the operational aspects, the scenario planning, and risk mitigation are critical for addressing the research questions. In the case of the ADAPTABLE trial, understanding the aspirin dose usage pattern under routine clinical practice from fit‐for‐purpose real‐world data sources, identification of factors associated with dose change from patients' perspective, a pilot study to identify potential issues earlier, and adoption of a double‐blind design would be helpful at the design stage to mitigate the treatment switching risk during the study conduct.

Similarly, across the case studies being reported here, the estimand framework of ICH [29] and intercurrent event handling strategies are not well documented. To facilitate more adequate evaluation and robust analysis of the DCTs, better capture of the intercurrent events needs to be proactively planned and their handling strategy should be prespecified at the design stage. First, attention should be paid to data collection to ensure sufficient capture of relevant data while balancing patients' reporting burden. Second, foreseeing the main types of intercurrent events and prespecifying their handling strategies is a necessary component at the study design stage. Some intercurrent event handling DCTs are similar to TCTs, for example, if treatment discontinuation or switch is due to lack of efficacy or intolerability, clinically these patients experience treatment failure and non‐responder imputation is appropriate for subsequent office visits. There are also intercurrent events that are unique to DCTs and require special consideration, for example, if data outside office visits fail to upload solely due to technical difficulties, these missing data are not directly related to the treatment and missing at random is the reasonable underlying mechanism. Mixed‐effects for repeated measurements or multiple imputations may be considered. Further, sensitivity analysis is recommended to assess the robustness of findings and the impact of confounders (including potentially unobserved ones). The companion paper [1] discusses in more detail the statistical considerations of planning, designing, executing, analyzing, and reporting a successful DCT.

In the cases being reviewed, there is also a general lack of analysis specifically addressing the heterogeneity of data collected from different mechanisms, for example, on‐site versus remote. The observed effects may reflect a mixture of many factors and could deviate from the intended research questions. Visualizing data patterns, conducting stratified analysis, and assessing consistency of effects across clinically relevant subgroups will be able to assess and account for the inherent heterogeneity. Careful assessment of the study findings relative to TCTs and other data sources will facilitate the interpretation and put the results in context.

Many DCTs are designed to validate innovative digital endpoints, and their successes will enable the wider adoption of such technology. Similarly, DCTs that are conducted to quantify decentralization as a clinical research platform will generate new insights and increase confidence in their usage.

4.2. Limitations

This paper reports the learnings from 23 DCTs from a focused literature search. These cases cover medication, vaccine, and digital intervention across different therapeutic areas. However, the search strategy does not follow the rigorousness of a systematic literature review; for example, no protocol, lack of PRISM for a detailed selection process, and thus a convenience sample up until March 2024. Therefore, selection bias may have been introduced, resulting in potential skewing of observations or recommendations. In the meantime, the mandatory registration of clinical studies at www.clinicaltrials.gov and a thorough search of this database with relevant key words mitigate to some extent this bias. For example, many cases were terminated earlier and still included in this assessment for lessons learned.

Another limitation from the estimand perspective is that the DCT cases are more concentrated in vaccines, rare diseases, and chronic conditions. Interventions mostly have an established safety profile or can be conveniently administered at home settings by patients. Outcomes are more likely to be DHT‐based or clinical outcomes that are hard to miss, for example, all‐cause mortality. Therefore, the suitability of DCT and recommended best practices for a broader patient population, diverse treatments, and various outcomes will require customized assessment on a case‐by‐case basis.

As this paper relies on publicly available information, there is inconsistency regarding the level of details across cases. This also highlights the importance of improving DCT reporting standards, especially study components that are unique and have implications for study conclusions, such as data collection, quality monitoring, intercurrent events and their handling strategy, and sensitivity analysis for robustness. Additionally, this paper does not address some of the common topics and concerns associated with DCTs, for example, ethics considerations, privacy, confidentiality, security, and data breaches. Lastly, many statistical considerations are more comprehensively reported in the companion paper Chen et al. [1] and are not covered again here.

4.3. Recommendations for Future DCT Designs

DCTs provide opportunities for improving the conduct of clinical research and generating findings that are more generalizable. There have been numerous successful examples of DCTs that greatly improved public health and benefited patients, for example, COVID‐19 vaccine trials, large pragmatic trials that are practice changing. It is also well positioned to execute complicated study design. The Beat Acute Myeloid Leukemia (BAML) Master Trial sponsored by the Leukemia & Lymphoma Society is built upon a streamlined operational approach and innovative use of digital technologies [30]. The master protocol simultaneously tested several targeted therapies using genomic profiling of specific acute myeloid leukemia subtypes to determine which sub‐study the patients will be enrolled in. The e‐technology beneath this master protocol includes clinical oversight, e‐protocol, e‐source upload, electronic data capture, and data review platforms around the “patient‐first” principle. Some of the challenges encountered in this study are relevant to DCTs in general, for example, more planning and ramp‐up time due to the complexity, a steep learning curve for study personnel to enter data, and delayed upload to the data platform.

Considering the importance of DHTs, FDA issued guidance on DHT for remote data acquisition [31]. The companion paper [1] details the statistical and practical considerations during both design and analysis stages. This includes the validity of outcomes captured by DHT, the heterogeneity resulting from the different data collection mechanisms, safety monitoring, the importance of sufficient training to all key players, pre‐specification of estimand framework and intercurrent events handling in study protocol and statistical analysis plan, documented implementation plan, reevaluation and improvement of operation among others. As evidenced by case studies being reported here, not all of them have been common practice in the current conduct of DCT. Table 7 summarizes the opportunities, challenges, and our recommendations across research questions, estimand attributes, and study execution. With the learnings from the ongoing validation trials comparing DCT versus TCT, best practices from successful case studies, and lessons learned from challenges, decentralized clinical trials will play an even more important role for drug development in the era of real‐world evidence.

TABLE 7.

Opportunities, challenges, and recommendations when designing and executing DCTs.

	Opportunities	Challenges	Recommendations
Research questions	Reflect patient‐centric philosophy and more generalizable findings Release of guidance from health authorities Ability to address unique research questions	TCTs are considered as gold standard when practical DCT as a platform still needs to establish credibility	Assess different study design options case‐by‐case and adopt DCT only when appropriate Proactively engage health authorities and incorporate feedback upfront
Estimand attributes
Population	Reach patients that are otherwise not candidates of TCTs Faster enrollment and larger sample size	Patients have different disease characteristics and may respond differently Fraud or not meeting inclusion/exclusion criteria at enrollment	Factor population heterogeneity into power calculation Implement patient authentication and PIs co‐sign the consent
Treatment	More convenient for patients Established mail order delivery, home visiting, or administration by local health care providers	Loss or damage to medications during transition Unclear instruction or patients not following instruction Suboptimal treatment adherence	Document patients' receiving medication and quantity, quality Track treatment administration and build in reminder
Variables	Integration of innovative outcomes, e.g., DHTs Enhanced data collection, e.g., high‐frequency collection	Validity of outcomes can be questioned Higher variability due to no central lab or central rating	Adopt hybrid model and central rating of critical outcomes Require common standards and employ validated outcomes Incorporate outcome validation within study
Summary measures	Mostly similar to TCTs	Handling of high‐frequency data Effect size differs from TCTs	Adopting appropriate statistical methods Conduct sensitivity analysis for robustness
Intercurrent events	Established handling strategies Methodology available for accounting for them	More complicated intercurrent event types More frequent intercurrent events Impacting conclusions if not appropriately accounted for	Forsee potential intercurrent events and prespecify handling strategies Collect relevant data including mechanism Piloting or run‐in
Study execution	Wide availability of electronic health records and continued revolution in DHTs that enable decentralization in trial conduct Overcome extrinsic constrains Saving cost and shorten trial study period	Patients have different preferences for data collection and familiarity with technology Participating HCPs may not be familiar with clinical research Risk of technology failure Comprised data quality and protection	Offer different data collection options Provide patient and site training, and timely technical support Build DCT platform that integrates ethics, security, privacy, different data capture mechanisms Implement data monitoring plan

Disclosure

The opinions expressed in this manuscript are those of the authors and should not be interpreted as the position of the U.S. Food and Drug Administration.

Conflicts of Interest

The authors declare no conflicts of interest.

Wang H., Daizadeh N., Shen Y.‐L., et al., “Decentralized Clinical Trials in the Era of Real‐World Evidence: A Critical Assessment of Recent Experiences,” Clinical and Translational Science 18, no. 9 (2025): e70328, 10.1111/cts.70328.