What the research says decentralised trials actually struggle with

Most conversations about decentralised trials still happen at the level of promise: wider reach, less site burden, participants who don't have to travel. A 2025 scoping review took a different approach and simply looked at what actually happened across 53 published reports of completed decentralised studies. The gap between the promise and the reported reality is instructive.

The studies themselves were more varied than the label suggests

The reviewed trials spanned mental health research (18 studies) and COVID-19 research (11 studies) as the two largest categories, with the remainder covering pharmaceutical interventions (24 studies), nutritional research, medical devices, and behavioural interventions. Most were conducted in the United States (34 of 53). Recruitment, screening, and consent were commonly handled remotely across the set, while outcome measurement varied more, mixing online, in-person, and telemedicine approaches depending on what the study actually needed to measure.

That variation matters because it undercuts a common assumption: that "decentralised" describes one consistent model being applied across research. It doesn't. It describes a spectrum of choices about which specific steps get moved off-site, and the review's authors flagged this directly, noting that terminology around decentralisation was inconsistent across the studies they examined. Two trials both calling themselves decentralised could mean genuinely different things by it, which makes comparing outcomes across studies harder than it should be.

Retention and recruitment bias, not technology, were the recurring problems

When the review looked at what actually went wrong, the two challenges that came up repeatedly weren't about software reliability, connectivity, or device compatibility, the problems a lot of vendor conversations focus on. They were:

• Participant retention. Studies frequently struggled to keep participants engaged through to completion, a problem that remote delivery doesn't automatically solve and can sometimes make harder if it removes the routine face-to-face contact that used to double as a retention check.
• Recruitment bias. Multiple studies reported biased recruitment, meaning the participants who ended up enrolling didn't represent the population the study intended to reach. Remote recruitment channels have their own selection effects, favouring people with reliable internet access, comfort with digital tools, and the flexibility to engage on their own schedule, which is not the same population a site-based recruitment drive would reach.

Neither of these is a technology problem in the narrow sense. They're study design and participant experience problems that happen to surface differently once a study removes the physical site as a fixed point of contact. A platform that renders forms correctly and syncs data reliably doesn't automatically solve either one.

The Denmark study points at a different kind of barrier entirely

A separate qualitative study of Danish policy stakeholders, drawn from patient groups, healthcare institutions, industry, and political bodies, found something that complements the scoping review's findings rather than repeating them. Rather than technical or regulatory detail, the recurring theme was what the authors called a "knot of paradoxical tensions" among stakeholders with genuinely different interests in how decentralisation should work.

Their recommendation is worth sitting with: implementation goes more smoothly when responsibility sits clearly with one stakeholder, or a small, defined group, who can actually guide the process, rather than being diffused across every party with an opinion. That's a governance finding, not a technical one, and it explains something the scoping review's data doesn't directly address: why some decentralised trials execute cleanly and others stall despite using comparable tools. The tools were rarely the differentiator. Clear ownership of the decisions around them was.

What this means for how a study actually gets designed

Put the two findings together and a clearer picture emerges than "decentralised trials are harder" or "decentralised trials are easier." They're different, in ways that shift where the effort needs to go:

• Retention needs a deliberate remote-equivalent, not an assumption. If a site visit used to be when a coordinator noticed a participant was struggling, quietly checked in, and kept them engaged, that function needs a designed replacement, not a hope that messaging and reminders will cover it.
• Recruitment channels need auditing against the intended population, not just against enrolment targets. Hitting a recruitment number doesn't mean hitting the right recruitment number, and a scoping review flagging bias as a recurring problem is a signal to check this explicitly rather than assume digital reach equals representative reach.
• Somebody needs to actually own the decentralisation decisions. Not a committee, not a shared responsibility distributed across sponsor, site, and vendor. A defined owner who can resolve the inevitable trade-offs between competing priorities before they turn into implementation drift.

None of this argues against decentralised or remote-first approaches. The reviewed studies covered a genuinely wide range of interventions and mostly got their trials done. What the research actually supports is more specific and more useful than a general endorsement or a general warning: the failure modes are known, they're about people and process rather than platforms, and a study that plans for them explicitly is working from real evidence instead of a marketing pitch in either direction.