Decentralized clinical trials: research that comes to the patient

A trial participant with lupus drives two and a half hours to an academic medical center, waits forty minutes, fills out a symptom questionnaire she could have answered at her kitchen table, gives a blood sample her local lab draws routinely anyway, and drives two and a half hours home. The protocol calls this a site visit. Her calendar calls it a lost day — one of fourteen the study will ask for over two years.

Most people who decline to join clinical trials are not declining the science. They are declining the commute.

What 'decentralized' actually means

A decentralized clinical trial moves some or all trial activities away from the traditional research site and to wherever the participant is — home, a local clinic or lab, a phone. It is a spectrum, not a switch. Fully remote trials exist, but most decentralized trials in practice are hybrids: screening and a few anchor assessments at a site, everything else moved outward.

The framing matters because regulators treat decentralization the same way. The FDA's guidance on conducting clinical trials with decentralized elements, finalized in 2024, describes a menu of design choices available to ordinary trials — not a separate category of research with looser rules. The investigational product is unchanged. The endpoints are unchanged. What changes is the geometry: instead of the patient orbiting the site, the trial orbits the patient.

The elements, piece by piece

Each decentralized element removes one reason to travel — and adds one obligation in return.

Read the right-hand column carefully; it is the honest part of the table. Decentralization does not make trials simpler. It relocates the difficulty from the participant's calendar into the sponsor's data architecture. That trade is usually worth making — but only if someone actually builds the architecture.

The pandemic was the forcing function

Remote trial methods were not invented in 2020, but 2020 made them unavoidable. When COVID-19 closed research sites, regulators allowed remote consent, televisits, and direct shipping of study drugs so that ongoing trials could survive — and sponsors learned in months what conference panels had debated for a decade: much of what happens at a site never needed the site.

Lawmakers then made the shift durable. The Food and Drug Omnibus Reform Act, passed by Congress at the end of 2022, directed the FDA to clarify how decentralized trials should be run, and the agency's final guidance followed in 2024. Large public projects had already shown the model at scale: the NIH's All of Us program enrolls and follows most of its participants without a traditional trial site ever entering the picture.

So "decentralized" is no longer an experiment in how to run experiments. It is a normal answer to the oldest operational problem in research — trials that cannot find or keep participants. We have written about that problem, and the large share of trials that miss their enrollment timelines, in why most clinical trials miss enrollment deadlines.

The plumbing nobody photographs

A decentralized trial generates data from more places than a site-based trial, and every place is a seam. The marketing photo shows a participant on a sofa with a phone. The actual system behind that photo collects a home nurse's vitals, a local lab's results, a wearable's continuous stream, an ePRO app's diary entries, and a televisit's clinical notes — five sources, five formats, one dataset that must hold up to inspection years later.

Three properties have to survive the seams. Identity: every measurement must attach to the right participant, even when no coordinator watched it happen. Provenance: every value must say where it came from and through whose hands it passed. Consent: the participant's permission must travel with the data, including the right to withdraw it mid-study. None of this is exotic. All of it is unforgiving, because a gap discovered at database lock cannot be repaired retroactively.

The practical foundation is the same standards work rewiring the rest of healthcare — FHIR for structure and exchange, and the disciplines of real-world evidence for turning routine, device-collected, home-collected measurements into research-grade ones.

Who gains the most

Chronic-disease patients — because travel burden is regressive. It falls hardest on exactly the people whose conditions research most needs to understand: people with fatigue and pain, mobility limits, caregiving duties, hourly jobs, and long distances to academic centers. Site-based research quietly samples for people well enough, free enough, and close enough to show up, then generalizes its findings to everyone else.

For chronic illness the mismatch compounds. These patients already manage care scattered across six specialists and six portals; a protocol that adds fourteen site days to that load selects for the least burdened — the mild cases, the early cases, the lucky ones. Decentralized designs invert the selection. When the trial comes to the kitchen table, the cohort starts to look like the disease.

This is the world Mithras Health is built for. The platform is patient-mediated research-data infrastructure for chronic disease — 28 conditions across 12 categories — where patients hold their own records, with FHIR R4 as the system of record, and grant each study a revocable, scope-limited consent. A decentralized trial needs consented, well-structured, findable participants before the first televisit is ever scheduled; that is precisely the starting state the platform maintains.

And because participation fees post only on IRB-approved research-use events, the economics travel with the data: the patient who carries half the trial in her pocket is paid when research actually uses what she contributes — not thanked in the acknowledgments and billed for parking.

Research that comes to the patient is not a convenience feature. It is a sampling correction. The trials that make the trip will describe disease as it actually lives — in people with jobs, distances, and bad days — and not merely as it appears in the subset well enough to drive.