Catching Non-Adherence in Decentralised and Home-Dosing Trials

Participants taking their medicine at home is nothing new. Most oral-dose Phase 2 and Phase 3 trials have worked this way for decades: dosing happens between visits and is reconciled when the participant next attends the site. What has changed is how much site contact a trial assumes. Decentralised and hybrid designs deliberately reduce visits, and some remove them almost entirely, which lengthens the stretches when no one sees how dosing is actually going.

Adherence has always been the hardest thing to see once dosing happens away from the clinic, and the fewer the site touchpoints, the harder it gets. This guide covers why the methods trials have relied on struggle as visits thin out, what home dosing actually looks like in our own data, and how per-pill capture with real-time alerting keeps adherence visible when there is no visit to catch a problem.

The blind spot between site visits

Most established adherence methods are anchored to a site visit. A pill count reconciles what is left in the bottle against what was dispensed; a diary is reviewed when the participant next attends. Both give you one reading per visit and infer everything in between.

In a conventional trial with frequent visits, those gaps are short. In a decentralised or home-dosing trial they are not. The more dosing happens away from the clinic, the longer the stretch of time a visit-based method cannot see, and the later any problem surfaces. By the time a pill count reveals a shortfall at week twelve, the non-adherence it reflects happened weeks earlier, when something could still have been done about it. For a full comparison of the established methods, see our guide to monitoring adherence in clinical trials.

Methods that depend on the participant fail where you need them most

A second group of methods asks the participant to do something: complete an electronic diary, open an app, or record and upload each dose for video-observed therapy. These can work well in engaged populations. The difficulty is that the participants least likely to complete an extra task are often the same ones whose adherence is slipping, so the method fades exactly where it is needed.

A decentralised trial also removes the scaffolding that normally props up those tasks. There is no waiting-room reminder and no coordinator on hand to prompt. What is left has to survive on the participant’s own initiative on an ordinary Tuesday at home. The design principle we work to is to ask nothing of the participant beyond taking their dose: a dose is dispensed with a button press, the event is captured and transmitted on its own, and there is no app to open or upload to remember. The lower the demand on the participant, the less the signal depends on the very behaviour it is trying to observe.

What home dosing actually looks like

We deployed dispensers in a technical evaluation to see how dosing behaves in a home setting. Eighteen volunteers dispensed twice daily for around two weeks. It was a small validation study with no interventions, so it does not support broad conclusions, but the pattern in the data is worth reporting, because it is the kind of thing only per-pill, timestamped capture can see.

Missed dispenses did not occur in isolation. When a volunteer missed, it tended not to stop at one: lapses clustered into consecutive days. Early behaviour was telling too. The three volunteers who missed their first two days went on to be the lowest adherers overall, which suggests a struggling participant can be identifiable within days of starting rather than months.

The contrast with a pill count is the point. For one volunteer, a count at the end would have shown six pills remaining in the bottle. What it could not show is that those six were three consecutive missed days rather than the occasional isolated dose spread across the fortnight. Depending on the compound, three missed days in a row may warrant a very different response from a handful of scattered misses, and a count cannot tell the two apart. The pattern is the information, and the pattern only exists if each dose is captured as it happens.

Reliable self-administration is what makes home dosing possible

Home and remote dosing only works if participants can be trusted to self-administer reliably, without a clinician observing each dose. That is the capability a decentralised design depends on, and it is worth being specific about the evidence.

In a study with the University Medical Center Groningen, published in the European Respiratory Journal in 2022, tuberculosis patients used our dispenser to self-administer their treatment as an alternative to directly observed therapy, with a dispense adherence of 99% recorded. Tuberculosis is a demanding test, because the standard regimen runs for months of daily dosing and has traditionally relied on a healthcare professional watching each dose. The study was a hospital-setting proof of concept rather than a decentralised trial, so we are careful not to overstate it. What it demonstrates is that patients can self-administer with a digital adherence technology at high reliability, which is the precondition for moving dosing out of the clinic. As with any dispensing-based method, it records doses leaving the device rather than confirming ingestion, a distinction we keep to.

Closing the loop when there is no visit

Capturing dosing at home solves visibility; it does not, on its own, change outcomes. What changes conduct is what happens to the data. Because each dispense transmits on its own, the events can be watched in real time rather than waiting to be read at a visit. When a pattern of missed or delayed dosing appears, an email reaches the site team with what happened and when, so the team can act on it while the participant is still enrolled and the dosing can still be corrected. The loop then closes on a record: what was detected, the action the site took, and the dosing that followed. In a decentralised trial, where there may be no near-term visit to surface the problem any other way, that push model is often the only thing standing between a developing lapse and a non-evaluable participant. This is the shift from seeing adherence to acting on it, covered in more depth on real-time monitoring.

What to look for in a decentralised adherence method

If dosing in your trial happens away from the clinic, the method you choose should hold up without a visit and without leaning on the participant. In practice that means looking for:

• Low participant burden. Capture that asks nothing beyond taking the dose, so the signal does not depend on an extra task.
• Data that transmits on its own. Events that reach you between visits, not only when a device is scanned or a bottle is returned.
• Alerting by exception. The signal reaches the site team, rather than waiting in a dashboard for someone to log in and find it.
• A documented record. Detected, acted on and resolved, as evidence for RBQM and audit.

Which device records what, and how far each sits from a confirmed dose, is set out in our guide to the smart pill bottle.

A decentralised trial does not have to mean flying blind between visits. The tools exist to keep adherence as visible at home as it would be in the clinic, and to turn a missed dose into an early conversation rather than a number discovered too late. See how the dispenser captures each dose at the bottle.