Wearable Technology in Clinical Trials 2026: Digital Endpoints, FDA COA Qualification, and Real-Time Safety Monitoring

Consumer wearables have been generating clinical-quality physiological data for years — continuous ECG, SpO2, step count, gait parameters, sleep staging. The bottleneck has never been data collection. It's been regulatory acceptance. The FDA's COA qualification process requires sponsors to demonstrate that a wearable-derived endpoint measures a specific, pre-defined construct (not "gait" in the abstract, but "free-living gait speed in habitual daily activities"), that changes in the measure correspond to changes patients would recognize as clinically meaningful, and that the measurement is reliable across populations, environments, and device versions. Several digital endpoints have now cleared or are near clearing this bar in 2026 — which changes what's possible in trial design in concrete, practical ways.

FDA-Qualified Digital Endpoints: What's Been Cleared and What's Pending

The Parkinson's qualification is the most instructive example of what the COA process actually requires. The critical distinction is not "measuring gait" — it's "measuring habitual free-living gait speed as the patient moves through their actual daily environment." This is fundamentally different from timed 10-meter walk tests performed in a clinic corridor, because the clinic performance is subject to test-anxiety effects, Hawthorne effects, and is a 30-second sample of a performance patients can temporarily optimize. Continuous accelerometer data over seven days captures habitual function without these biases. The validation studies had to demonstrate correlation with the timed walk test at a pre-specified r threshold, and patient focus groups had to confirm that changes in the wearable-derived measure correspond to changes patients would recognize and find meaningful in their daily lives.

What the COA Qualification Process Actually Requires

The FDA's Digital Health COA qualification pathway requires three demonstrations that go beyond standard analytical validation:

• Concept of interest (COI) definition: A precise, operationally testable definition of what the endpoint measures. "Spontaneous free-living gait speed during habitual activities in the home environment" is a COI. "Mobility" is not. The specificity of the COI definition determines the scope of the endpoint's regulatory utility — a narrowly defined COI may be well-validated but limited in generalizability across disease stages or populations.
• Measurement equivalence or anchor study: A study demonstrating that the digital endpoint captures the same clinical construct as the established comparator measure — with quantitative correspondence established at a pre-specified threshold. For gait, this meant demonstrating that accelerometer-derived gait speed correlates with supervised timed walk tests at a level that supports the claim that both measures assess the same underlying construct. The anchor study design (which gold-standard comparator, which population, which correlation threshold) requires pre-agreement with FDA via a qualification meeting.
• Meaningfulness to patients (MTP) demonstration: Patient focus groups and quantitative threshold analyses (often using Rasch analysis or anchor-based methods) demonstrating that a minimum clinically important difference (MCID) in the digital endpoint corresponds to a change patients would identify as meaningful in their functioning or quality of life. This step exists because a statistically significant change in a digital biomarker is worthless if patients cannot perceive it.

The qualification submission is resource-intensive and takes 18–30 months to complete. The incentive to do it: once qualified, the endpoint can be used as a primary endpoint by any sponsor for the indicated disease and patient population, without repeating the validation work. The Digital Medicine Society (DiMe) maintains a Library of Digital Endpoints cataloging qualification status and methodology for endpoints that have completed or are undergoing the process.

Real-Time Safety Monitoring: Where Wearables Change Trial Operations

Beyond endpoint measurement, wearable devices in 2026 trials are generating continuous safety data that fundamentally changes adverse event detection timelines. The applications with the most documented impact:

• Cardiac safety in oncology Phase 1 trials: Continuous ECG monitoring via wearable patch detects QTc prolongation and arrhythmias within hours of drug administration, versus the standard 12-lead ECG performed at scheduled visits — typically every 4 weeks in early-phase oncology trials. Real-time cardiac surveillance has enabled same-day dose holds in multiple oncology Phase 1 programs, preventing serious arrhythmic events that would have been undetected between visits. This is particularly important for kinase inhibitors and antibiotic-oncology combinations with known QT-prolonging potential.
• Fall detection in neurology trials: Wearable accelerometers detecting fall events in Parkinson's disease, multiple sclerosis, and ALS trials capture safety data that patients substantially under-report in self-reported diaries. A 2024 comparative study published in NEJM Evidence demonstrated 3x higher fall event capture with continuous accelerometry versus patient-reported diary entries — with falls being a critical safety endpoint for drugs targeting motor function and potentially confounding efficacy assessments if systematically missed.
• Sleep-based safety signals in CNS trials: In trials targeting insomnia, anxiety, or mood disorders, overnight actigraphy provides a continuous, objective measure of sleep fragmentation that may serve as an early safety signal for investigational CNS drugs that would otherwise only appear in weekly clinical assessments. Several CNS Phase 2 trials now include wearable-derived sleep endpoints as pre-specified safety secondary outcomes specifically because of this detection sensitivity advantage.

What Participation in a Wearable-Enabled Trial Looks Like

For participants, the practical experience varies considerably depending on the device used. Patch-based continuous ECG monitors (iRhythm Zio, Cardiac Insight CORR) are typically worn for 14-day periods, require minimal interaction, and are mail-returned for analysis. Wrist actigraphs or research-grade activity monitors are worn continuously, sometimes for the entire trial duration, and synchronized via Bluetooth to trial systems at home or clinic visits. CGM devices require sensor insertion (subcutaneous for most approved devices) and calibration.

Adherence is the practical limitation that underlies all of the statistical power arguments. A device worn 60% of nights generates data of limited interpretability for circadian and sleep analyses. Most wearable trial protocols specify minimum wear time requirements for data quality exclusion criteria, and some trials include behavioral adherence support (reminder notifications, participant engagement applications) to maintain data completeness above the threshold needed for valid endpoint analysis. If you're considering a trial with wearable components, asking specifically about the wear time requirements, what happens to your data when the trial ends, and whether there are visit implications if device data quality falls below minimum thresholds are reasonable questions to raise at the screening visit.