From Hype to Protocol: Where Digital Biomarkers Actually Live in 2026

For most of the last decade, digital biomarker announcements outpaced digital biomarker adoption. Every ASCO, DIA, and SCOPE Summit produced another wave of decks about how wearables and smartphones would revolutionize clinical development. The wave rarely translated into modified protocols. In 2026, that gap is finally narrowing, but not evenly.

A recent scoping review of sensor-based digital health technologies used to capture clinical trial endpoints found that more than 1,000 drug trials have now embedded wearables at some point in their design, spanning phases 1 through 4 and multiple therapeutic areas.1 Yet the same body of work makes clear that most of these deployments are exploratory. Sensor-derived measures are being collected, but they are not the endpoints on which regulatory decisions are being made.

The distinction matters. Collecting accelerometer data as a secondary or exploratory endpoint is now routine at large sponsors. Building a submission where a digital endpoint changes what regulators actually approve is still rare. That is the sorting axis for 2026: which therapeutic areas have made the leap from data collection to decision-grade endpoints, and which are still gathering evidence?

1,021Drug trials that have used wearables through 2025, per a Brain Trials analysis of registered studies2
0Sensor-derived endpoints fully qualified through the FDA Drug Development Tool program as of late 20252
1Digital measure qualified by EMA as a primary endpoint in a pivotal drug trial: stride velocity 95th centile in DMD3

The gap between activity and regulatory acceptance is the story of digital biomarkers in 2026. Sponsors have moved past whether to include a wearable. They are now grappling with whether the endpoint that wearable produces will withstand a regulatory pre-submission meeting. In most cases, it will not, at least not yet.

The practical filter for sponsors in 2026. The right question is not “should we include a digital measure?” The right question is “what is our regulatory strategy for the specific measure we are proposing, and does the therapeutic area precedent support treating it as anything more than exploratory?”

Neurology Leads: Parkinson’s, MS, and DMD Set the Template

Neurology is the therapeutic area where digital biomarkers have come closest to fulfilling the promise the industry has been making since 2018. The reasons are structural. Neurological diseases produce movement signatures that are hard to capture in a clinic visit but easy to capture continuously. Symptoms fluctuate over hours and days. Traditional clinical scales are burdensome, subjective, and coarse. A sensor that records gait or hand motor function at home for a week produces a signal that is genuinely different from what a rater can record at a study visit.

Duchenne Muscular Dystrophy: The First Qualified Digital Primary Endpoint

In July 2023, the EMA issued a Qualification Opinion for stride velocity 95th centile (SV95C) as a primary endpoint in trials of ambulatory patients with Duchenne muscular dystrophy aged 4 years and older.4 SV95C is derived from a magneto-inertial wearable, worn continuously in daily life, and represents the top 5% of stride speeds a patient produces in their natural environment. The measure was developed by Sysnav Healthcare and validated across a multi-year evidence package.

The EMA opinion is a genuine milestone. It is the first digital outcome measure qualified as a primary endpoint for pivotal drug approval anywhere in the world.5 Head-to-head analyses have shown SV95C detects functional decline over shorter intervals than the 6-Minute Walk Test or the North Star Ambulatory Assessment, the standard clinician-administered scales in DMD.6 That sensitivity matters. Shorter time to detection means smaller studies, faster answers, and more efficient use of scarce rare-disease patients.

What is instructive for the rest of the field is how narrow the qualification is. EMA qualified SV95C as an alternative primary endpoint in superiority studies, only when supported by consistent findings on established scales in secondary endpoints. Regulators moved forward, but they did not remove the safety net.

Parkinson’s Disease: Wearables Everywhere, Primary Endpoints Nowhere

Parkinson’s disease is the therapeutic area with the highest volume of active digital biomarker research and the widest gap between exploratory data collection and confirmed clinical utility. A 2025 systematic review of digital outcomes in early Parkinson’s found dozens of measures spanning gait, tremor, bradykinesia, sleep, and speech, but no consensus on which digital biomarkers reflect disease progression well enough to serve as pivotal endpoints.7

The signals of maturity are real. Wearable movement data have shown the ability to identify Parkinson’s up to seven years before clinical diagnosis, opening the door to prodromal-phase trials that are otherwise impossible to run.8 Multiple upcoming disease-modifying platform trials, including the UK EJS ACT-PD platform trial, the French NS-Park Master trial, and Norwegian MAMS trials, are adopting digital endpoints as exploratory outcomes alongside conventional clinical scales.7

The bottleneck is not sensor technology or algorithms. It is agreement on which measures are meaningful. The field has more validated ways to quantify tremor than it has agreement on what quantified tremor tells us about a disease-modifying therapy’s efficacy.

Multiple Sclerosis: Floodlight and the Long Adherence Question

Roche and Genentech’s Floodlight MS program is the most mature smartphone-based digital biomarker program in any therapeutic area. Floodlight administers self-directed tests of cognition, hand motor function, gait, and postural stability, and generates continuous digital outcome measures in people with MS.9 The Floodlight Open study, a global open-access research platform, has now enrolled thousands of participants across dozens of countries.

Adherence has been the recurring question. Recent published data from Floodlight suggest that 70% adherence is achievable in a 24-week study window, with adherence exceeding 90% over shorter three-month intervals.10 Smartphone-based gait assessment through Floodlight has demonstrated sensitivity to change over a three-year follow-up period, an important marker of clinical validity in a slowly progressive disease.

MS also illustrates a challenge that generalizes beyond neurology: the difference between a validated remote assessment (Floodlight qualifies here) and a regulator-accepted primary endpoint (Floodlight is not there). The measures are strong. The regulatory package to convert them into pivotal endpoints has not yet been assembled at the same scale as SV95C in DMD.

Sakara Digital perspective. Neurology has produced the qualification precedents because neurology sponsors were willing to spend the time and money to build a full V3 validation package for a specific, narrow, high-value use case. Every therapeutic area that wants a similar outcome will have to do similar work. There is no shortcut, and no sensor vendor can do it for the sponsor.

Oncology, Cardiovascular, Respiratory, and Mental Health: The Second Wave

Outside neurology, digital biomarker adoption is real but earlier in its arc. Each of these therapeutic areas has produced meaningful pilots. None has produced a qualified endpoint at the EMA-DMD or FDA-DDT level.

Oncology: Slower Adoption, Real Signal

Cancer clinical trials have been slower to embrace digital outcome measures than trials in other areas.11 Part of the reason is structural. Oncology endpoints have historically been survival, tumor response, and clinician-reported adverse events, all of which are hard to replace with sensor data. Another part is cultural. Oncology trialists have deep, hard-won trust in RECIST and OS. Sensors are additive, not substitutive, in most oncology designs.

Where digital biomarkers are landing in oncology in 2026 is around functional decline, chemotherapy tolerability, and remote monitoring of frailty and activity. Wearable-derived physical activity, sleep, and heart rate have shown correlation with treatment tolerability in solid tumor trials, and are being embedded as exploratory endpoints in a growing share of protocols.12 Cancer cachexia is a particularly promising early adoption area, because it is a symptom syndrome where continuous behavioral measurement genuinely adds information the clinic cannot capture.13

Cardiovascular: Physiq’s FDA-Cleared AFib Biomarker

Cardiovascular disease has one of the clearer digital biomarker success stories. Physiq holds multiple FDA 510(k) clearances for AI-based cardiac analytics, including an atrial fibrillation detection biomarker cleared in 2018 and used in the Phase III RESTORE-1 study evaluating orally inhaled flecainide.14 Wearable-transmitted data, run through the FDA-cleared algorithm, allows the trial to confirm AFib symptoms and enroll eligible patients within hours.

It is important to note that these are device-level clearances (510(k)), not drug endpoint qualifications, but they are exactly the class of infrastructure required to make wearables usable as inclusion criteria and safety monitoring tools in pivotal drug studies. Cardiovascular sits closer to genuine adoption than most therapeutic areas because the underlying biosignals (ECG, heart rate variability, arrhythmia detection) already have decades of clinical acceptance.

Respiratory: Connected Spirometry Reaches the Protocol

Bluetooth-connected handheld spirometers now allow high-resolution FEV1 and PEF collection remotely across the course of a respiratory trial. Feasibility studies in both asthma and COPD have demonstrated concordance between clinic and home spirometry measurements, with the potential to increase statistical power and reduce required sample sizes.15 Koneksa’s at-home spirometry, developed in partnership with Regeneron, is an example of a digital biomarker that has moved from feasibility to embedded protocol use.16

Vocal biomarkers are the more speculative branch of respiratory digital measures. A 2025 prospective cohort study is evaluating respiratory-responsive vocal biomarkers for asthma exacerbation monitoring, an area where the underlying signal is genuinely novel.17

Mental Health: Digital Phenotyping Enters Real Studies

Digital phenotyping, the passive collection of behavioral signals from smartphone use, has moved from academic curiosity to embedded feasibility in depression, bipolar disorder, and schizophrenia trials.18 A treatment-resistant depression digital health monitoring study opened enrollment in January 2025 and will run through late 2026, combining electronic data capture with wearables to build behavioral biomarker models of severity change.19

The mental health application illustrates the sharpest form of the validation gap. Sleep duration variance, step count drops, and typing patterns can be measured with high resolution, but converting those signals into a clinically meaningful measure of depressive symptom change requires establishing that the digital signal predicts something a clinician would act on, and doing so in a population diverse enough to generalize. Most published digital phenotyping studies have not yet cleared that bar.

Regulator Acceptance Patterns: What FDA and EMA Are Actually Signing Off On

The pattern of regulator engagement is telling. Both FDA and EMA are actively engaging on digital biomarkers, but they are moving with visible caution and along different tracks.

FDA: Structured Programs, Zero Sensor Endpoint Qualifications

The FDA’s Drug Development Tool (DDT) program is the formal pathway for qualifying biomarkers, clinical outcome assessments, and other tools for use across drug development programs. As of mid-2025, the program has accepted 61 projects, roughly 30% safety biomarkers, 21% diagnostic, and 20% pharmacodynamic response.20 Only eight biomarkers have been fully qualified through the program over its full history, and Letter of Intent and Qualification Plan reviews frequently exceed FDA target timelines by several months.20

For digital measures specifically, the picture is starker. As of late 2025, no sensor-derived digital health technology endpoint has been fully qualified through the DDT program for use in drug development trials.2 The FDA’s ISTAND (Innovative Science and Technology Approaches for New Drugs) program, which handles novel tools that do not fit established DDT paths, has accepted eight submissions to date, including three AI-based tools and Sibel Health’s nocturnal scratch digital endpoint for atopic dermatitis, which remains under review.21

EMA: The First Mover on Digital Primary Endpoints

The EMA has moved somewhat further, most visibly through the SV95C qualification. Between 2013 and 2022, the EMA evaluated a range of digital health technologies proposed for endpoint measurement in clinical trials. Accelerometers were the most commonly proposed sensor, followed by continuous glucose monitors and smartphones, with accelerometers most often proposed for nervous system diseases.22

The EMA’s Regulatory Science Strategy to 2025 committed the agency to establishing a multi-stakeholder platform for digital health technology-derived endpoints, complementing the existing qualification and scientific advice procedures.22 The DEEP project, a proof of concept for navigating EMA qualification of novel digital methodologies, has publicly documented the process gaps and lessons learned that other sponsors can now use.23

Common Ground: Meaningfulness Is the Bar

Both agencies converge on one requirement: digital endpoints must reflect clinically meaningful changes in how participants feel, function, or survive.24 High-resolution measurement is not sufficient. Sensors can detect statistically significant variations that have no meaningful impact on patient quality of life. Regulators have made clear that “over-measurement” of the biologically insignificant is a real risk they are guarding against.25

FDA

Structured, cautious, still catching up

Zero sensor-derived DDT qualifications through 2025. Robust program infrastructure, active engagement, and clear guidance, but a very high bar for full qualification. Individual device 510(k) clearances (Physiq, Sibel, others) are the pragmatic route for many programs.

EMA

First mover, disease-specific

SV95C qualification is the reference point. DEEP project and Regulatory Science Strategy to 2025 signal further openness. Qualification remains extremely narrow: specific measure, specific disease, specific patient population.

MHRA / PMDA

Watching, aligning

Both regulators are engaged in the underlying scientific dialogue but have not yet issued digital endpoint qualifications comparable to the EMA’s SV95C opinion. Sponsors targeting global submissions should assume EMA and FDA precedent will lead the field.

Regulator alignment

Feel, function, survive

All agencies converge on the same test. Digital measures must be shown to reflect changes patients notice and clinicians act on. Measurement precision alone is not the same as clinical meaningfulness, and regulators will push back when the two are conflated.

The Validation Gaps That Keep Digital Endpoints Out of Pivotal Studies

The Digital Medicine Society’s V3 framework (verification, analytical validation, clinical validation) has become the de facto standard for evaluating whether a digital clinical measure is fit for purpose. Since publication in 2020, V3 has been accessed more than 30,000 times, cited in over 250 peer-reviewed papers, and leveraged by more than 140 teams, including NIH, FDA, and EMA.26 In 2025, DiMe extended the framework to V3+ with the addition of a fourth pillar, usability validation, reflecting real-world lessons about scalable deployment.27

The framework is clear. The evidence packages sponsors bring to regulators are often not. The recurring shortfalls we see when reviewing digital biomarker submissions and internal validation documentation cluster around a small number of predictable gaps.

1

Verification packages that skip real-world conditions

Bench testing shows a sensor produces the expected raw output. Real-world verification confirms it does so in the actual environment of use, on the actual patient population, over the intended wear period. Sponsors often provide the former and stop short of the latter.

2

Analytical validation done against reference standards that do not reflect the disease

An algorithm validated in healthy volunteers walking on a treadmill has not been analytically validated for detecting gait changes in early Parkinson’s. Reference standard selection is the most common analytical validation gap, and it is the one regulators catch most consistently.

3

Clinical validation that shows correlation, not meaningfulness

A digital measure that correlates with a clinical scale is not automatically clinically meaningful. Regulators want evidence that the digital measure captures a change patients experience and clinicians act on, in the specific context of the proposed use. Correlation is a starting point, not a conclusion.

4

Usability validation treated as a courtesy

V3+ formalized usability as an equal pillar because deployment failures are the largest single source of digital biomarker dataset attrition. Adherence, device wear, and data completeness need their own validation plan, not a paragraph in the operational section of the protocol.

5

No plan for algorithm change control

Digital biomarker algorithms evolve. Sensors are updated. Cloud platforms are patched. Regulators expect a defensible plan for change control that preserves validation state across the life of the study. Sponsors who treat the algorithm as static discover the problem at inspection.

The failure mode we see most often. A sponsor invests in the sensor and platform, generates a rich exploratory dataset, and then discovers at pre-submission that the validation package was built to demonstrate the technology works, rather than to demonstrate the endpoint means what the sponsor claims it means. The two are not the same. The distinction is the difference between an exploratory endpoint and a pivotal one.

Vendor Landscape: Who Is Instrumenting Real Studies

The vendor market has consolidated meaningfully in the last two years. A relatively small number of vendors now power the majority of pharmaceutical digital biomarker deployments. The distinction that matters is not marketing polish but whether the vendor has actually generated regulator-facing evidence with a pharmaceutical sponsor.

Vendor Focus Notable 2024 to 2026 activity
ActiGraph Research-grade accelerometry; end-to-end regulated sensor data platform Strategic partnership announced with VivoSense to enhance secure access to regulated sensor data for pharmaceutical clinical trials.28
VivoSense Sensor data analytics; biomarker derivation from continuous biosensor streams Named among the market leaders in digital biomarkers globally; core partner for pharmaceutical sponsors deriving new biomarkers from continuous biosensor data.28
Koneksa Health Validated digital biomarker pipeline: spirometry, actigraphy, gait, vital signs Announced clinical pipeline of validated digital biomarkers; partnership with Regeneron on mobile spirometry and with Sanofi on gait biomarkers in neurological disease.16
Clario (formerly ERT + Bioclinica) Precision motion (Opal V2C) plus full clinical trial technology stack Acquired APDM Wearable Technologies. Thermo Fisher announced an $8.88B acquisition of Clario in October 2025, consolidating one of the largest wearable-plus-eCOA platforms.29
Physiq FDA-cleared AI cardiac analytics; continuous ambulatory monitoring platform Six FDA 510(k) clearances for digital biomarkers, including AFib detection; deployed in the Phase III RESTORE-1 AFib study with Syneos Health and InCarda Therapeutics.14
CareEvolution (MyDataHelps) Decentralized clinical trial platform with device and EHR integration Powered the Scripps-led PROGRESS study (Nature Medicine), enrolling more than 1,000 participants in a fully digital diabetes risk study; supports the Long COVID LoCITT trial.30
Withings Health Solutions Consumer-grade hardware validated for research use; sleep, cardiac, weight Sleep Rx received FDA 510(k) clearance for at-home sleep apnea diagnostic support; Withings Sleep Analyzer used in dementia and Alzheimer’s remote monitoring studies.31
Sysnav Healthcare Magneto-inertial wearable; SV95C digital endpoint Developer of the first EMA-qualified digital primary endpoint (SV95C in DMD); reference point for how a full digital endpoint validation package can be built.4
Roche / Genentech Floodlight Smartphone-based MS assessment; largest deployed MS digital biomarker program Floodlight Open enrolled thousands globally; three-year gait sensitivity data published; cost-effectiveness modeling recently published from Austria.9

What we tell sponsors about vendor selection. The correct question is not “which platform is best?” It is “which vendor has run this specific measure in this specific disease with this specific patient population, and can share a defensible validation package?” Very few vendors can. The list of vendors above is the shortlist worth starting from.

An Adoption Maturity Model by Therapeutic Area

Digital biomarker adoption is not a single frontier. It is a set of therapeutic-area-specific frontiers, each moving at its own pace based on regulatory precedent, sponsor investment, and the character of the underlying disease. The maturity model below is a snapshot of where each area sits mid-2026, expressed in the same categories DiMe uses to talk about deployment readiness.

Therapeutic area Maturity stage What is real in 2026
Duchenne muscular dystrophy Qualified primary endpoint SV95C is EMA-qualified as a primary endpoint. Only therapeutic area with a fully accepted digital primary endpoint.
Multiple sclerosis Validated secondary and exploratory endpoint Floodlight is deployed at scale with published sensitivity and adherence data. Not yet qualified as primary endpoint, but well past feasibility.
Parkinson’s disease Widely deployed exploratory endpoint Multiple platform trials incorporate digital measures as exploratory outcomes. Prodromal detection is scientifically compelling but not yet regulatory-ready.
Cardiovascular (arrhythmia, HF) 510(k)-cleared devices in pivotal trials Physiq AFib biomarker cleared and deployed in Phase III. Underlying signal (ECG, HRV) has decades of clinical acceptance.
Respiratory (asthma, COPD) Embedded exploratory endpoint Connected spirometry moved from feasibility to protocol. Vocal biomarkers in early prospective study.
Oncology Exploratory endpoint, symptom-focused Activity, sleep, and heart rate used exploratorily for tolerability and functional decline; cachexia is emerging as an early-adoption pocket.
Mental health Feasibility to early clinical validation Digital phenotyping deployed in feasibility studies for depression and severe mental illness. Clinical meaningfulness case still being built.
Endocrine (diabetes) CGM-native, novel biomarkers emerging Continuous glucose monitoring is decades-established as a regulator-accepted measure. New digital biomarkers layering on top (activity, sleep) still exploratory.
Sleep and neurodegeneration Emerging clinical validation Withings-based studies exploring sleep as a digital biomarker for Alzheimer’s pathology and dementia progression; strong early signal.
Rheumatology, dermatology, GI Feasibility only Isolated proof-of-concept studies. No clear path to regulator-qualified primary endpoint in the near term.

Reading the model. If your therapeutic area is at “qualified primary endpoint” or “validated secondary and exploratory endpoint,” a digital biomarker strategy is table stakes for competitive protocol design. If your area is at “feasibility only,” a digital biomarker is a nice-to-have that will not affect your submission, and the money is better spent shoring up primary and secondary conventional endpoints while you monitor the field.

What Sponsors Should Do in the Next Twelve Months

The mistake we see repeatedly is sponsors treating digital biomarker strategy as a technology decision. It is not. It is a regulatory strategy question, a therapeutic area precedent question, and a validation investment question. Technology is downstream of all three. A practical twelve-month agenda for a mid-sized sponsor looks like this.

Set a therapeutic area strategy first

Map your active pipeline against the maturity model above. In neurology, cardiology, and respiratory, expect digital biomarkers to be a required component of any competitive protocol design. In areas at feasibility only, do not force adoption for the sake of it. The wrong digital measure in the wrong therapeutic area is a source of regulatory friction, not competitive advantage.

Anchor to a specific regulator question

Every digital biomarker program should have a named regulator question it is answering. Are we replacing a clinical scale? Adding a supportive secondary endpoint? Enabling remote inclusion? Reducing sample size? Each of these implies a different validation strategy, and asking regulators to accept the wrong one is where programs stall.

Build validation packages against V3+

Adopt V3+ as the internal framework, not just V3. Usability validation is now expected, and the sponsors that catch this in their study design phase will save six to twelve months at the back end of a program.

Choose vendors on evidence, not marketing

Require every prospective vendor to share the specific validation package for the specific measure in the specific disease. Vendors who cannot are running a general-purpose platform, which may still be the right choice, but should be priced accordingly and not credited with regulatory readiness they have not earned.

Plan for algorithm and platform change control from day one

The vendors your team selects in 2026 will update their firmware, algorithms, and cloud platforms during your study. Your protocol and validation package need to survive those changes. Change control is a structural risk in every digital biomarker program.

Do not skip the pre-submission conversation

Both FDA and EMA offer scientific advice and qualification advice procedures specifically for digital tools. Sponsors that engage early, with a clear question and a defensible evidence plan, get materially better outcomes than sponsors who show up with a completed dataset asking for retrospective validation.

Conclusion

Digital biomarkers in 2026 are past the hype cycle but a long way short of the destination the industry has been promising. A short list of therapeutic areas, led by neurology and rare disease, have produced qualified endpoints, mature validation packages, and defensible regulatory dialogue. The rest of the field is at exploratory endpoint, feasibility, or watching-and-learning. That is not a failure of the technology. It is what maturation looks like when the underlying question, whether a sensor-derived measure is clinically meaningful, is harder than the underlying question of whether the sensor works.

The sponsors moving fastest are the ones that stopped treating digital biomarker adoption as a technology procurement. They treat it as a regulatory strategy, a validation investment, and a therapeutic area precedent question. That reframing is the single most useful piece of advice we can offer teams that are wrestling with where to start.

Sakara Digital works with pharma and biotech organizations building the regulatory, validation, and operational infrastructure that turns digital biomarker ambitions into submission-ready evidence packages. If you are wrestling with a specific therapeutic area strategy or a validation gap and want an independent perspective on where to start, we are happy to have that conversation.