Wearables, smartphone sensors, and other sensor-based digital health technologies (sDHTs) can now capture continuous, objective measures of how a patient actually functions between clinic visits — step count, gait speed, sleep architecture, speech patterns, heart rate variability. The methodological question this raises for trial sponsors and research administrators is not whether the data can be collected, but whether a given digital measure is fit to serve as a digital biomarker or, further still, a digital endpoint a regulator will accept as evidence of efficacy or safety. Those are two different bars, and conflating them is the single most common planning mistake sponsors make when budgeting timelines for a digital-measure-based trial.
This guide is scoped specifically to that validation and qualification question: how a raw digital signal becomes a scientifically defensible, regulator-accepted measure. It is a companion to, not a replacement for, CASRAI’s guide on digital twins and synthetic control arms, which covers a related but distinct use of digital/AI methods in trials — generating a synthetic comparator population rather than validating a novel outcome measure. If your question is "can I reduce or replace a control arm with modeled data," see that guide. If your question is "is this sensor-derived measurement good enough to serve as a primary or secondary endpoint," this is the relevant page.
Digital biomarker vs. digital endpoint: the terms are not interchangeable
The field uses these terms precisely, and regulatory reviewers will hold sponsors to the distinction:
- Digital health technology (DHT) — the hardware/software system used to collect data (a wrist-worn accelerometer, a smartphone app, a connected spirometer). See CASRAI’s dictionary entry on wearable devices and digital health technologies for the FDA’s operational definition and the related decentralized-trial context.
- Digital measure — the specific metric a DHT produces (e.g., stride velocity, sedentary time, keystroke dynamics).
- Digital biomarker — a digital measure that has been shown to be a valid, reliable indicator of a biological, physiological, or functional state relevant to a disease or its treatment. Not every digital measure clears this bar; most never leave exploratory use.
- Digital endpoint (or "digitally-derived endpoint") — a digital biomarker (or a clinical outcome assessment collected through a digital modality) that has additionally been established as clinically meaningful enough to serve as a trial outcome measure — the basis on which a regulator will judge efficacy or safety. This is the qualified biomarker or COA sitting inside a specific trial’s endpoint hierarchy, distinct from the broader concept of a clinical outcome assessment.
Put simply: verification and analytical validation can establish a digital biomarker; clinical validation, tied to a specific context of use, is what can elevate it to a digital endpoint. A sponsor can have a well-validated digital biomarker that is nonetheless the wrong choice as a primary endpoint for a given indication because its clinical meaningfulness hasn’t been established for that specific population and outcome.
The V3 framework: the field’s standard validation chain
The most widely adopted validation framework for sensor-based digital measures is V3 (Verification, Analytical Validation, Clinical Validation), published in 2020 in npj Digital Medicine by Goldsack, Coravos, Bakker, Bent, and colleagues, and maintained by the Digital Medicine Society (DiMe). Since publication it has become the reference framework cited by NIH, FDA, and EMA reviewers and has been extended (V3+, 2024) to add a fourth component for usability validation of sensor-based DHTs.
- Verification — does the sensor hardware/software accurately and reliably capture the raw physical signal it claims to capture, under the conditions the trial will actually use it in? This is an engineering question (signal fidelity, sampling rate, device-to-device consistency), not yet a clinical one.
- Analytical validation — does the algorithm that transforms raw sensor output into a derived metric (e.g., raw accelerometry into "steps" or into a gait-speed estimate) do so accurately and precisely, against an accepted reference standard, across the relevant patient population (age, disease severity, device-wear location, skin tone where optical sensors are involved, etc.)? Analytical validation performed in healthy volunteers does not automatically transfer to a target patient population with impaired mobility, tremor, or altered physiology — this is a frequent gap reviewers flag.
- Clinical validation — does the resulting digital measure meaningfully reflect a biological, physiological, functional, or clinical state relevant to the specific context of use? This step is what ties a digital biomarker to clinical meaningfulness and is prerequisite to using it as an endpoint.
- Usability validation (V3+, added 2024) — can the intended user population (which may include elderly, cognitively impaired, or motor-impaired participants) use the DHT as intended, reliably, without differential burden that itself biases the data?
A digital measure that has only cleared verification is not yet a biomarker. One that has cleared verification and analytical validation is an analytically sound digital biomarker but not necessarily clinically meaningful. Only after clinical validation, scoped to a specific context of use, is a sponsor on solid ground calling it a validated digital biomarker fit to inform an endpoint.
Context of use: validation is never generic
A recurring error is treating validation as a property of the device or algorithm alone. Under both DiMe’s V3 framework and FDA’s biomarker qualification framework, validation is always scoped to a context of use (COU): the specific disease, population, measurement conditions, and intended interpretation the biomarker is being validated for. A stride-velocity digital biomarker validated in a Parkinson’s disease population under supervised in-clinic conditions is not automatically valid for unsupervised, free-living use in a Duchenne muscular dystrophy population — even though the underlying sensor and algorithm may be identical. Sponsors planning to reuse a digital measure across indications or populations should expect to re-run some portion of analytical and clinical validation for the new context, not simply cite the original publication.
Regulatory qualification pathways
Once a digital biomarker is validated for a context of use, a sponsor has two broadly different routes to regulatory acceptance, and the choice has real consequences for reusability and timeline:
1. Program-specific acceptance within a single IND/trial
A sponsor can propose a digital biomarker or endpoint as fit-for-purpose within the context of a specific investigational new drug (IND) application, through the ordinary review process (e.g., in a Type B/C meeting, a pre-IND meeting, or as part of a marketing application). If FDA agrees, the measure is accepted for that program only. This is faster to pursue but non-portable: another sponsor developing a different drug for the same or a related condition cannot rely on that prior acceptance and must independently justify the measure to FDA.
2. Formal qualification via FDA’s Biomarker Qualification Program (BQP)
FDA’s Biomarker Qualification Program, established on its current statutory footing by Section 507 of the FD&C Act (added by the 21st Century Cures Act, signed December 13, 2016) and reviewed jointly by CDER and CBER, offers a route to a qualified biomarker: one that, once qualified for a defined context of use, can be used by any sponsor across multiple, unrelated drug development programs without needing to re-establish its validity each time. This portability is the core value of formal qualification and the reason it is worth the additional process investment for a digital biomarker a sponsor expects the field to reuse broadly.
The qualification process runs in three sequential stages with FDA target review timeframes: a Letter of Intent (LOI, target 3 months; the LOI content-element outline was revised in February 2026), a Qualification Plan (QP, target 6 months), and a Full Qualification Package (FQP, target 10 months). The 21st Century Cures Act also requires FDA to publish and update the status of Drug Development Tool (DDT) qualification submissions, including digital-measure-based ones, at least biannually via its public DDT Qualification Project Search database — a useful place for research administrators to check whether a digital biomarker relevant to their program is already in the qualification pipeline before starting a duplicative effort. See CASRAI’s dictionary entry on the FDA Biomarker Qualification Program for the general (not digital-specific) mechanics of this pathway, and CASRAI’s guide to surrogate endpoint validation for how a qualified biomarker can, separately, be evaluated and validated further as a surrogate endpoint for accelerated approval.
The BQP process was designed before sensor-based digital measures were common submission material, and it applies to digital biomarkers without a digital-specific variant — the same LOI/QP/FQP structure and evidentiary bar apply, with the added burden of demonstrating the V3 verification and analytical-validation steps as part of the supporting evidence package.
FDA’s digital health technology guidance and the current regulatory conversation
Separate from biomarker qualification, FDA’s Digital Health Center of Excellence and CDER/CBER review divisions have issued guidance specifically on using DHTs to acquire trial data. The final guidance Digital Health Technologies for Remote Data Acquisition in Clinical Investigations was issued December 22, 2023, finalizing a draft first issued December 23, 2021; it addresses selecting a fit-for-purpose DHT, verification/validation expectations, and data management, but is written at the level of data-acquisition practice rather than providing a digital-specific biomarker qualification track.
Two developments worth tracking, as of this writing (mid-2026), sharpen the open question of how FDA will evaluate the statistical properties of endpoints derived from continuous digital data specifically:
- FDA published a notice in the Federal Register (March 31, 2026) requesting public information and comment on the use of digital health technologies in clinical investigations for drugs and biological products — a step that typically precedes new or revised guidance, though FDA has not committed to a specific guidance timeline as of this writing.
- FDA and the Duke-Margolis Institute for Health Policy are co-hosting a workshop, Digital Health Technologies and Statistical Considerations for Digitally-Derived Endpoints, scheduled for August 27, 2026. Its stated focus is data standards and analytical methods for producing clinically meaningful digitally derived endpoints. As of this writing the workshop has not yet occurred; sponsors and research administrators tracking this space should watch for published proceedings or a workshop summary afterward rather than assume specific conclusions in advance.
Sponsors building a multi-year development program around a digital endpoint should treat this as an active regulatory-science area rather than a settled one: statistical handling of high-frequency, continuous digital data (missingness patterns from device non-wear, aggregation windows, handling of within-person variability) does not yet have the same settled precedent that traditional clinic-visit endpoints do, and FDA’s own public engagement in 2026 signals it is actively working through those questions rather than treating them as resolved.
Practical validation and evidence-planning checklist
- Define the context of use before validating anything. Population, disease stage, measurement setting (clinic vs. free-living), and the specific clinical concept of interest all need to be fixed first — validation performed against an undefined or shifting COU is not reusable evidence.
- Budget for all V3(+) stages, not just analytical validation. Verification and usability validation are frequently under-resourced relative to the algorithm-development work of analytical validation, but gaps here are a common source of reviewer questions.
- Decide early whether you need program-specific acceptance or formal qualification. If the digital biomarker is intrinsic to a single asset’s development, program-specific acceptance through ordinary IND-stage interaction with FDA is usually faster. If the field will plausibly reuse the measure across sponsors and programs (e.g., a mobility or cardiac-rhythm digital biomarker relevant to an entire disease area), formal BQP qualification is the pathway that produces a portable, citable regulatory asset.
- Check the DDT Qualification Project Search database before starting fresh validation work — a relevant digital biomarker may already be in FDA’s qualification pipeline, in which case joining an existing consortium effort (biomarker qualification submissions are frequently led by multi-sponsor consortia, not single companies) may be faster than an independent submission.
- Document the DHT selection rationale against FDA’s remote-data-acquisition guidance even when the digital measure itself isn’t going through formal qualification — reviewers expect a documented fit-for-purpose justification for the device/algorithm pairing regardless of qualification pathway.
- Distinguish your digital-endpoint validation plan from any synthetic-control-arm strategy in the same protocol. They rely on different evidence bases and different FDA review considerations; conflating the two in a briefing document is a common avoidable source of reviewer confusion. See CASRAI’s digital twins guide if a synthetic or externally-derived control arm is also part of the trial design.
Frequently asked questions
Is a digital biomarker automatically a digital endpoint?
No. A digital biomarker is a validated indicator of a biological or functional state. It becomes a digital endpoint only once its clinical meaningfulness has been established for a specific context of use and it is incorporated into a trial’s endpoint hierarchy as the measure on which efficacy or safety will be judged.
Does FDA have a qualification pathway specific to digital biomarkers?
Not a separate one. Digital biomarkers go through the same statutory Biomarker Qualification Program (Section 507 of the FD&C Act) as any other biomarker type, with the V3 verification and analytical-validation evidence folded into the standard LOI/QP/FQP submission package.
What is the difference between analytical validation and clinical validation in the V3 framework?
Analytical validation confirms an algorithm accurately converts raw sensor data into a derived metric against a reference standard. Clinical validation confirms that resulting metric meaningfully reflects the biological, functional, or clinical state relevant to the intended context of use. A digital measure can pass analytical validation and still fail clinical validation for a given population or indication.
Can a digital biomarker validated in one disease population be reused in another?
Not without additional work. Validation under the V3 framework and FDA’s qualification framework is scoped to a specific context of use; reusing a digital measure in a new population, measurement setting, or indication generally requires re-validating at least the analytical and clinical validation steps for that new context.
Where can a sponsor check whether a digital biomarker is already being qualified by another sponsor or consortium?
FDA’s public DDT (Drug Development Tool) Qualification Project Search database, updated at least biannually as required by the 21st Century Cures Act, lists biomarker qualification submissions in progress, including digital-measure-based ones.
