For decades, medicine has answered the question "did this drug work?" through scheduled snapshots of disease: a scan every six months, a lab result every quarter, a structured questionnaire at each clinic visit. These assessments are validated, standardized, and trusted, and they will remain foundational. Yet they share a structural constraint: they capture disease episodically, while disease itself unfolds continuously.
The promise of digital biomarkers as a solution to this gap has been discussed for over a decade, first formally defined in 2015. What has shifted is not the concept of digital biomarkers, but the infrastructure beneath it. With maturing regulatory frameworks, stronger validation science, and operational governance models, the field is moving from technological possibility toward system-level readiness, even if that transition is still unfolding.
What Are Digital Biomarkers: From Promise to Proof
Digital biomarkers are objective, quantifiable physiological and behavioral data collected through digital devices. Enabled by wearable sensors, smartphones, and connected technologies, they capture continuous signals in natural settings; physiological measures such as heart rate or glucose levels, behavioral patterns such as gait or sleep quality, and environmental factors that influence health outcomes.
The value in drug development extends beyond operational efficiencies. Digital biomarkers are proving capable of detecting subtle treatment effects and early disease progression, particularly in slowly evolving conditions such as neurodegenerative diseases, that traditional endpoints may miss. They enable deeper phenotyping to identify earlier signals of efficacy, distinguish likely responders, and refine target populations. In some cases, this sensitivity has translated into materially smaller trials, reportedly reducing required enrollment by 50–70% and accelerating timelines.
But technological capability alone does not translate into regulatory acceptance, particularly in a system built around historically discrete endpoint categories.
Regulatory Navigation: Where Acceptance Meets Categorization
Let’s first look at digital health technology as a whole. Both the FDA and the EMA have established pathways for digital health technologies in drug development. Guidance on verification, validation, usability, and risk management now exists where it did not before. Precedent is accumulating.
The clearest milestone to date is the EMA's qualification of stride velocity as a primary endpoint in studies of ambulatory Duchenne muscular dystrophy — a rigorous, precedent-setting decision demonstrating that a digitally derived measure can meet the evidentiary standards required for a drug trial's main outcome. The same measure is under FDA review, and a previous FDA agreement to use wearable-derived moderate-to-vigorous physical activity as the primary endpoint in a Phase 3 idiopathic pulmonary fibrosis study (although the trial did not achieve statistical significance) is significant. However, in both instances, the measures were categorized as functional outcomes rather than biological surrogates.
This pattern reflects a fundamental tension; digital measures are gaining acceptance primarily as functional outcome assessments — tools capturing how patients feel or move — while qualification as biomarkers reflecting underlying biological processes remains far rarer. This distinction has real consequences. Biomarkers can serve as surrogate endpoints, enabling accelerated drug approval in serious diseases where definitive outcomes take years to observe. That pathway, compressing development timelines, potentially accelerating patient access, remains largely unlocked for digital measures.
Understanding why requires a brief detour into regulatory mechanics. Before any digital measure can enter a development program credibly, sponsors must define two things: the context of use (specifically, how the measure will inform decision-making in a trial, whether as a primary endpoint, an exploratory signal, or a safety indicator) and the concept of interest, meaning the underlying clinical characteristic the measure is actually trying to capture. These definitions are not administrative formalities. They determine which evidentiary pathway the measure must travel, what validation is required, and ultimately what claims can be made.
The challenge is that digital measures frequently make both definitions genuinely ambiguous. Take gait speed in Parkinson's disease, continuously captured via a wearable accelerometer. Define the concept of interest as neurodegeneration — the progressive biological process underlying the disease — and the measure looks like a biomarker, requiring a different and more demanding validation pathway. Define it as functional capacity ( the patient's ability to walk and move — and it looks like a clinical outcome assessment, more accessible but differently constrained in terms of the claims it can support. The signal is the same. The regulatory journey diverges based on how the question is framed.
Regulators have generally resolved this ambiguity by categorizing such measures as functional outcomes — the more accessible pathway, but the lower-impact one in terms of what can ultimately be claimed about biological mechanisms or surrogate validity. Global inconsistency compounds this: the Mobilise-D consortium's digital mobility outcomes were evaluated under the EMA's biomarker qualification pathway, while similar measures are discussed as clinical outcome assessments at the FDA. The underlying science is the same. The categorization differs. Whether this reflects a principled distinction or a structural artifact of different frameworks is a question the field has not fully resolved.
The Surrogate Endpoint Opportunity
Perhaps the highest-value application of digital biomarkers would be as qualified surrogate endpoints; measures that stand in for definitive clinical outcomes in accelerated approval decisions. The evidentiary bar is high: sponsors must demonstrate a durable link between the digital measure and meaningful outcomes across populations and disease contexts. But the potential return is proportionate. Compressing a ten-year development program by even a few years carries enormous clinical and commercial value.
Digital biomarkers appear well-suited for this role. Their continuous measurement captures disease progression with greater granularity than episodic assessments. Their sensitivity to early biological changes could detect treatment effects months or years before clinical outcomes manifest. These characteristics align precisely with what surrogate endpoints need: reliable, sensitive measures that predict meaningful clinical benefit.
Yet achieving surrogate qualification requires crossing a high evidentiary bar. Sponsors must demonstrate a mechanistic or epidemiologic rationale linking the digital measure to meaningful outcomes and provide robust clinical data showing predictive validity. Even then, qualification applies narrowly to specific diseases, populations, and often particular mechanisms of action.
The opportunity is substantial—materially shorter development timelines, smaller trials, earlier patient access—but it requires disciplined investment in the right evidence at the right time. The current regulatory acceptance pattern suggests the field hasn't yet cracked this code at scale.
Navigating Technology Evolution
Complicating these categorization challenges is a structural tension: digital products evolve on two- to three-year lifecycles, while drug development spans a decade. A measure validated at trial initiation may be technically obsolete by regulatory submission. This mismatch initially appeared insurmountable but is now driving pragmatic governance innovation.
Predetermined change control plans allow controlled algorithm updates under defined oversight. The FDA's discussions on AI and machine learning specifically address data quality, bias, and transparency risks unique to adaptive models. Sponsors now implement version control, performance monitoring, and drift detection as standard practice.
More fundamentally, standardization efforts through consortia are shifting how measures are defined. Sponsors increasingly validate "daily step count derived via triaxial accelerometry using algorithm X" rather than "this specific wearable device." This measure-specific approach is foundational for scalability, enabling validation evidence to persist even as hardware iterates.
Fit-for-Purpose Validation: Flexibility With Rigor
Regulatory thinking has converged around fit-for-purpose validation. A digital measure must be validated to the level appropriate for its intended use. An exploratory Phase II endpoint requires different evidentiary depth than a primary endpoint in a pivotal trial.
This tiered framework has lowered the barrier to early adoption without compromising rigor. Sponsors are increasingly using exploratory digital endpoints—for mechanistic insight, faster go/no-go decisions, patient stratification, and composite endpoint prototyping—to build validation evidence in parallel with drug development. The FDA's 2023 guidance clarified expectations around verification, validation, usability, and risk management.
Where Innovation Continues
While regulatory frameworks mature, the technology itself continues advancing. Platform companies like Koneksa, Empatica, Evidation, and ActiGraph are evolving toward validated measurement ecosystems, publishing methodologies, and building evidence packages that multiple sponsors can leverage.
Sensor fusion technologies are moving from single-parameter tracking to integrated profiling. Cognitive and mental health applications are expanding through digital phenotyping. Therapeutic momentum is concentrating, where digital measures address clear gaps—neurodegenerative diseases, rare neuromuscular conditions, psychiatric disorders, and pulmonary diseases.
An Ecosystem Aligning
Digital biomarkers require coordination across stakeholders that historically operated independently. Pharmaceutical companies seek predictability. Technology firms iterate rapidly. Academic groups generate foundational evidence. Regulators refine guidance. Payers determine whether endpoints translate into reimbursable value.
Alignment is imperfect but advancing; precompetitive collaborations, like the Critical Path Institute's efforts and the Mobilise-D consortium, are distributing validation costs; regulatory precedents are creating templates that reduce uncertainty; remote patient monitoring reimbursement codes signal payer engagement is evolving from skepticism toward framework development.
Ultimately, digital biomarkers are no longer speculative tools; they're increasingly integrated into early-phase trials, exploratory endpoints, and, in select cases, pivotal development strategies. But the path from today's selective adoption to routine deployment remains unclear. Stakeholders need to harmonize divergent global regulatory expectations, standardize validation approaches while technology continues evolving, and build frameworks that accommodate measures bridging biological processes and functional outcomes, categories that the current system wasn't built to handle. The infrastructure is catching up, but how quickly, and through what mechanisms, remains the central challenge.