Digital Biomarkers in Clinical Trials: The Signal Is the Easy Part

Digital Biomarkers in Clinical Trials: Turning Data Into Decisions

“Data from smartphones, wearables and sensors promises a more continuous, objective picture of patient health. Turning that promise into a measure a trial can actually act on is where the real work begins - and where the technology is rarely the obstacle”, says Agnete Bjerregaard Nielsen, Director of Clinical Consulting at NNIT. 

Clinical development is moving closer to the patient. Trials are more decentralised, more data-driven and far more interested in what happens between site visits. Digital biomarkers in clinical trials sit at the center of that shift. By capturing movement, sleep, gait or heart rate from devices people already wear, they can show how someone actually lives with a condition - not only how they present during a scheduled visit. 

The promise is easy to describe. Delivering on it is where most programmes get stuck. 

"The sensors are good and getting better, so the technology is rarely the bottleneck anymore. The hard part is deciding what you are trying to measure and why, and then proving that the measure holds up under scrutiny," says Agnete Bjerregaard Nielsen. 

Start with the decision, not the device 

More data is not the goal. A digital biomarker earns its place only if it answers a clinical question better than the measures already available and that question has to come first. 

"The most common mistake I see is a team falling in love with a signal before anyone has asked what decision it is supposed to inform", says Agnete Bjerregaard Nielsen and continues: 

"Is the measure there for exploratory insight, for patient stratification, for safety, for adherence, or is it going into a regulatory submission as a digital endpoint? Each of those needs a completely different level of evidence. Skip that conversation and you end up with an impressive dataset that nobody can act on." 

The first step, in other words, is strategic rather than technical. The right measure is chosen because it connects to the disease biology, the patient's experience and the objective of the trial - not because a device happens to produce it. 

Measure what matters to the patient 

The strongest case for digital biomarkers is that they capture what a clinic visit misses. In neurological disorders, wearables in clinical trials can track gait, tremor or mobility as they change through an ordinary day. In sleep conditions, they add context on patterns and variability over weeks rather than a single night in a lab. In chronic disease, near-continuous monitoring surfaces fluctuations that scheduled assessments never see. But volume is not relevance. 

"A wearable will happily generate millions of data points. That is not the same as measuring something the patient would recognize as meaningful", says Agnete Bjerregaard Nielsen. If a number moves, the clinical team has to be able to say whether that movement matters in daily life, and whether it matters for the therapeutic area.  

Getting there means bringing clinical, scientific, regulatory and patient-facing perspectives together from the start, rather than validating a measure after the fact and hoping it means something. It is one of the questions we will be discussing with industry experts from Novartis and Boehringer Ingelheim in our upcoming webinar.

Data you can stand behind 

Digital biomarkers live or die on data quality, and this is usually where implementation gets hard. Device placement, battery life, connectivity, firmware updates, missing data and mid-study algorithm changes all move the signal around. On top of that, a trial needs provenance, auditability, privacy controls and a clean path into the wider clinical data flow. That is a design problem, not a procurement one. Can patients use the device correctly for the full period? Can sites onboard and troubleshoot without drowning? Is the pipeline validated, secure and inspection-ready? Those questions decide whether a measure survives contact with a real study. 

Digital biomarker validation is the throughline. Analytical validation, clinical validation and regulatory expectations belong in the protocol from day one, not bolted on before a submission - the same discipline that underpins decentralised clinical trials and the data integrity regulators expect from them. 

AI helps - and raises the bar for governance 

AI is increasingly part of how digital health technologies in clinical trials are built. Machine learning can find patterns in high-volume sensor data, support endpoint discovery and sharpen how longitudinal signals are interpreted. What it cannot do is remove the need for validation. If anything, it raises the stakes. 

"The interesting question is never whether a model can find a signal. It always can. The question is whether that signal is clinically meaningful, reproducible, and defensible for the decision you want to make", says Agnete Bjerregaard Nielsen and continues: "The moment you bring a model in, training data, bias, version control and change management all become part of your evidence package." 

Build it, or partner for it? 

Pharma is increasingly looking beyond its own walls. Collaborations between drug developers and consumer technology companies reflect a broader rethink of where value comes from. The ambition is often wider than data. For many pharma companies, the question is no longer only how to develop and deliver a medicine. It is also how to support the patient journey around that medicine. That may include symptom tracking, adherence support, behavioural interventions, patient education, lifestyle support, remote monitoring, and better feedback loops between patients and healthcare professionals. 

This points toward a wider care ecosystem, where digital tools complement the therapeutic product and help generate a more complete understanding of the patient experience. In some cases, the digital component may help collect relevant data in daily life. In others, it may help patients manage their condition more actively or help clinicians intervene earlier. Often, both ambitions are closely connected. 

Each side brings something the other cannot easily build. Pharma has therapeutic depth, clinical development experience and regulatory understanding. Technology partners bring user experience, device ecosystems and faster innovation cycles. 

"Some organizations will build these capabilities in-house, some will partner, and most will land somewhere in between. What matters is keeping strategic ownership internal while being honest about what you genuinely need to source outside," argues Agnete Bjerregaard Nielsen. She concludes: "And do not treat choosing a partner as a purchasing exercise. The device is the easy part to evaluate. Whether a partner can support validation, data governance and regulatory scrutiny over years is what actually decides success." 

From pilot to practice 

Plenty of teams have run a digital biomarker pilot. Far fewer have embedded these measures across a development programme. The difference is rarely the technology - it is the discipline around it: a clear clinical purpose, a scientifically justified measure, validation planned early, a data foundation you can defend, and partners chosen for the long haul. 

That is the shift now underway: from experimenting with digital biomarkers to depending on them. NNIT and industry experts from Novartis and Boehringer Ingelheim are exploring exactly this in an upcoming webinar - from proving strategic value to integration, validation and the build-or-partner decision.  

Sign up to join us, or register to receive the recording.