For me, the primary focus when working with clients has been to help them leverage their existing domain and scientific expertise through integrated technology platforms – identifying the pivotal role digital biomarkers play and using them to drive long term success. In this article, I want to place biomarkers in context alongside two other key digital adoption factors. The first is the relationship between AI and domain experience, while the second is the network effects that can ensure long-term data-led market leadership for digital platforms.

A digital biomarker can be any physiological or biological signal that is captured digitally to provide objective measurement of a patient’s condition. They range from general measurements such as glucose levels or blood pressure to something very specific like a genetic marker.

Clinically, these biomarkers can be applied in different ways during the patient care pathway. At the early stage sit susceptibility or risk biomarkers, which indicate the potential for an individual to developing a disease. Cognitive changes in healthy subjects could denote a risk of developing Alzheimer’s, for example. Then there are diagnostic biomarkers that are already in common use to detect the presence of conditions such as asthma. Further up the pathway are pharmacodynamic response biomarkers used to show that a biological response has occurred after a patient has taken a medical product.

Uses for digital biomarkers

Biomarkers, then, are applicable for screening, diagnosis and disease management purposes. At the commercial level, they represent a route to longer lasting business value by enabling data insight to shift from swathes of siloed data to patient- or carer-actionable insights through descriptive, diagnostic, predictive and prescriptive analytics. For digital platforms, this allows the introduction of nudge behaviours and long-term patient engagement and ultimately behaviour changes. Drug companies have the opportunity to add value to their offering through aspects such as improved adherence and efficacy.

Central to all this opportunity is transformative sensing technology, which has moved on apace in recent years by incorporating innovations and learning from the first generation of sensing-based drug delivery devices. It’s now possible to provide continuous monitoring of a wide range of diseases, including diabetes, asthma and neurological conditions. Many inhaler companies have added sensing, initially to measure and monitor the way that devices are being used. But some, like Teva, are taking it further.

Their new inhalers can – in limited clinical settings – collect data from patients that has the potential to predict problems well before they start to feel ill effects. Data from the inhaler can be used to spot minuscule changes in flow of inhalations, and AI algorithms detect the characteristic signs of declining lung function ahead of an asthma attack. During 2021, Teva highlighted study results indicating that exacerbations could be predicted five days in advance. Such foresight will empower both patients and healthcare professionals to take action, offering the potential to alter treatment and avoid serious illness and death.

Keeping with digital biomarkers, the team here at CC instigated an in-house research project called Verum to establish whether they could help in mental health treatment. This is a notoriously under-served and challenging area of medicine where accurate diagnosis and treatment can often fall short. Our project objective was to develop a wearable capable of spotting symptoms of stress. To do this we explored a range of biomarkers and determined which of these had the most independent and predictive information, settling on the three that were most effective and minimally patient impacting.

Analysis of three digital biomarkers

Firstly, the system senses changes in the voice – specifically the correlation between the way certain key phonemes were expressed and the level of stress. Secondly, electromyography measures tell-tale muscle activity while thirdly, a clinical grade wearable was used to track heart rate variability. Digital analysis of the three biomarkers in combination gives an accurate, effective and reliable record of stress levels in a subject.

Another research initiative within CC is focusing on determining combining multiple sources of information and sensing patterns for a particular tissue in the context of surgery. The idea is for the biomarker analysis and data fusion to take place in real time, giving instant feedback to the surgeon. By taking the data and recontextualising it into a familiar setting, it can be used to increase the situational awareness of the surgeon and reduce the mental overhead of integrating these new sources of information, augmenting their skills and experience. The new data techniques are so transformative to the end user experience that their value is readily apparent to all. This has captured the enthusiasm of team members so much that they are continually egging each other on to improve various aspects of the design late into the evening.

Across the innovation landscape, much of the focus right now is on developing and identifying key sensing technologies that can measure accurately enough to be medically useful – but slip seamlessly into peoples’ lives. A promising emerging technology here is integrated photonics, which allows unobtrusive through-skin sensing of biomarkers such as glucose and lactate via a wrist-worn device. A prime example of players in this space is Rockley Photonics, who recently announced partnerships with consumer and healthcare companies for their ‘clinic on the wrist’. My physicist colleague Tom Watson has written an article exploring the extraordinary potential of photonics.

It is of course machine learning and AI that have given us the necessary tools to extract more from data. But I want to make a clear statement here: they are not the universal answer to everything. Domain experience, both from the technology and clinical perspective, are still the essential factors when it comes to making sense of data. Statistical methods, for example, remain the most effective way of determining effects in small studies – and they require expert insight to develop and power them.

Data curation, 90% of which is human effort, can lead to the biggest gains in predictive performance. It’s also true that there is an information limit with all data sources, so it is better to find a specific biomarker than to infer it from lower quality data sources. And at the end of the day, much of it comes down to asking the right question to ensure that insight is useful to the healthcare professional or patient.

Digital platform adoption

I want to widen the focus now to the digital pharma market of the future. The goal of digital platform adoption for connected health is the stability that comes from the network effects, i.e., when external participants contribute and add value. As more people use the platform, the value of its service offering improves. A virtuous circle of positive feedback and user growth ensues.

My message for the ambitious players is that there is a danger of being half-hearted in digital – and that defensive positions are increasingly placing companies at greater risk than offensive ones. Advances in AI now mean that when data, interlinks and interoperability are in place, user-created value could create a ‘Google moment’ suddenly and with surprising consequences. This tipping point could very well create one winner in a medical data market – or at least one Amazon, Strava or Zoom for a number of niche fields.

History from adjacent technology sectors shows us that the first to achieve a platform lead tends to go on to achieve a ‘winner takes all’ long-term dominant position. Being first to market with a product that delivers actionable user value is therefore essential.

Please メール if you’d like to discuss any aspect of this significant topic. We’re currently at a pivotal stage as the foundations of the world’s connected health future finally begin to firm up.

専門家