Biomarkers remain a critical measurement in almost every clinical trial, with this type of data facilitating the accurate characterisation of a drug or medical device’s safety and efficacy.
While traditional blood-based biomarkers have been employed for decades, the biomarker landscape is evolving, with newer variants experiencing a strong surge over the last few years.
Discover B2B Marketing That Performs
Combine business intelligence and editorial excellence to reach engaged professionals across 36 leading media platforms.
One type of biomarker that is garnering attention is the speech-based biomarker, which works by measuring the pace, pitch and clarity of a patient’s voice to provide researchers with data that blood-based biomarkers may not detect.
Speech-based biomarkers can be applied in a broad range of contexts, including everything from screening and tracking disease progression to measuring treatment effects.
According to Ulrik Zeuthen, ex-data and artificial intelligence (AI) director at Novo Nordisk and current CEO of speech-based behavioural analytics biotech, Adalyon, speech-based biomarkers can also facilitate better patient retention. This comes at a time when research estimates that clinical trial dropout rates can exceed 30%, putting economic and operational pressure on sponsors.
Note: This interview has been edited for length and clarity.
US Tariffs are shifting - will you react or anticipate?
Don’t let policy changes catch you off guard. Stay proactive with real-time data and expert analysis.
By GlobalDataAnnabel Kartal-Allen (AKA): What type of insights can speech-based biomarkers provide?
Ulrik Zeuthen (UZ): The interesting thing about speech biomarkers is their broad applicability. Asking what insights they provide is a lot like enquiring about what a blood sample can be used for; they can both play many roles.
Speech-based biomarkers are often used in disease-specific contexts – with a common use centering around the diagnosis of mental health conditions, as well as neurodegenerative indications like Alzheimer’s disease. In this context, they can also be used to predict treatment effects during a trial – helping researchers to determine if a medication is working and if the patient has experienced any improvements before you’d be able to pick that information up through more traditional mechanistic biomarkers.
AKA: Is there an area you see particular promise for speech biomarkers?
UZ: Speech biomarkers can be particularly helpful within the broader general population health domain.
Currently, a lot of operational pain points surrounding clinical trials revolve around patient dropout risk. Though some attrition is seen due to side effects, research has shown that many also leave trials due to practical matters, such as travelling to sites or filling in rigorous questionnaires. As speech-based biomarkers can reduce the necessity for these activities in clinical trials, they can best help with selection, transportation, retention and ongoing monitoring within this setting.
AKA: How do speech-based biomarkers benefit patients enrolling in clinical trials?
UZ: For patients with chronic diseases, many report that – even with a normal blood sample and markers indicating their disease is under control – they still experience health issues, as it can be difficult to measure and categorise wellbeing. This means that there is often a discrepancy between what the researchers and sponsors see and what the patient is experiencing.
Because of this, the industry is making steps to enhance patient centricity in studies. This is something which speech-based biomarkers can help with, as they reduce the need to attend a trial site, fill in a paper or tablet-based questionnaire, instead opting to do this through a web-based solution at home. This means that patients can journal about their experiences in the trial, covering key aspects like when medication was taken and what side effects they were experiencing, while giving the entire context to principal investigators to help evaluate the drug’s impact on a patient’s everyday practical life.
AKA: Are there any privacy and ethical considerations that need to be considered when employing speech-based biomarkers?
UZ: I think it’s highly important to consider quality first in this scenario to avoid falling into the trap of creating a minimum viable platform (MVP) or point of care (POC) system to see if it works, and then worrying about quality, regulations and privacy later. Quality and security by design are crucial, and implementing secure and compliant solutions from the outset is the best way forward.
When dealing with speech biomarkers, we only need to pick up their markers, which already allows for a degree of anonymisation – just like with a blood sample. Researchers do need to be able to see the data, but depending on the intended use and what the patient has consented to, you can easily apply an anonymisation strategy. It’s also possible to synthesise the data for that specific use case.
AKA: Are there any other considerations that may need to be made when integrating speech-based biomarkers into clinical trials?
UZ: It’s important to consider your or your contract research organisation’s (CRO’s) ecosystem, because clinical platforms are at different levels of maturity. For those with modern technology, it would make sense to take this tool and ‘lift & shift’ it into an existing ecosystem. However, some companies are still using paper-based systems, so their approach to integration will differ.
If there’s one thing patients do not want, it’s more tools; many would prefer one, if possible, so that’s what every company should strive for. It’s also important to keep in mind that an implemented tool will reduce the strain on patients who are already feeling the burden of study participation, as the goal is to prevent dropouts.
Speech-based biomarkers fall under behavioural science, which is not just running large language model (LLM) wrappers and then outputting something. This is a domain where companies may have to spend years researching to get it to work, highlighting the importance of picking a reliable provider.
AKA: Do you think a majority of CROs and pharma companies currently have the suitable infrastructure to employ speech-based biomarker technology?
UZ: Any pharma or CRO could use this tomorrow if they wanted to, but they might be more or less dependent on a provided platform – subject to where they are. Sometimes, an app can be applied to a current ecosystem, meaning it just blends in with the existing protocol. In the scenario that they already have a journaling application, you could run a model within their secure infrastructure. If not, speech biomarker technology providers may offer that secure infrastructure to the company.
AKA: What do you think the future of speech biomarkers will look like?
UZ: They mainly prevail in the cognitive domain, as you can use speech biomarkers as an additional signal on top of blood samples and imaging, which does not always give you all the information you need.
While they are used fairly widely in diagnostics, I think we will see speech-based biomarkers used to measure treatment effect, too. They may also gain ground as a holistic tool, meaning they can be harnessed to evaluate the behaviour of a patient and gauge their wellbeing. When this technique is employed, principal investigators can better listen to patients and determine a therapy’s efficacy outside of traditional biomarkers.
I think we will also start to see speech-based biomarkers used in other therapy areas outside of mental health more commonly moving forward.
