The art of capturing better patient data: Designing trials around the lives of patients

‘In this age of AI, doesn’t having an actual device make you look historic?’, was the provocative opening question from moderator David Cristina, Principal of Vesalius Biocapital, at his expert panel, on the ‘Role of pharmatech in clinical trial management,’ held at the LSX World Congress Europe in Lisbon.

As Mikesh Udani, co-founder & CEO of Albus Health, riposted, the question is rather, ‘where does AI get its data from?’ If you ask patients to fill out forms and scorecards and other ‘paper-age’ tech, you lose a lot of data. After all, really, who wants to fill out endless forms and diaries when they are feeling top fit, let alone suffering from a chronic disease? And the longer the trial is, the harder it is to rely on patients to consistently fill out the paperwork, meaning that trial sponsors, usually pharma companies, do not get a clear view on how symptoms are resolving, or, worse, progressing. Applying innovative hardware, which in the case of Mikesh’s company, as a sleek box that sits unobtrusively on the bedside table, enables the continuous and objective collection of high-quality clinically meaningful data. Patients don’t notice the hardware as they do not need to do anything different in how they sleep, as the monitoring is invisible, they literally just plug the box in and forget about it. Clinicians appreciate it too, as they capture rich data on vital variables including respiratory rate, cough, heart rate, and more, to inform decision making. Types of data captured can include detecting early warning signs of child asthma attacks up to five days before onset.

At Adalyon, CEO, Ulrik Zeuthen, outlined how his company is developing biomarkers from speech to optimise clinical trials, with a focus on reducing the increasing burden placed on patients. The idea is that speech is such a rich source of information, that if natural conversations are ‘listened’ to more carefully and consistently, it will paint a fuller picture of how the patient is really feeling, not only what blood samples might suggest. It is also not just the words that are used that are important, but the ‘acoustics’ in how words are said. With the aid of AI on top of a mountain of published behavioural research, this novel approach could help predict flare ups in diseases such as irritable bowel disorder or avert mental health breakdowns. By better stratifying patients in this way, clinicians can spend more time speaking with the patients who really need their attention.

Designed around the lives of patients

This can also make ‘patient reported outcomes’ during clinical trials much more insightful. Attrition rates in trials is a big issue for pharma, as these can be as high as 20-40% in a typical phase 2 trial. Clearly, this can have a significant impact on the primary endpoint. And for any trial, not hitting the endpoint is a costly disaster, so why not improve the odds of success? The way to do that is to improve clinical trial design, by designing trials around the lives of patients, to better capture quality patient data.

Putting together a clinical trial dossier itself is very labour intensive, typically taking many months, with protocol amendments once a trial has been approved causing significant delays and pushing up the cost of trials. That hard problem was what inspired Pedro Coelho, CEO of BIORCE, to make things faster, much faster, for phase 1 to 3 clinical trials of therapeutics. Before the AI era it wasn’t possible to generate meaningful text, but now you can. Following a lot of work crunching relevant data – around 1 million real clinical trials – BIORCE is making the distant future of trials rapidly become the present day. In effectively a jump to light speed in dossier production, BIORCE, with ‘Aika’, its AI platform accelerating clinical trials, can now produce a fully-fledged, regulatory compliant, protocol in less time than it takes to eat to drink a coffee and eat a pastel de nata.

Getting the site right & the patient data sorted

An important element of enabling more efficient trials is to start right. Focus on site selection, and the feasibility of specific sites to deliver on patient recruitment given possible changes to inclusion and exclusion criteria, which are at the heart of patient selection and trial success. If a sponsor can access in a compliant way site-related patient data when doing its regulatory planning in designing the trial, that enables the tweaking of these key variables. This helps overcome the intrinsic problem that the person designing the study is not the person selecting the study sites. By helping the clinical team accelerate its trial setup, and reduce risks, sponsors can provide faster access to life-saving therapies for patients.

However AI is not magic, as Pedro observed. LLMs are not great at assimilating numerical data, that’s not what they do, they are language models, plus as we don’t understand how the body works 100%, how could you teach a model if we can’t explain what it is being taught?

Ulrik added, AI is a tool in drug development. If it can be positioned correctly, it can support human work flow. But when inventing a novel approach, you absolutely need to be able to understand how it works, hence Adalyon’s technology is built on behavioural science, with AI being used to pick out the signals – the speech biomarkers – in real time and enabling the company’s vision that ‘speech becomes the new blood.’

Crucially, the data needs to be solid. As Mikesh highlighted, pharma is highly regulated, and needs traceability of every data point, which is not a strength of generative AI. But, as David asked, can there be too much data? As Mikesh put it, the importance is to collect clinically meaningful data, not just more data and reduce the need for patients to visit a site for measurements that could be done at home. Surely it is better to measure someone’s sleep in their own bed, following their own routines, rather than have them go off to a sleep clinic? No one’s going to want to do that for an extended period, but with at home monitoring you can do just that. Ulrik added that more data is not necessarily better, it is about better separating the noise from the signal. If the signal can be strengthened, then better disease management becomes possible. Plus it is not about changing the nature of well-established measurements, such as pain scales, but changing the way the patient answers the scale: this is where disruptive innovation driven by companies such as Adalyon and Albus Health are rapidly moving the needle.

The art of the sell to pharma

With the three panellists all looking to work with pharma, David asked about the art of making sales to pharma. Not all pharma are the same, as Ulrik said: there’s a lot of heterogeneity and it is a very broad landscape. You need to work out who you must convince, and what their objections might be. Some don’t want to be a first mover. Having patient groups and clinicians on board can help. Also, as Mikesh said, if you are not solving a real problem that pharma has, there will be barriers! Pharma has a lot of assessment processes, as the patient comes first and these companies need to minimise risk. Tools need to be validated against the highest clinical standards, and meet all the legal, regulatory and privacy requirements. That takes time, but it is worth it, as pharma are valuable customers that are in it for good. Plus, as Pedro said, be there as a partner, not a service provider, and recognise that any sales are likely to take 12-18 months to emerge as it takes time to build relationships. Finally, it’s a healthy sign that it takes time to do a deal, as you need to build trust particularly for something so novel, added Ulrik.

Finishing up, David asked about the future: what will clinical trials looks like in five years from now? For Pedro, it’s quite simply going to be a revolution, and potentially a ‘one-click’ trial that’s managed by AI-agents. For Mikesh it’s patients having normal lives, with standardised, non-intrusive, data collection. And for Ulrik it is making hospital site visits much more focussed on those patients who really need to be treated, ideally before they have a flare up, or break down, to prevent those debilitating outcomes. And across all three spokespeople it’s about dramatically improving successes for trial sponsors and reducing patient burden. Crack that, and it’ll be a historic change.