Christopher Rudolf
Founder & CEO, Volv Global SA
Léon van Wouwe
Clinical Innovation Director, Volv Global SA
Algorithms can scan through anonymous patient data to help clinicians to discover those who may be at risk of having a rare disease. It could significantly speed up the diagnostic process.
Why can early diagnosis of rare diseases be a challenge?
Christopher Rudolf, Founder and CEO, Volv Global SA: Finding patients with rare diseases is like looking for a tiny constellation among thousands of stars. To put the problem in perspective: there are vastly more data points in a patient record than there are visible stars in a clear night sky. We’re asking doctors to look through all that data to find a disease — or, in our analogy, scan the skies to spot a ‘constellation’ — that they may never have even heard of. That’s not humanly possible. It can only be done with AI technology.
What are the benefits of early diagnosis?
CR: People living with a rare disease are diagnosed seven years too late, on average. Take Fabry disease and Pompe disease. People with Pompe disease can wait up to 12 years before they receive a diagnosis, and Fabry disease patients can wait even longer. Sanofi is working with us to discover patterns for both diseases using complex but accurate algorithms so that patients can receive a much earlier diagnosis.
What is meant by an ‘earlier’ diagnosis?
CR: For example, in the USA, we can use algorithms to find patients with a disease called Alpha-1-antitrypsin deficiency, or AATD, two years earlier than a clinician would typically be able to. In the UK, we think we can find some Fabry and Pompe disease patients even earlier. It’s only possible now due to the amount of data — and I stress this data is anonymous to us — that is available in GPs’ patient health records.
What is the difference to patients and clinicians?
Léon van Wouwe, Clinical Innovation Director, Volv Global SA: When people living with disease get a diagnosis, they can start to make informed choices that reduce the impact that the disease has on them, their families, friends, carers and even work. After all, access to tailored care can only happen after the right diagnosis is made. Also, finding patients earlier helps eliminate some of the costs associated with delay in diagnosis and saves GPs time. This technology is not there to replace clinicians. Clinical reasoning is sovereign. It’s there to support them and make their job easier.
How does the technology work?
LvW: It sifts through data and alerts doctors that there is a patient in their practice who has an elevated risk of rare disease. Those doctors then receive accurate background information about the disease and the standard practice for diagnosis which helps them make better decisions. It’s important to say that the technology is applied to the data — we don’t see the data. We never know who the patients are or even extract patient-identifiable information, so anonymity is always kept. The decision about what to do with the information is up to the doctor.
This technology is not there to replace clinicians. It’s there to support them and make their job easier.
Léon van Wouwe
How accurate is the technology?
LvW: A doctor has validated our model with a review, so we are confident that our performance for Pompe has an 80% precision of plausible people that should be tested for the disease; and for Fabry it’s 88%.* Avoiding getting it wrong must be the priority because doctors and nurses must be able to trust the performance of the technology, and — critically — we don’t want to distress patients with a misdiagnosis. Apart from worrying them unnecessarily, it can put them on a therapy path not tailored to their disease.
Can this technology be used for all rare diseases?
CR: Many rare diseases can be screened for at birth, so it’s not necessary. Some rare diseases are obvious after the onset of symptoms. This technology focuses on diseases that are incredibly difficult to diagnose. The challenge is to constantly improve it and think of other ways to discover all the other disease ‘constellations’ in our night skies.
*Precision is calculated @k
