Machine Learning to Help Predict PD Progression?
Doctors and patients alike generally agree that everyone with Parkinson's Disease (PD) has their own PD journey, experiencing many different motor and nonmotor symptoms, a wide range of ages when they experience their first PD symptom, the rate at which their symptoms worsen, and how they respond to treatment. This notorious diversity makes the attending physician’s job all the more difficult: with no objective lab test or x-ray to point to a particular therapy for each individual patient, much of the therapy decision-making needs to be done on a trial-and-error basis. Naturally, this can also be highly frustrating for patients. Also, on the clinical research side, since it is impossible to identify clearly what “type” of PD a trial participant has or how quickly it may progress, trials need to include many patients and last long, making the development of new therapies much more expensive.
Until now, doctors would assign patients to a PD subtype based on the age at onset (early versus late onset), the rate at which their disease progresses (slowly progressing versus fast progressing), and clinical symptoms (with and without dementia, tremor-dominant versus postural instability with gait disorder). But this is little more than a very rough classification system, since each of these categories represents a spectrum without any clear limits. Such a rough subtyping method doesn’t give doctors, patients, and researchers much insight to work with.
To solve this problem, modern approaches to data analytics are now being explored. Machine learning and artificial intelligence have become buzz words across many industries; nonetheless, recently an international group of researchers published a paper1 with some very promising results. By looking at the progression of people's motor (movement) symptoms, sleep functions, and cognitive changes over a 5-year period, they found 3 clusters of disease progression: slower progressing, moderate progressing, and fast progressing. They then showed that they could predict people's rate of progression based only on their symptoms at disease onset with great accuracy – specifically, with 92%, 87%, and 95% accuracy for the slow-, moderate-, and fast-progressors, respectively.
But what does this mean for individual patients and their treating physicians? Interestingly, by looking at each item of the clinical scales separately, some intriguing insights were found. For example, it may be possible to help predict how a specific patient’s disease may progress by looking at specific symptoms at baseline (meaning when they are diagnosed with PD). Daytime sleepiness was found to be the strongest predictor, followed by doing hobbies and activities, getting dressed, and urinary problems.
The authors of these papers also tried to identify additional predictors of how a specific patient’s PD may progress over time. Given the great interest in genetics, patient characteristics (such as height, weight, blood pressure, and demographic background), and blood and urine tests to try and sub-type Parkinson's, the authors tried to look for a pattern here too. They found a significant but weak predictive effect of genetic biomarkers. Also, a blood test showing high levels of neurofilament light (NfL) at diagnosis, as well as rapid increases of Nfl, predicted a more rapid progression
Further work in these directions may help doctors and patients in making personalized treatment plans. It may also help in addressing specific symptoms – for example, it may be possible to identify in advance patients at high risk of falls, cognitive decline, or rapid progression, and possibly offer
Source: Dadu, A., Satone, V., Kaur, R. et al. Identification and prediction of Parkinson’s disease subtypes and progression using machine learning in two cohorts. npj Parkinsons Dis. 8, 172 (2022). https://doi.org/10.1038/s41531-022-00439-z https://www.nature.com/articles/s41531-022-00439-z#citeas/
THIS ARTICLE DOES NOT PROVIDE MEDICAL ADVICE and is not a substitute for professional medical advice, diagnosis or treatment. If you or any other person has a medical concern, you should consult with your health care provider or seek other professional medical treatment immediately. Never disregard professional medical advice or delay in seeking it because of something that you have read on this website or in any linked article, blog or other materials.