Why look for Parkinson's before diagnosis?
Parkinson's disease is a disease of the nervous system that affects movement and gets worse over time. It has a long "latent phase": brain cells are damaged for years before a diagnosis is made. Today there is no treatment able to change the course of the disease.
To test future treatments that would protect the brain, they would need to be given as early as possible. But reliable signals, called biomarkers, to identify these people in advance are lacking.
What did the researchers do?
The team used the UK Biobank, a huge British health database. They studied accelerometry data, meaning movement measured by a wearable device, in people from the general population.
Machine-learning models (programs that learn to recognise patterns from many examples) were trained to tell apart people with diagnosed Parkinson's (153 people), people diagnosed later (113 people, up to 7 years after the measurement) and the rest of the population (33,009 people). These models were then compared with others based on genetics, lifestyle, blood biochemistry or early signs.
What did the study show?
Models based on accelerometry did better than every other type of data tested, both for people already diagnosed and for those not yet diagnosed, up to 7 years before diagnosis.
The score used, AUPRC, measures whether the people flagged are truly cases when the disease is rare. It runs from 0 to 1. For people not yet diagnosed it is 0.07, against 0.01 for genetics, 0.03 for lifestyle, 0.01 for blood biochemistry and 0.01 for early signs.
Key figures
- 7 yearsbefore diagnosis: the longest span over which the models flagged cases not yet diagnosed
- 33,009people from the general population used as comparison
- 0.07AUPRC score for cases not yet diagnosed, against 0.01 to 0.03 for the other data types
What does this change for patients?
Nothing concrete for a patient yet. The authors see accelerometry as a potential low-cost screening tool, to spot people at risk and to pick participants for trials of neuroprotective treatments.
Such a tool would not say that someone is ill, but that closer follow-up or joining a trial might be worthwhile.
What are the limits?
A score of 0.07 is low in absolute terms: many of the people flagged would never develop Parkinson's. Used alone, the tool would not be a reliable test. It simply does better than the other sources tested. Because Parkinson's is rare, even a better score can still mean many false alarms among thousands of people, which is why the absolute value matters.
The study relies on data already collected in a cohort. It does not show that spotting these people earlier improves their lives, nor that the method will work in other populations or with other devices. That has to be checked.
Parkinson's and daily life: a helping hand from Mima
This study is research: no watch today can tell you that you have Parkinson's disease. It shows that everyday movements could one day give valuable clues.
If you live with tremors or stiffness, or support someone who does, Mima helps you describe what is changing and keep a record for the neurologist. A question to ask: "Which signs should I report to you straight away?"
Frequently asked questions
Can my smartwatch detect Parkinson's disease?
Not today. The study analyses research-grade accelerometer data and shows a lead for screening, but performance is low and it is not a diagnostic test.
Why detect the disease so early?
Because the disease develops for a long time without obvious symptoms and future protective treatments would need to be tested as early as possible. That requires finding people at risk.
What is the UK Biobank?
It is a large British health database holding information on a very large number of participants, used by researchers around the world.
Is this tool better than blood tests or genetics?
In this study, yes: accelerometry performed better than genetics, lifestyle, blood biochemistry and early signs, according to the study published in Nature Medicine in 2023.
The original study
Schalkamp AK, Peall KJ, Harrison NA, Sandor C. “Wearable movement-tracking data identify Parkinson's disease years before clinical diagnosis.” Nature Medicine, 3 July 2023. DOI : 10.1038/s41591-023-02440-2
Read the full study in Nature MedicineThis article summarises a scientific study and is not medical advice.