Reacción a "Reactions: movement data can predict Parkinson's risk years before diagnosis"

The work is very interesting and of high quality, and provides very eloquent data on early mobility impairment in Parkinson's disease. It is one example of many that will be completed in the near future in Parkinson's and many other diseases using new data collection tools - such as the mobile phone - and the application of artificial intelligence algorithms.  

What they measure is acceleration in the dominant hand, and the fact that that simple measure ends up being better than more sophisticated genetic or biological measures [in predicting the development of Parkinson's] is very important and makes a lot of sense. On the other hand, the fact that the alteration does not exist in other neurological diseases - except in other parkinsonisms, which is normal - also gives it a lot of value. Moreover, as the authors point out, the method of data collection is easy and inexpensive.  

The sticking point in the interpretation and dissemination of such studies is the classic process of "amplification" between what the authors say, what they themselves suggest, and what is finally communicated. Towards the end of the discussion, the authors very correctly write the following: "In training our model we have prioritised sensitivity over specificity, thus creating a screening tool rather than a substitute for clinical diagnosis". This means that the test gives many false positives, so if the test is positive it certainly increases the risk of developing Parkinson's disease, but being a rare disease (Parkinson's disease is not covid-19) the probability of developing Parkinson's disease is still very low. 

For this reason, the second part of the paper's conclusion is overly optimistic: "In conclusion, our results suggest that accelerometry collected with wearable devices in the general population could be used to identify those at elevated risk for Parkinson's disease on an unprecedented scale and, more importantly, would allow individuals who are likely to develop Parkinson's disease in the coming years to be included in studies for neuroprotective treatments.  

In summary, the work provides a simple, low-cost and very valuable tool for early screening for Parkinson's disease (where a positive test identifies people at higher risk of developing the disease, but it should be clear that they are still much more likely not to develop it). The work does not provide a tool for early diagnosis of Parkinson's disease. This "subtlety" is critical to correctly assessing the scope of the study.