Neurologists often remark that Parkinson’s is a highly variable disease. Persons with Parkinson’s not only acquire the disease at different ages, they also manifest different clinical profiles, displaying symptoms that can progress at strikingly different rates. Persons with Parkinson’s also report divergence in how they respond to treatments such as dopamine replacement therapy. Some are exquisitely sensitive to L-dopa medication, some much less so. Patients also vary widely in their susceptibility to medication-related side effects, from dyskinesias to hallucinations.
So, while Persons with Parkinson’s may superficially look similar, they are in fact presenting with a vast spectrum of disease profiles or “phenotypes.” Bas Bloem, Professor of Neurological Movement Disorders at Radboud University Medical Center in the Netherlands and founder of ParkinsonNet, argues that it is time we tried to unpack this profound puzzle. “Why are some people with Parkinson’s completing crossword puzzles 15 years into the disease and others are demented… Why are some people playing tennis or skiing after 15 years and others, just five years after diagnosis, are unstable or even dependent on a wheelchair?”
Currently, most neurologists intuitively bundle their patients into a handful of clinical subtypes. Patients who initially present with tremor, for example, tend to have a more benign disease course than those who display prominent posture and gait symptoms. People with an older age of disease onset (over 75) tend to become demented more quickly than people with a much younger age of diagnoses. And medication responses like “on/off” fluctuations typically develop more rapidly in young onset patients – those who acquire Parkinson’s under 40 – than those who develop Parkinson’s disease in middle age.
But to Bas Bloem these are very crude subdivisions. In the age of precision medicine, he says, surely we can do better. “So now we have, what is it, five or six Parkinson’s disease phenotypes, whereas in reality there are 5 million Parkinson disease phenotypes: corresponding to the 5 million people worldwide with Parkinson’s, each with their own individual profile.“ Now, while identifying five million phenotypes may be unrealistic, the goal of Bloem and his colleagues is to move towards a much finer granular profiling than we have currently.
Welcome to the Personalized Parkinson Project (PPP), an international collaboration involving Radboud University Medical Center, Radboud University, ParkinsonNet (all of them in The Netherlands) and Verily Life Sciences, an Alphabet spin off and sister company to Google (in the US). The PPP aims to track 650 patients with rather early Parkinson’s disease diagnosis (5 years or less), initially for a period of two years. Researchers plan to periodically measure a plethora of biological and performance metrics. They will take detailed structural and functional brain images. They will sample and analyze spinal fluids, blood serum, plasma, DNA, and stool. Trained assessors will annually conduct detailed clinical exams. And, thanks to the Verily Study Watch, participants will be followed 24/7 outside the clinic as well.
Verily have built a high-tech watch that gathers a variety of physiologic and environmental signals from study subjects including movement, body position, pulse, ECG, galvanic skin response and more. As Dr. William Marks, Head of Clinical Neurology at Verily, puts it, “We designed this study device to be a wrist watch. It’s very lightweight, it’s very nicely designed, and it tells time. So subjects can substitute it for their wristwatch and then just forget about it and go about their daily lives.” With the wearable device, researchers can acquire high quality raw data, monitor changes over time and eventually evaluate responses to therapeutic interventions such as new pharmaceuticals.
The overall project will generate a feast of health data of many different types, but unlocking the insights buried in that so-called “multidimensional” data will require advanced data analysis and machine learning techniques. Says Marks, “Our aim is to ingest large amounts of different data types, to curate that data and then join it together for multidimensional analyses.” So, for example, researchers might apply machine learning to particular data types such as genomic data - in which case they might learn to what extent the known Parkinson’s disease genes account for the variation in symptoms and rates of progression. Researchers might also analyze many other data types, say tau protein levels in cerebrospinal fluid (CSF), to see if levels vary between PD patients and whether those concentrations of tau protein in patients’ CSF help to predict progression.
“But then the really interesting question,” says Marks, “is how do we bring these different data types together. “ Using multivariate analytic techniques, researchers hope to discover particular combinations of biological and performance factors that can account for the variation among Persons with Parkinson’s. Says Bloem half jokingly, “What you would like to do in the end is to say to a patient ‘You’re 73, left handed, with red hair, you live in Michigan, you have two children, you used to be a professional soccer player. After looking at your brain scan and CSF profile, we can say that …. people just like you probably do best if they follow this regime of medication, diet, exercise etc.’”
The data generated in the Personalized Parkinson Project will be securely stored and shared with other projects that are exploring Parkinson’s, including the Michael J Fox Foundation’s Parkinson's Progression Markers Initiative (PPMI). In order to maintain patient confidentiality, Radboud computer scientist Professor Bart Jacobs has developed an advanced data management system called Polymorphic Encryption and Pseudonymisation.
The 13 million Euro project has been made possible by financial contributions from Verily, the Radboud university medical center, the Dutch national government, state government and the City of Nijmegen. The first patients will be enrolled in July 2017; spontaneously, already more than 170 eligible patients have volunteered to participate. The only inclusion criterion for the study is that the patients must have been diagnosed with Parkinson’s for less than 5 years. The rationale is twofold. One: it enables researchers to follow the cohort for a long time (well beyond the initial two-year period). And two: because clinicians believe that Parkinson’s is most amenable to remediation in its early phase.
Bloem and Marks hope that the project will eventually lead to clinical decision support tools that clinicians can use all over the world. As Bloem puts it, “Most Persons with Parkinson’s in the world never get to see a Parkinson’s expert… so my dream is to end up with a decision support system, a black box, that sits on the clinician’s desk. And now he can say, ‘based on your kinetic profile, your CSF result, your stool analysis, your brain scan, your age, your gender, your disease duration, perhaps the color of your hair etc, the best evidence suggests that people like you do well when they get this particular treatment.’ With a better quality of life as a result. That’s my dream."