Recently, I had the great pleasure of sitting down with neurologist and internationally recognized movement disorders specialist Dr. John Caviness to get his take on the use of vocal biomarkers in the early detection and progression of Parkinson’s disease and other neurodegenerative disorders. Dr. Caviness has played a key role in the research and treatment of such disorders for over 25 years, and is an esteemed member of our scientific advisory board. The opinions expressed below are his own and are not official positions of the Mayo Clinic.
As a neurologist, what is your role in managing the condition of your Parkinson’s patients?
Most of my practice is centered around Parkinson’s Disease. My role is to evaluate patients to either confirm, question, or deny their diagnosis. I then render an opinion as to whether or not they’re on the correct symptomatic treatment. We don’t currently have treatment of the underlying disease, so it’s about managing symptomatic treatment. Since Parkinson’s is a progressive disease, they are seen periodically as the patient’s medication requirements naturally change.
Can you walk me through the experience of a Parkinson’s disease patient, following diagnosis?
Sure. So, let’s say they come in with a tremor or they’ve noticed that their motions are becoming slower, or they are displaying less facial expression. During the initial visit, I confirm the diagnosis of Parkinson’s disease, and only if their disability justifies it, I suggest a trial of Levodopa therapy or some other intervention. Levodopa is a precursor to dopamine in the brain. It works to improve dopamine function in the motor system predominantly and can give them approximately >50% symptomatic relief. However, I have to tell the patient that it’s not a cure, that the medication is going to help with their motor symptoms, and they’ll need to adopt a healthy lifestyle with good exercise, good nutrition, and a good mental attitude.
Often, the medicine has a good response and the patient is impressed. I am in tertiary care, acting as a consultant to the patient and will see them periodically. Over time, the patient will notice the medication isn’t working as well as it once did and we’ll start supplementary medication. We may start other medications to try to even out the variability of the symptoms as the levodopa kicks in and subsequently wears off. Patients report this variability as best they can, along with their families or caretakers. We do our best to even out the ups and downs while working to minimize supplementary medication and the variability that the patient is experiencing.
Longer term, say 5-10 years out, patients are on a good amount of longer duration therapy in addition to the initial treatments. It’s common for patients to also take their medications more often than what they’re initially prescribed and they’re having fairly significant ups and downs to their medication effect to the point where it affects their quality of life.
Another alternative is deep brain stimulation surgery, which can significantly minimize the fluctuation with levodopa therapy, when supplementary medication is not effective enough. Over longer periods (10+ years) there are limits on what any intervention can do. Unfortunately, the patients just live with the limitations of what the medication and surgery can do. But importantly I would tell you Daniel, what the medication can do is also limited by the quality of information that the physician receives. Here I see a lot of opportunity and need for new biomarkers, especially digital tools to better empower physicians and patients.
You mentioned the quality of information can limit the effectiveness of the medication. That’s interesting, can we talk more about that?
First of all, every patient is different. I’ve told you about an average patient of Parkinson’s disease, but there really is no such thing. Everybody is different. Further, I can only see them and examine them at a single point in time. They are varying throughout the day. Getting better information about what’s actually happening objectively can be difficult given that variability. It really affects how well you can titrate their medications and dial in their care. It would be greatly beneficial to have more data – objective, clinically relevant data – to review along with the interaction with the patients directly at a point in time. This type of data could lessen doctor visits and drive down healthcare costs.
Have there been any significant developments in the treatment of Parkinson’s disease since over the last few decades?
Unfortunately, not really in terms of preventing the progression of this horrible disease.
I think the biology has been tough. We’ve gone after some things. We’ve been optimistic about data shown in animal models and we’ve misled ourselves in thinking what happens in the test tube, so to speak, has application to human beings. Thus, we’ve been testing treatments that really don’t get to the basic biology of Parkinson’s disease to the degree needed. However, I am very optimistic that this situation is changing significantly. People with the genetic causes of Parkinson’s disease getting genetic diagnosis based therapies in clinical trials is an example of a potentially good approach. This is just one example of several exciting new approaches.
An important goal of our speech analytics technology is to improve the outcomes of patients. As an expert in movement disorders, what do you see as the lowest hanging fruit for our technology to make a positive impact on the patient population you serve?
Your tool is what I would regard as digital medicine; interacting with the patient in a digital way such as to gain an understanding as to what their current level of symptoms are and/or how well the medicine is or is not doing for them. The patient can do that themselves today, subjectively by coming into the clinic to tell me about how they are feeling and the variable response that they are seeing with their medication. That is purely subjective and months can go by between visits. Your tool gives us and the patient a much more frequent and objective data set that is valuable.
So, if I had a way to gather a sampling of objective data over the six months in between visits, that would be very useful. It could be for confirmation that the patient is doing great, or it could tell a story that the patient, over the last couple of months, has been doing not as well as they should. It could tell a story that the patient is doing well during some hours, and not as well in others. Understandably, patients and families not being in the medical field can struggle with quantifying that or even evaluating that very well.
How do you believe technology like Aural Analytics’ speech analytics platform will impact the diagnosis, differential diagnosis, longitudinal tracking, or treatment of Parkinson’s disease?
I think there’s a lot of important potential. Speech is known to change with Parkinson’s disease for a lot of patients and a tool like this could be used in several different ways, via as I put earlier, digital medicine. For instance, you might find big changes are happening to a patient say 30 minutes before or 30 minutes after taking their medication. I would be a lot more confident in not only what is occurring with that patient, but also what to try to help mitigate the medication response variability that we discussed earlier.
We view clinical validation and interpretability as paramount. Can you talk a little about your viewpoint on new digital tools?
For those in clinical and academic medicine, we’re going to want to see that any digital tool correlates with a physician’s objective observation as well as correlates with the patient’s symptoms and quality of life. If you can show that, then you have the analytics tool acting an extension of the clinical examination. That is tremendously useful.
I can think of several “blue sky” ideas assuming clinical validation has been met. Beyond the concept I discussed above which I termed digital medicine, I could easily see this extending to what we define as “personalized medicine” whereby patients take charge of when they take medicine, with periodic consultation with a physician.
You mentioned there are some successful treatments for addressing the symptoms of Parkinson’s disease but not the underlying causes. Do you see this changing and is there a role for speech analytics in these kinds of studies/trials?
In the sense of a digital biomarker and more specifically a vocal biomarker, definitely yes. This is especially true in terms of utility in clinical trials and studies. In addition to addressing the basic biology of Parkinson’s Disease as we discussed before, the studies could be testing the treatments better with your tool. Thus, the second part of the story is; do we really have the objective measures that we need which are sensitive and precise enough to pick up real changes that reflect an improvement in the biology? Overall, it feels like we don’t have the right tools for that. I believe speech data and your platform could be an important tool to address that problem. Your tool could also work well, I believe, with other assessments for this purpose.
About Dr. Caviness
John N. Caviness, M.D., is a board-certified neurologist at Mayo Clinic in Phoenix, AZ. His clinical practice includes the diagnosis and treatment of movement disorders such as Parkinson’s disease. Dr. Caviness has played a key role in the research and treatment of such disorders for over 25 years and is an esteemed member of our scientific advisory board.