View complete transcript
And please tell us [00:01:00] a little bit about your work in the Laboratory of Gait and how did you come with this specific question about falls in PD patients? According to the state of Levodopa medication.
[00:01:13] Prof. Vrutangkumar Shah: Yeah. Thank you for the good introduction, Sarah. So, my main area of the work is on applying machine learning and statistical analysis to identify the digital biomarkers of gait and balance in people with Parkinson's disease that we can use for clinical trials. So fall is really a major source of disability in Parkinson's disease, and that leads to hospitalization burden of cost, additional health costs and then also that will lead to reduced quality of life. So I think understanding fall from a balance and gait perspective is very important, especially when we have different drug effects. And it's also multifactorial problem. So according to literature people with Parkinson's disease, they fall almost eight [00:02:00] times more than healthy controls of similar age. So it's people with Parkinson's disease, they are at higher risk of falls. So that's how I started to understand what we can understand from a medication point of view, because most of the people they have off levodopa state on levodopa state. And generally during the daily life, they're mostly on stage depending upon how optimized the drug dose is. So we wanted to investigate what are those effect and then what are the different gait and balance parameter that are significantly contributing to discriminate fallers and non-fallers.
So that was our aim.
[00:02:38] Dr. Sarah Camargos: Perfect. Dr. Shah, why is studying these patients in different scenarios, for example, in the lab or in daily practice?
[00:02:48] Prof. Vrutangkumar Shah: Yeah. So generally in a lab or a clinical setting, we ask subjects to do only like single task or dual task, but it's very focused environment. So in the lab [00:03:00] everything is very focused. It's not like our daily life, which is we walk on the road. We might get in something like a bump or a dog to our way, and then we need to adapt accordingly.
We might be walking over phone and then crossing the road, which is not possible in the lab or even in the clinical setting. So daily life is kind of reflecting how your daily actual working activities is happening. But in lab it's kind of reflecting what is your best performance. So we see that it's kind of having a white-coat effects. So where you will tend to perform better or if somebody's always looking for you or if they want to observe how much or how better you want to perform.
So compared to if we have only normal daily activity, you don't even care about what other people are looking, you just do your normal daily activity, which could be very slower compared to when you do your activity in the lab. So that's the main [00:04:00] difference.
[00:04:00] Dr. Sarah Camargos: And less focused. As you said.
[00:04:02] Prof. Vrutangkumar Shah: Less focused, yeah.
[00:04:04] Dr. Sarah Camargos: And there is maybe more tendency to fall.
[00:04:07] Prof. Vrutangkumar Shah: Yep.
[00:04:08] Dr. Sarah Camargos: So, Dr. Shah, please most of us are not so familiar with these sensors, could you please walk us through the sensors? Are they comfortable how they work?
[00:04:24] Prof. Vrutangkumar Shah: Yeah, so that sensor is very similar to, we have like opal sensor. So that's inertial measurement units, basically. It has accelerometer, gyroscope, and magnetometer. And that is developed by APDM Wearable Technology, which is acquired by Clario now. So it's a part of the Clario company. So it is being used for a lot of clinical trials and more than 500 studies from the academic setting. So when we talk about the daily life we have very similar sensor. It's called instrumented socks, or it's a smart socks, which is similar to your ankle wrap. The [00:05:00] sensor is wrapped around your foot and ankle, and we have a Velcro to adjust the size. So basically, It's the same sensor with the same configuration, but it's a different form of factor.
So it becomes a lot more easier to wear 8 to 10 hours a day. Compared to if we just want to have a sensor on the foot in the clinic. So that is still on the R&D stage. We are planning to put into the market very soon.
[00:05:26] Dr. Sarah Camargos: Very interesting, and I understood that these patients were using these wearable sensors for one week and at least eight hours a day. Right?
[00:05:38] Prof. Vrutangkumar Shah: Yes that's correct.
[00:05:39] Dr. Sarah Camargos: Wow. But did they have a good compliance using this in your research? All of them.
[00:05:45] Prof. Vrutangkumar Shah: Yeah, definitely. This is the third or fourth iteration of the Smart Socks, so we had gradually increased on the compliance. We asked subject to wear for seven days and so far in our data we have seen on and average at least 6 to 6.5 days and at [00:06:00] least 9 to 10 hours of data.
So that's a pretty good of compliance in terms of varying consistency.
[00:06:04] Dr. Sarah Camargos: Yes. Very interesting. And did you predict some falling in some patient, or did you find some patient falling with the sensor? Can you see when a patient falls with the numbers of your sensor?
[00:06:21] Prof. Vrutangkumar Shah: Yeah, so that is still what needs to be done in that area? We haven't figured out just by looking at the sensor data, can we predict the fall because that is a lot of interest in current direction and it might lead to lot more false positives. So we need to look more into the raw data and then see if we get significant kind of a pattern that we can predict. But so far we have worked on, based on the fall history and trying to use some of the measures to predict the future falls.
[00:06:51] Dr. Sarah Camargos: Yes. It's important to mention that patients from both groups, you were comparing patients who fall and [00:07:00] patients who doesn't fall. Both had a similar UPDRS. Both had same age, time of disease and levodopa dosage. Did you thought of studying postural abnormalities in these groups?
[00:07:17] Prof. Vrutangkumar Shah: Yeah, we do have that information. But we did not find that also to be significant between those groups. We had lot of other clinical skills that we used to get those. So even we had the PIGD sub score that was not significantly different between the groups.
[00:07:33] Dr. Sarah Camargos: Tell us a little bit about your results in both lab scenario and in daily life.
[00:07:41] Prof. Vrutangkumar Shah: Yeah, sure. To start with, we had 17 non-fallers and 17 fallers. And we have three scenario. In lab we have two scenario. One is on levodopa and off levodopa stage, and the third stage is the daily life. So when we looked at, in the lab, we find that none of the gait [00:08:00] measures that we looked at was significant in on levodopa state. But in contrast, we find that the turning measures and the gait speed variability measures that were significantly different in off levodopa state. So that was most important because that kind of gives a suggestion that if we want to analyze or even identify the followers, it might be lot more useful if you can evaluate the patient in off levodopa state, and not on levodopa.
So that was the first finding. And the second one from the daily life we do find that turning variability is most discriminative between fallers and non-fallers. So again, turning is what we are finding consistently turning seems to be most abnormal in people with Parkinson's disease.
And also now continuation with the fallers also. So that was our main results.
[00:08:49] Dr. Sarah Camargos: Very interesting. So turning is a very consistent measure which discriminates fallers from non-fallers. And did these [00:09:00] patients have freezing off gait? And if so, did the sensors have the ability to distinguish between freezing and the number of steps in turning.
[00:09:10] Prof. Vrutangkumar Shah: Yeah, so we did add freezing of gait subjects in that also. And we looked at that. That score was not also significantly different between fallers and non-fallers. So we looked at the new freezing of gait questionnaires. Based on that we identify with whether subject is freezers or non freezers. But we did not find a difference between that kind of interacting with the fallers and non-fallers. And with the sensors we use something called freezing of gait ratio from the sensor data. And we have a separate results. And this published now, it's open source code that we use to identify freezers versus non freezers versus non-free, just based on the variable sensor data.
[00:09:48] Dr. Sarah Camargos: So the sensor has the ability to see the freezing of gait of the patient and how this disability can impact [00:10:00] on falls.
[00:10:01] Prof. Vrutangkumar Shah: Yeah, generally we tend to see as the people we have freezing a lot, they have a higher chance of falling because there are almost like the tripping state. But in our dataset, we were not able to clearly figure it out whether that is leading to the falls, basically because we had the similar number of fallers and freezers and non freezers in both of the groups. So it's kind of difficult to figure out when we don't have that information, but I think that's a really important direction to investigate.
[00:10:32] Dr. Sarah Camargos: Perfect. So what are your next steps and how do you plan to translate this research to the clinical practice?
[00:10:41] Prof. Vrutangkumar Shah: Yeah, sure. So I think based on this preliminary results I think the next step is to find out whether this is generalizable enough in a large cohort of the study. So we are collecting a large study with the fallers and non-fallers, almost 6,200 subjects. So next plan is to apply the [00:11:00] same analysis and see if we get the similar results.
If that is a case, then it's a strong suggestion that it might be useful to develop the patient in the off state compared to on state. That might be even useful for a clinician to know that they might get a better understanding about fallers and non-fallers if they evaluate the patient in the off state.
So that is the one. And then next thing that we want to do is kind of build a multimodal for this assessment based on this and prospectively validate on the larger cohort to see how this model is a kind of giving similar results. So in that case we can build the confidence and then suggest that the variable sensors and the digital measures of gait and balance can be used to predict the future fallers.
[00:11:46] Dr. Sarah Camargos: So, in the future maybe we have some very good data about the fallers and non-fallers, and how can we examine this patient in a better scenario [00:12:00] to predict these falls. So as your study points a very good measure should be an off state and especially when turning.
Is that correct?
[00:12:11] Prof. Vrutangkumar Shah: That's correct. Yes.
[00:12:13] Dr. Sarah Camargos: Perfect. Thank you very much for all your time and for your study. And we had learned a lot of you. I hope the listeners read the paper and had some insights about falling and turning.
Thank you. [00:13:00]