Digital Biomarkersfor Huntington Disease

Friday, November 411:15am-12:15pm

Chair: Ralf Reilmann, MDGeorge Huntington Institut

Presenters

HSG 2016: DISCOVERING OUR FUTURE

Max Little, PhD (virtual attendee)Aston University

Spyros Papapetropoulos, MD, PhDTeva Pharmaceuticals

Gaurav Sharma, PhDUniversity of Rochester

Objective measurement of HDsymptoms using smartphonesDr Max Little (maxl@mit.edu) Research Director, NumericAnalysis LtdAssociate Professor, Aston University, UK Senior Research Fellow, Oxford University, UK Visiting Associate Professor, MIT, US

Smartphones as serious toolsfor symptom measurementKey aims:• Reducing logistical difficulties for measurement of HD

symptoms• Improve objectivity (repeatability, reliability) of testing

methodology• Enable high-frequency measurement• Improve quality and frequency of follow-up measurements

in clinical trials

Structured smartphone tests:hardware and protocol

Raw sensor data collected using specialised Android smartphone softwareUsers performed specific test protocols in clinic:• Gait, balance (accelerometry)• Tapping, reaction time (touchscreen)• Voice (microphone)

Accelerometry pre-processing • Smartphone orientation

identification (top left)• Orientation signal in

spherical coordinates (top right)

• Impulsive events extracted from dynamic acceleration (bottom left)

• Residual signal (bottom right)

Objective-HD pilot study cohortstatistics

Age Gender MOCA UHDRStotal motor score

Controls (N=5) 54 (21) 40% male 28 (1) 0 (0)

HD (N=15) 57 (7) 60% male 21 (4) 42 (13)

Gait test results • Gait low-frequency

spectral entropy feature

• Validation: strongly correlated with 10m walking test time (left)

• Discriminates controls from HD (right, Cohen’s d=1.2)

Balance test results • Balance dynamic acceleration

magnitude interquartile range feature

• Discriminates controls from HD(Cohen’s d=1.2)

Touchscreen data pre-processing• Left/right tapping clusters identified from x-y touchscreen

coordinates• Extract: tap timing events, tap placement statistics from

cluster properties

Tapping test results • Tapping time

coefficient of variation feature (horizontal)

• Tapping cluster placement spread feature (vertical)

• Discriminates controls from HD

Combining tests • Predict UHDRS total motor

score, linear regression• 14 features from tapping,

gait, balance tests• Select features using single

feature regression significance

• Optimal model 2 or 5 features (top)

• Prediction error ~10 UHDRSpoints (bottom)

Conclusions• Small pilot study: Smartphone-based testing

discriminates controls from HD across tapping, balance and gait

• Smartphone-based gait test validates against standard 10m walking test

• Smartphone-based testing can predict UHDRS total motor score within ~10 UHDRS points

• First steps on the road to using standard smartphones as serious tools in clinical and research practice in HD

Spyros Papapetropoulos MD, PhDVP, Neurodegenerative diseases and

Movement Disorders

Implementing Innovation:Rewiring Clinical Research

October 27th – 30th, 2016 ◦ Boca Raton, Florida

Digital Technology Disrupted the World as we knew it

15Yesterday’s advantage will be replaced by today’s trends

Steve Jobs, Co-Founder of Apple

“The biggest innovations of the 21st century will be at the intersection of biology and technology. A new era is

beginning.”

CommonPedometerGyroscope

AccelerometerGeomagneticSleep activity

Heart rate

SpecialtyPulse oximetry

Sun exposureECG, EEG, EMG

PK, Respiratory rateHeart Rate Variability

Pulse rateStress

Brain Activity Sweat

Blood pressure Skin temperature Skin conductance

Activity (steps)Climbing/elevation

GyroscopeAmbient light sensor

AccelerometerGestureProximity

Tracking chipRGB light sensor

Barometric pressureOutdoor Temperature

HumidityVoiceGait

Urine analysisWeight

Blood PressureGlucose

Water qualityInfrared

Outdoor temperatureVoice

Ocular pressureOther

Smartphones, Wearable Devices and Health Sensors are capable of quantifying health and disease

On SMART device Wearable Portable

Objective, Real world, eSource, Remote, Real time, Continuous

Patients are looking for a change– By 2021, average person will

have 3 personal smart devices– Relationship with physician/site

changed– Self motivated to find answers– Direct to patient marketing has

changed expectations

18

A sad reality – the cascade of drug development

Clinical trials have been centered around the site for decades mimicking delivery of healthcare

Sponsor

CRO

IRB

FDA

Patients

Site

Long, difficult to enroll and execute, expensive clinical trials with high failure rates and inconclusive data – Is it the drug or the trial?

The missed opportunities of traditional trials

A 6-month trial = 4,380 individual patient hoursOnly ~ 50 hours at a clinical site

4.330 hrs of missed dataPatient and family burden

Costs

Technology is creating a new research paradigm inside and outside the clinic

22

CTI@TevaClinical Trial Transformati

on

Smart

Making our trials smart - generating more insights!

23

Social MediaTelemedicineTrainingGamification

Smart PillsBYOD - ePRO Closed loop delivery

Adherence

Biometric Monitoring

Electronic consenting

In 2017 Teva will incorporate virtual visits (and more) into its existing studies

24Adapted from M Alsumidaie, Applied Clinical Trials 2013

At home drug delivery

Adding to the expanding Clinical Research Toolbox

Small Molecules and Biologics for Disease Modification

Effective Symptomatic Therapies

Novel Mechanisms, Pathways and delivery methods

Personalized medicine (-omics, imaging)

Biomarkers Throughout the Drug Development Process

The new clinical research paradigm will disrupt healthcare

New technologies support real time, continuous, self care/monitoring

Source: Tectonic Shifts in Healthcare. James R Mault MD, FACS VP & Chief Medical Officer Qualcommm

That can be leveraged by ALL stakeholders

Opportunities to Meet Stakeholder Needs

PHYSICIANS PATIENTS

PROVIDERS PAYERS

Open PRIDE Digital Health Sub-studyA Teva-Intel Collaboration

Case Study

Background

– Motor symptoms in HD are typically evaluated by physicians using a rating scale; UHDRS-TMS

– Clinician-rated scales are inherently subjective and may lead to intra- and inter-rater variability, and to a substantial placebo effect

– Easy-to-use digital health solutions can supplement clinical evaluation by providing rich, reliable, and sensitive datasets during and between clinic visits

– May allow objective real-time monitoring of symptoms and progression, treatment customization and reduce patient and caregiver burden

Open-PRIDE Digital Health Sub-study

–Exploratory sub-study– 60 Patients; Enrolment starts in 2016

–Delivery: The HD Algorithm– Detect and quantify Chorea

– Co-Developed by Intel and Teva

30

Validation Data Gathering (In-Clinic and @Home)

31

– Devices are continuously collecting data for the entire 6 months days of the trial

– Each device collects 3D accelerometer data that reflects the intensity and direction of movements of the device

Pebble Smart Watch

iPhoneSmartphone

X - Forward

Z - Down

Y - Right

Major hurdle for algorithm development: Filter normal from abnormal movement

Open-PRIDE Digital Health Sub-study*

32

Manage the Disease using Data

Data for Analysis

Researcher

INSIGHT / VALUE

Patient and Clinician Clinically

Meaningful Data

Smart Watch

Smart Phone interface

Disease platform

Big Data Analytics

(*) Almost Virtual; A Medical IoT Setting

The mobile application

Medication diary Pop-up reminders Testing Instructions

Chorea severity rating

In-Clinical Assessments

1.Timed Up and Go (TUG) test2.Sitting at rest (2 minutes) with arms

relaxed3.Sitting at rest (1 minute) with arms

extended 4.Standing at rest (30 seconds) 5.Ten Meter Walking Test6.Drinking from a cup test (repetitive

5 motions)7.Pronation-supination test (30 secs)

At-Home Assessment

Assessment Tasks1. Sitting at rest (2 min) with arms

relaxed2. Standing at rest (30 sec)

35

THANK YOU!

“Digital Biomarkers” for Huntington's Disease using Multiple Body affixed,

Light weight Sensors

Sensor MD Team†

University of Rochester

†Represented by: Gaurav Sharma

MC10 BioStampRC Sensor: Specifications and Advantages

Mode Sampling Rate

Dynamic Range

Recording Time (Max)

Accelerometer (Accel.)

31.25,50,100,200 Hz

2,4, or 8G 8-35 hours

ECG 125,250 Hz 0.2 V 17 hours

EMG 250 Hz 0.2 V 17 hours

Accel.+ECG 50 Hz(Accel.),125, 250 Hz (ECG)

2,4, or 8G(Accel), 0.2V

(ECG)

11-22 hours

Accel.+EMG 50 Hz(Accel.) 2,4, or 8G(Accel), 0.2V

(EMG)

11 hours

Gyro.+Accel. 25,50,100,250 2,4,8,16 2-4 hoursHz G(Accel) Off,

250,500,1000,2000 /sec(Gyro)

● Light weight (7 grams)

● Unobtrusive, body affixable

● Low power

● Long recording time

Pilot Study Overview

● Focus on motor symptoms in Huntington's and Parkinson's Diseases (HD/PD)●

●

●

10 HD, 4 pHD, 16 PD, and 15 Controls enrolled Five accelerometers for each participantIn clinic assessment + two day in home recording

Body affixed vs Body worn Sensors

More than 93% of participants are●

●

●

●

Comfortable with sensors Experience no interference with daily activitiesPleased with overall experience Ready to re enroll in future

Contrast with body worn sensors●

●

●

●

………...

Advantages of Multiple Sensors

● Potential for better/more information through

● Targeted selection of individual sensors for analysis

● Joint exploitation across sensors● Allow for effective motion analysis without being invasive to individuals' privacy (as compared to video alternatives)

Preliminary Study Results

Lack of Coordination in HD (walk)

Normalized vector cross

correlation of the sensor data

from left leg and right leg for

control

Lack of Coordination in HD (walk)

Normalized vector cross

correlation of the sensor data

from left leg and right leg for HD

Lack of Coordination in HD (walk)

Normalized vector cross

correlation of the sensor data

from left leg and right leg

Control vs HD

Lack of Coordination in HD (walk)

Scatter plot Control vs HD

Step Duration Identification

Effect of Medication on HD

For one individual●

●

On/off TetraBenazine Three 10 m walk tests, each

● Mean step duration (HD off) = 0.67 seconds

● Mean step duration (HD on) = 0.55 seconds

On/Off Medication for Parkinson'sPatient with severe at rest tremors

Spectrograms of principal acceleration component

On-medication (Levodopa)

Off-medication ((LLevodopa)

On/Off Medication for Parkinson'sPatient with severe at rest tremors

Relative power in characteristic 5Hz band and first harmonic

On/Off Medication for Parkinson'sPatient with mild at rest tremors

Spectrograms of principal acceleration component

Off-medication (Levodopa)

On-medication (Levodopa)

On/Off Medication for Parkinson'sPatient with mild at rest tremors

Relative power in characteristic 5Hz band and first harmonic

Summary● The time is ripe for broad adoption of sensors health data

analytics●

●

Light weight, body affixed, low power, long duration recording abilities Effective in combination with data analytics/signal processing

●

●

Multiple sensors are advantageous: analysis can target specific individual sensors or exploit jointly

Preliminary analyses show clear signatures of clinically observed motor symptoms in Huntington's and Parkinson's

●

●

●

Lack of limb co ordination in HD: apparent in cross correlation analysis between sensors on left and right legsSlowing of gait in HD upon going off medication apparent in auto correlation analysis of chest sensorAt rest tremors in PD apparent in spectral analysis of the hand sensors, also impact of medication

More Information

● Come see our poster:▪ “Wearable Sensors for the Objective Measurement

of Motor Features of Huntington Disease a Pilot Study”, Jamie Adams et al, Presidential Boardroom A, Nov. 4, 10:30 am 11 am and 2:45 pm 3:15 pm and Nov. 5, 11:30 am – 12:15 pm (Presented by: Mulin Xiong)

▪ Catch us for a conversation▪ We're looking for partners to take the work

further

Team Members

Karthik Dinesh Mulin Xiong Jamie Adams

Nirav Sheth A.J. Aranyosi Ray Dorsey

Gaurav Sharma

Thank You