From Big Data to Precision Medicine: New
Opportunities for Behavioral and Social Scientists
William Riley, Ph.D. Acting Director, Office of Behavioral
and Social Sciences Research Chief, Science of Research and
Technology Branch National Cancer Institute
Research Methods in a Data Poor Environment
• Priority is on prospective design and data collection • Limited data collection opportunities • Predominately cross-sectional or minimally
longitudinal designs • Unable to assess or control myriad
confounds • Control confounds via randomization
Research Methods in a Data Rich Environment
• Temporally Dense • Noisy But Precise • Computational • Predictive
A Brief History of a Data Rich Science: Meterology
• Local, limited measurement • Leverage communications technologies
(telegraph) to connect data across sites • Set standards for data integration • Continued leveraging of technical advances
in measurement and communication • Powerful, integrated data computationally
modeled to explain and predict phenomena Is it possible for behavioral medicine to become a data rich science?
"Nearly all the grandest discoveries of science have been but the rewards of accurate measurement." Lord Kelvin, 1872
Dawn of a Data Rich Behavioral Science
Rapidly accelerating technology development Ecological Momentary Assessment (EMA) methods
improved and delivered on cell phones Capture of digital traces from daily interactions with
technology Social media Call data records Consumer sensors
Sensors that can passively and continuously monitor cancer risk behaviors in context Physical activity sensors Smoking sensors Sun exposure sensors Environmental exposure sensors Dietary intake sensors (sort of)
Applications of computational modeling and new statistical modeling approaches that provide the analytic capabilities for intensive longitudinal (temporally dense) data.
6
Ecological Momentary Assessment
Ginexi EM et al. The Promise of Intensive Longitudinal Data Capture for Behavioral Health Research. Nicotine & Tobacco Research 2014 16 (4): S73-S75.
Growth Mixture Models: Elkhart Group Ltd
Click to edit Master title style
Computerized Adaptive Tests
0 1
2 3
- 1
- 2
- 3
high physical function
0 1
2
Question #2
1 2
Question #3
Wide Range High Precision Minimal Burden
low physical function
Question #1
New Modalities for Prospective Data Collection
• Citizen Science and Crowdsourcing Efforts – Mechanical Turk for:
• Volunteer Data Collection and Sharing (Quantified Self)
• Cognitive Testing of Survey Items • Environmental Field Assessments
• Opt-In Internet Panels • See Summary Report of the AAPOR Task
Force on Nonprobability Sampling • See Roshwalb et al. (2012)Towards the use
of Bayesian credibility intervals in online survey results. NY: Ipsos Public Affairs
Archival Big Data Sources in the Behavioral Sciences
“Digital Breadcrumbs” (Pentland, MIT) • Behavioral Data Traces gleaned from
consumer-based data sources – Social Media (Twitter, Facebook)
• Twitter opens up its 200 million users with 500 million tweets per day to researchers (2/10/2014)
– Internet Searches (Google) – Cell phone Use (# calls and texts) – Cable Box Data (hours of TV) – Auto Black Box data (miles driven, seat belt use)
Sensor Technologies
Population Scale Activity Measures • Population-scale measurement of physical
activity • Miniature, low-cost devices that measure
human motion using redesigned accelerometers in a user-friendly format
Stephen Intille, PhD, Northeastern University NHLBI, U01HL091737
Emerging Technologies and Assays for Adherence Monitoring
Xhale SMART “breathalyzer” for GRAS drug taggants
GlowCaps Proteus pill microchips and sensor
Drug (metabolite) concentrations via hair samples or dried blood spots
Psychophysiology
Autosense Santosh Kumar University of Memphis
Wearable Chemical Sensor System
• Chemical exposure varies by context, need personal exposure
• Selective detection of VOCs (hydrocarbon and acid vapors) Sensitive: ppb – ppm Real-time: sec. – min. Spatially resolved Wearable: cell phone size Cell phone based interface
Nongjian Tao, Arizona State University, NIEHS, U01 ES016064
http://www.airnow.gov
Implantable Biosensors • Measurement of analytes (glucose, lactate O2 and CO2) that
indicate metabolic abnormalities • Miniaturized wireless implantable biosensor that
continuously monitors metabolism – Inserted by needle subcutaneously – Operated remotely using a PDA – Multi-analyte sensor – One month continuous monitoring
Diane J. Burgess, University of Connecticut NHLBI, R21HL090458
Big Data Analytics – Pattern Recognition
Beck et al., Sci Transl Med 2011; 3:1-11
Big Data Analytics Computational Modeling
The fluid analogy depicts accumulation-depletion of the output (MVPA) as a result of changes in the input (self-efficacy). A controller / self-regulator relying on a sensed value of the output attempts to compensate for the input change, resulting in potentially significant variability.
Self-Efficacy
MVPA
What We Mean by “Model”? Conceptual Model
What Do We Mean by “Model”? Statistical Model
Anderson-Bill et al., J Med Internet Res 2011;13(1):e28
Theory of Planned Behavior from a Control Systems Perspective
Barrientos, Rivera, & Collins (2010). A dynamical model for describing behavioral interventions for weight loss and body composition change. Mathematical and Computer Modeling of Dynamical Systems.
Social Cognitive Theory (SCT)
• Dominant, influential theory of health behavior (Bandura, 1986)
• Lineage from Social Learning Theory • Core Concepts:
– Triadic reciprocity (reciprocal determinism) – Self-efficacy - Perceived ability to succeed in specific situations – Outcomes expectancy - Perceived
likelihood that performing the behavior will result in expected outcomes
– Cue to Action – stimulus that prompts behavior
Control Systems Model of SCT
Advantages of Computational Models
• Requires explicit and detailed system identification (even if only as hypothesis) and provides simulation capabilities
• Model is testable via precise and temporally dense data collection
• Ability to test “controllers” and their impact on the model to build JITAIs
• Utilize a series of N-of-1 trials to optimize the intervention
• Ability to be pre-emptive, not just reactive
THE U.S. PRECISION MEDICINE INITIATIVE
A Step Toward a Data Rich Cohort
President Obama’s State of the Union Address: January 20, 2015
“And that’s why we’re here today. Because something called precision medicine … gives us one of the greatest opportunities for new medical breakthroughs that we have ever seen.”
President Barack Obama January 30, 2015
www.nih.gov/precisionmedicine
Precision Medicine Concept is not new • Consider prescription eyeglasses, blood transfusions… • Prospects for broader application raised by recent advances
in basic research, technology development, genomics, proteomics, metabolomics, EHRs, Big Data, mHealth, etc.
• Reinforced by 2011 National Research Council report
And not new to behavioral medicine • MATCH
• Tailored Interventions – especially eHealth
• And increasingly adaptive interventions - mHealth
Nature, 2004; 429: 475-7
US Cohort Envisioned a Decade Ago
Nature, 2004; 429: 475-7
GWAS from $10 mill to $1K
$10 mill
Nature, 2004; 429: 475-7
ONC and HiTech Act Effects on EHR
$10 mill 13% EHR Adoption
Nature, 2004; 429: 475-7
New Assessment Methods
$10 mill 13% EHR Adoption
Nature, 2004; 429: 475-7
Explosion of Wearable Sensors
$10 mill 13% EHR Adoption
Nature, 2004; 429: 475-7
Mobile Technologies
$10 mill 13% EHR Adoption
Nature, 2004; 429: 475-7
Mobile Technologies
$10 mill 13% EHR Adoption
Nature, 2004; 429: 475-7
Mobile Technologies
$10 mill 13% EHR Adoption
0
Precision Medicine Initiative (PMI) Vision: Build a broad research program to encourage creative approaches to precision medicine, test them rigorously, and, ultimately, use them to build the evidence base needed to guide clinical practice. • Near Term: apply the tenets of precision
medicine to a major health threat – cancer • Longer Term: generate the knowledge
base necessary to move precision medicine into virtually all areas of health and disease
PMI: Policy and Privacy Prerequisites
To develop a new research and care model, PMI will: Engage Stakeholders: White House, HHS, other federal
agencies will solicit input from patient groups, bioethicists, technologists, privacy experts, civil liberties advocates, etc.
Modernize Regulations: Food and Drug Administration (FDA) to review regulatory landscape, support changes to advance precision medicine and protect public health
• Safeguard Privacy: Office of the National Coordinator for Health Information Technology (ONC) will develop interoperability standards, enable secure data exchange
PMI Proposed Support: FY16
Agency $ Million
NIH • Cancer • Cohort
$200 $70 $130
FDA $10
ONC $5
TOTAL $215
PMI: Longer Term
Generate knowledge base needed to move precision medicine into the whole range of health and disease • To reach this goal, the Initiative will support research to:
– Create new approaches for detecting, measuring, analyzing a wide array of biomedical variables: molecular, genomic, cellular, clinical, behavioral, physiological, and environmental
– Test these approaches in small, pilot studies – Utilize the most promising approaches in greater numbers of
people over longer periods of time to collect data of great value to both researchers and participants
PMI: National Research Cohort • Will comprise:
– >1 million U.S. volunteers – Numerous existing cohorts (many funded by NIH) – New volunteers
• Participants will be: – Centrally involved in design, implementation – Able to share genomic data, lifestyle information,
biological samples – all linked to their electronic health records
• Will forge new model for scientific research that emphasizes: – Engaged participants – Open, responsible data sharing
with privacy protections
EHRs Patient Partnerships
Data Science
Genomics Technologies
Why Mobile Technologies? Better Characterize Phenotypes and Outcomes
Why Mobile Technologies? Better Characterize Treatments
Why Mobile Technologies? Assess Treatment Predictors beyond Genetics
Why Mobile Technologies? Intensively Measure Behavioral and Environmental Risk Factors
Why Mobile Technologies? Fully Engage Participants as Partners
Advisory Committee to the NIH Director Precision Medicine Initiative Working Group
Co-Chairs: Richard Lifton, MD, PhD, Yale Univ School of Medicine, New Haven, CT Bray Patrick‐Lake, MFS, Duke Univ, Durham, NC Kathy Hudson, PhD, National Institutes of Health
Members: • Esteban Gonzalez Burchard, MD, MPH
University of California, San Francisco • Sekar Kathiresan, MD
Harvard Medical School, Boston
• Tony Coles, MD, MPH Yumanity Therapeutics, Cambridge, MA
• Sachin Kheterpal, MD, MBA University of Michigan Medical School, Ann Arbor
• Rory Collins, FMedSci University of Oxford, UK
• Shiriki Kumanyika, PhD, MPH Perelman School of Medicine, Philadelphia
• Andrew Conrad, PhD Google X, Mountain View, CA
• Spero M. Manson, PhD University of Colorado, Denver
• Josh Denny, MD Vanderbilt University, Nashville
• P. Pearl O’Rourke, MD Partners Health Care System, Inc., Boston
• Susan Desmond‐Hellmann, MD, MPH Gates Foundation, Seattle
• Richard Platt, MD Harvard Pilgrim Health Care Institute, Boston
• Eric Dishman Intel, Santa Clara, CA
• Jay Shendure, MD, PhD University of Washington, Seattle
• Kathy Giusti, MBA Multiple Myeloma Res Foundation, Norwalk, CT
• Sue Siegel GE Ventures & Healthymagination, Menlo Park, CA
Looking to the Future • Advisory Committee to the NIH Director Working Group
– Public workshops to inform report in September 2015 • NIH coordinating with White House, FDA, other agencies Seeking input from:
– Potential participants – Leaders of current cohorts – mHealth developers and researchers – EHR developers and researchers – Potential international partners
• Oct. 1, 2015—PMI launched! (pending FY2016
appropriations)
www.nih.gov/precisionmedicine
Questions?
