MDEpiNet: Accomplishments and Capabilities Danica Marinac-Dabic, - - PowerPoint PPT Presentation

MDEpiNet: Accomplishments and Capabilities

Danica Marinac-Dabic, MD, PhD, MMSc, FISPE Director, Division of Epidemiology, CDRH/FDA On Behalf of MDEpiNet

2

Summary

• Brief MDEpiNet history/in the context of FDA

vision for the national system

• Scalable Domains
• Examples of Partnerships/Capabilities: A

Strategic look

3

2010 2011 2012 2013 2014 2015 2016 2017

MDEpiNet Launch International Consortia (e.g. ICOR, ICCR) Reports:

• Planning

Board

• MDEpiNet

Registry Task Force

• IMDRF

MDEpiNet Methodology Center at Harvard MDEpNet Science and Infrastructure Center at Cornell

MDEpiNet Partnership Coordinating Center at Duke

Develop and test drive novel methods and scientific infrastructure for device evidence generation synthesis and appraisal nationally and internationally

THE VISION FOR NATIONAL SYSTEM LAUNCHED FDA 4- day Public Meeting Day 1. Launch of FDA strategy Day 2. MDEpiNet Annual Days 3-4. Registries

4

MDEpiNet – Quick Facts

• Over 130 partnering public and private sector organizations from the device

ecosystem globally (academia, professional associations, registries, industry, patient organizations, health systems and more);

• Over 100 national/regional registries from 37 countries;
• 11 National CRNs and 4 International Registry Consortia;
• Over 750 clinical experts across clinical areas;
• Over 150 methodologists (statisticians, epidemiologists, data scientists,

informaticians, health care researchers and more);

• National and State claims data (e.g. CMS, NY SPARCS)
• National and International EHRs (e.g. PCORNet, Sentinel, VA, UK, Australia)
• Integrated Health Care Delivery Systems (e.g. Kaiser)
• Over 60 demonstration projects including projects with PCORNet and

Sentinel

• High Performance Integrated Virtual Environment (HIVE) collaborative space

with petabytes storage space and analytical tools

Access to over 30 million patient/device encounters in registries Access to hundreds of millions of claims records

www.fda.gov

6

Global MDEpiNet

Future MDEpiNet Chapter Academic Centers Data Sources Existing MDEpiNet Chapter

7

Registry Development National/International Consortia Development Electronic Device Data Capture (UDI) Task Force -Coordinated Registry Networks(CRNs) PASSION Initiative Active Surveillance Comparative Effectiveness Evidence Synthesis Claims Validation Linkage with other Data Sources Big Data Analytics Translational Epidemiology Augmenting Registries with PROs and Explant Analysis for Precision Medicine Assessing Minimally Important Difference (MID) for implants Patient and Family Engagement Committee Patient-led Device/Disease Specific Round Table

Infrastructure

Methods

Patient Engagement

MDEpiNet Work Streams

8

Scalable Domains of MDEpiNet

Tools Infrastructure Methodology Partnerships

9

Tools

10

Data Extraction and Longitudinal Track Analysis (DELTA)

• DELTA software was produced by collaborative work and support from

FDA, NLM, Harvard and Leahy at Tufts.

• It was first validated on outcomes and clinical trial databases and shown to

efficiently identify very low frequency events utilizing an array of Bayesian and frequentist inference methods.

• It now supports multiple simultaneous analyses, tracking the accumulating

experience of multiple devices, while monitoring multiple independent datasets simultaneously.

• Scalability demonstrated through series of retrospective then prospective

studies: a. Health Care System Level - including Partners Healthcare in Boston b. State level – including MA Angioplasty Registry sites linked to Massachusetts inpatient Claims DB, and MA Vital Statistics c. National level – NCDR Cath PCI registry ( e.g. vascular closure devices).

• By the end of the year it is expected to be installed in ICD registry and TVT

registry

• Earlier this year – DELTA became an open source available to all

stakeholders at no cost.

11

UDI as a Key Component of MDEpiNet Methodology Work

RAPID Registry Assessment of Peripheral Interventional Devices http://mdepinet.org/rapid/ BUILD Building UDI Into Longitudinal Data Sets for Medical Device Evaluation CARDIOVASCULAR DEVICES See http://mdepinet.org/build/

UDI – part of Real World Data Use DI to pull data from GUDID and auto-populate fields

12

maxi-HIVE location: White Oak /CDRH HPC storage: ~2 Petabytes cpu:3000 cores, extensible to 5000 wan : 10Gb Þ Internet2 lan: 40Gb Þ Infiniband platform: metal + SunGrid goal: ATO approved regulatory next-gen support platform for long term storage and large scale computations; to support regulatory submissions for NGS and standardization portal for NGS evidence submissions mini-HIVE location: White Oak/CBER server room storage: ~1 Petabytes cpu: ~1500 cores wan: 10Gb lan: 56 GB platform: barebone metal goal: research and scientific NGS portal with cutting edge production quality tools

Public HIVE as MDEpiNet Resource

Public-elastic HIVE location: ColonialOne/Ashburn datacenter storage: extensible to Petabytes cpu:+1000 cores wan : 10Gb Þ Internet2 lan: 10Gb Þ Infiniband platform: Lustre open source goal: to become extensibility platform for public HIVE users for their large scale computational needs for large clinical research projects. Public-HIVE location: GWU Dr Mazumder’s lab storage: ~300 Terabytes cpu: ~400 cores wan: 1Gb lan: 10 GB platform: metal goal: support and integrate wider community of researchers into HIVE process, allow access to cutting edge regulatory complaint tools and standards, perform pilot free projects with academic, industry and government entities to promote and ease the access to novel NGS techniques. To incorporate HIVE into education.

HIVE-ZONE

location: virtual cloud storage: ~extensible cpu: ~extensible wan: ~extensible lan: ~extensible goal: support large scale sporadic surge usage patterns for extra large clinical computations through openFDA .gov initaitve for FDA and for

• penHIVE platform for wider scientific research and education community.

Amazon

HIVE is an ecosystem of hardware/software/expertise allowing developers and users to implement variety

• f infrastructures/ tools/ visual interfaces

manipulate large amount of complex data through sophisticated computational pipelines in this “sandbox” like environment.

13

Methods

14

Examples of Expertise

• Causal inference with high-dimensional, complex, observational

data:

– Bayesian propensity score using regularizing prior distributions – Robust likelihood approaches for comparing many devices via Super Learning – Making better use of claims codes via Bayesian hierarchical prior distribution

• Meta-analysis of individual participant data for identification of

device-effect heterogeneity and for OPC construction

• Device surveillance that incorporate regulator loss functions
• Handling missing data in CRNs that use a distributed system
• Common Data Elements (CDE)
• Distributed models
• And much more!!!

15

Infrastructure

16

• Approach is to facilitate and/or leverage national

investments in registries and other relevant data systems (dual purposing) to create ‘National Medical Device Evaluation System on a fairly immediate basis, greatly minimizing the cost or development resources needed’

– Major ‘Quality and safety’ registries initiated by professional societies, states, healthcare systems, NIH/AHRQ, other – CMS claims including Part A,B,C,D – Commercial claims – PCORI CDRNs – All payer State databases – Comprehensive EHRs

CRNs

Focus is to provide timely information on device experiences for decision making by patients, physicians, regulators and all

• ther stakeholders

17

Evolving CRN Portfolio

National

• Ortho CRN (aka ICOR- USA
• Vascular CRN – VISION
• Cardiac CRN
• Neurology CRN – DAISI
• Gastrointestinal CRN – Obesity
• SPARED CRN – Prostate ablation
• Robotic Surgery CRN
• TMD/TMJ
• VANGUARD
• Women’s Health Technologies –

Uterine Fibroids, Pelvic Floor Disorders, Sterilization Devices,

• Breast Implants (NBIR pilot under

way)

• Abdominal Hernia CRN

International

• International Consortium Orthopedics

Registries (ICOR)

• International Consortium Vascular

Registries (ICVR)

• International Consortium of Cardiac

Registries (ICCR)

• International Collaboration of Breast

Registries Activities (I-COBRA)

18

Researchers Health Care Providers Patients Hospitals Integrated Delivery Systems EHR Vendors Government

ONC

Structured Data Capture (SDC) EHR

States Data

Harmonization/Interoperability Device Industry

Address ecosystem questions efficiently

Coordinated Registry Network (CRNs) as part of NEST

Registry Registry Registry

19

Selected Data Sources

• 10-20 years of clinical registries data (national/international)
• Hospital discharge data
• State data
• Medicare/Medicaid data
• AHRQ data sources
• Truven data
• RoPR - Searchable central listing of patient registries
• Gene Expression Omnibus (GEO/NCBI)
• Ingenuity Pathway Analysis (IPA)
• Electronic Medical Records and Genomics Network at the
• National Human Genome Research Institute (eMERGE/NHGRI)

including MCRI – Marshfield Clinic Research Institute)

• And more!

20

Partnerships

21

Example 1. RAPID/VISION: A Scalable Framework to Efficiently Conduct Real-world Enabled Vascular Clinical Trials

• Two MDEpiNet projects VISION (led by Cornel) and RAPID (led by Duke) were launched to

complement each other

• VISION - to develop National Vascular Implants Surveillance and Outcomes Network by linking VQI

registry to claims and other data sources for longitudinal assessment

• RAPID - to developed a core data element (CDE) set informed by multiple registries, EHR systems,

and stakeholders.

• Leveraging both efforts the SPEED pilot is now being designed to demonstrate how clinical trials

sufficient to support regulatory decision-making can be efficiently implemented into existing healthcare systems for expanding the indications

• The SPEED trial that will involve 12600 patients from the VQI registry with (estimated 4200

prospective and 8400 retrospective patients data and 2940 patients with 1 year follow-up data available for developing Objective Performance Criteria – OPCs

• The projects leverage the GUDID Database to enable comparative effectiveness research of

peripheral vascular technologies.

• It will involve data refinement and analyses to develop contemporary OPCs using RWE, data sharing

with industry stakeholders for potential labeling modification (e.g., longer lesions, heavy calcified lesions, diabetic patients).

• The new paradigm is that industry will have access to the conglomerate of data plus their own data.

Thus, if pre-specified statistical plan is met for a given subgroup analysis population as per examples above, a labeling modification could be requested/granted.

21

22

Example 2. Growth Across Clinical Areas

23

Example 3. International Registries

• 2011 MDEpiNet/ICOR registry launched in orthopedics
• 2012-2014 – international efforts replicated in cardiac and

vascular space - ICCR and ICVR

• 2014 – MDEpiNet proposed to IMDRF to create IMDRF

Registry WG – and produce essential principles documents

• 2015 and 2017 – Three IMDRF registry principles documents

produced (the concept of international CRN (iCRNs) endorsed by IMDRF along with the proposed methodology pilots)

• The MDEpiNet/ICVR is now working with manufacturers,

regulators and professional associations to champion the first international pilot to for expanding the indications for rAAA devices while applying the principles of IMDRF registry documents

• The goal is to finalize the protocol by December 2017

Danica.Marinac-Dabic@fda.hhs.gov