MD MEDICAL SAFETY Real World Data: How It s Used at a Medical Device Company Myoung Kim, Ph.D., MBA. Andrew Yoo, M.D., M.S. Epidemiology and Health Informatics Medical Devices Johnson & Johnson May 29, 2015
Content RWD in the Medial Device Sector Example 1: Effectiveness of Bariatric Surgery Example 2: Occult Malignancy in Hysterectomy In Sum 2
Real World Data (RWD) vs Clinical Trials RWD Observational Actual Practice Retro- or Prospective (Could be randomized) Sensors Healthcare Information System Social Pragmatic Trials (Randomized RW Trials) Clinical Trials Interventional Ideal Practice Prospective Randomized Although focus has been on EMR and Payer Claims Data, other sources of RWD are increasingly available and integrated into health systems EMR Payer Claims Analytics Data Registry Biomarkers SoC or 1-Arm Trials Head to Head Trials EMR = Electronic Medical Records Payer Claims = the administrative records of health insurance companies and employers (billing records) 3
Most Common RWD Sources Today Health Insurance Claims Databases Truven MarketScan, Pharmetrics, Payor-specific claims databases (UHC, BCBS, Humana, WellPoint, etc.) Hospital Billing Records Premier database Electronic Health Records Inpatient, Outpatient (GE Centricity, UK s CPRD, etc.) Integrated Health Systems Databases Kaiser, Geisinger, Intermountain, etc. Dx or Tx Specific Registries 4
What is RWD? Experimental Observational (>> RWD/PMD*) Prospective (>> 1st Data) Retrospective (<< 2nd Data) Randomized Controlled Trials Types of Studies & Data Registries & Pragmatic Trials Large Database Studies Internal Validity External *PMD=postmarket data 5
Clinical Trials vs Real World Data Clinical Trials Real World Data Question Efficacy--Can it work? Effectiveness--Does it work? Setting Controlled clinical trial Real world clinical practice Purpose Regulatory approval Performance in real world Protocol mandates Many Few ( care as usual ) Comparator Often placebo Other treatments Subjects Homogeneous Diverse N Smaller Larger Compliance Higher Lower Outcomes E.g., BP, HbA1c, LDL, pain intensity, etc) Internal Validity Higher Lower External Validity Lower Higher E.g., MI, stroke, hospitalization, medication use 6
What is Big Data? Large volumes of high velocity, complex, and variable data that require advanced techniques and technologies to enable the capture, storage, distribution, management and analysis of the information Big data encompasses such characteristics as variety, velocity and, with respect specifically to healthcare, veracity Existing analytical techniques can be applied to the vast amount of existing (but currently unanalyzed) patient-related health and medical data to reach a deeper understanding of outcomes, which then can be applied at the point of care. Source: Transforming Health Care through Big Data Strategies for leveraging big data in the health care industry. 2013. 7
Challenges in Studying Medical Devices Inherent nature of the industry Greater diversity and complexity Iterative nature of product development The short product life cycle The learning curve in adoption Proper device operation requires: Optimal device design User training/ competency Adherence to directions Use environment 8
Challenges in Studying Medical Devices with RWD Due to lack of RWD infrastructure Absence of unique device IDs in health care databases today Can t identify patients treated with specific devices Can t follow them over time for outcomes or surveillance Limited exposure to a specific device or class (smaller patient population) Limited information on exposure/denominator Often requires more granular and/or procedural data than for Pharma research Surgical notes are in text format 9
RWD in Med Dev Industry is approaching a tipping point Mostly Class-Level Data (with exceptions) Surge of Product/Brand-Level Data Occasional surprises expected even before 2017 Unique Device Identifiers come on stream CMS SharedClarity?? 2014 2015 2016 2017 2018 2019 2020 ACA Episode-Based Bundle Payments (2013) UDI Implementation (2014) ACA Value-Based Physician Payment (2015) NOTE: Projected Growth inclusive of all aspects of RWD: (1) Volume of brand-level RWD, (2) Publications, (3) Market & regulatory expectations, (4) Industry spend on data & FTEs, etc. 10
Use of RWD Today Generalizability of RCTs needs to be tested Restrictive patient selection into RCTs Highly trained/experienced surgeons Example Study 1: Effectiveness of Bariatric Surgery RCTs are not feasible or ethical for many questions Example Study 2: Occult malignancy in hysterectomy 11
RWE Opportunities in Pharma vs. Medical Devices RWE in Pharma The RWE opportunity is approaching a tipping point Multiple and global rich data sources Sophisticated analyses Stakeholders using RWE for making critical decisions Healthcare Stakeholders worldwide are tapping into the RWE opportunity Pharma companies are responding at scale RWE in Medical Devices RWE in Medtech has been limited by poor data sources Sharp increase of brand-level RWD expected in 2-4 years, per UDI implementation RWE opportunities currently exist in procedure/ class level analyses in select product areas Efficiency through prioritization & flexibility is critical Select stakeholders starting to tap into RWE Scope, scale, and depth of RWE expected to remain limited in the immediate future, i.e., 1-2 years Medical device companies still limited by the same data constraints However, major increase in all aspects of RWE expected in 2-4 years 12
MD MEDICAL SAFETY Bariatric/Metabolic Surgery A Study using Real World Data
Bariatric / Metabolic Surgery Most common procedures in the US Gastric Bypass (Roux-en-Y) Sleeve Gastrectomy Adjustable Gastric Banding 14
Surgical Treatment and Medications Potentially Eradicate Diabetes Efficiently (STAMPEDE) The largest RCT, sponsored J&J To evaluated the efficacy of Bypass and Sleeve vs medical therapy among obese patients with uncontrolled type 2 diabetes (T2DM) Small sample size: 50 per arm (total 150 patients) Single center: Cleveland Clinic Sponsored by Johnson and Johnson The study results at 1 and 3 years published in NEJM in 2012 and 2014 The evidence of the bariatric surgery s effectiveness on the long term T2DM complications is yet to be provided, requiring a Potential study with a Long follow-up, at least 4 years Large sample size, a few thousands per arm Additionally, Medical Therapy and Surgical Techniques continue to evolve making State of the Art a dynamic concept 15
Structure of RCT STAMPEDE 16
Building a RWE Metabolic/Bariatric Surgery Study Identify the intervention Medical Coding Does not give potentially important technique differences 17
Real World Data: Bariatric/Metabolic Surgery Develop a control group 18
Real World Data: Bariatric/Metabolic Surgery Challenges in identifying an adequate control group 1. BMI is often missing 2. BMI codes are non-specific 3. Comorbidity severity are not clinically characterized 4. Prior major events may occur outside of continuous enrollment 5. Surgery group has run up period where insurance documentation and pre-op fitness may capture more codes 6. Reimbursement can be driven by non clinical reasons **Estimate of only 1% of eligible patients actually undergo surgery 19
Real World Data: Bariatric/Metabolic Surgery Identify the Outcomes 20
*time to event based on: Patient Age, Disease Duration, Severity of Disease 21
Chronic Comorbidity Complications 22
RWD: Bariatric/Metabolic Surgery Study 23
A RWE Bariatric Surgery Study The objective of this study is to evaluate the effect of laparoscopic bariatric surgery In Patients with medical comorbidities: Type 2 Diabetes (T2DM), Dyslipidemia, Hypertension (HTN), Depression, and Obstructive Sleep Apnea (OSA) Data Source: Optum Clinformatic database United Healthcare Insurance Claims data Medical claims, Pharmacy claims, and Lab results. Years: 2005 to 2013 Outcomes Healthcare Resource Utilization and Reimbursement Surgical Procedure Safety / Complications Metabolic Comorbidity onset and progression 24
Laparoscopic Surgery Patient Selection Process # of Patients Inclusion/Exclusion With at least 1 Laparoscopic BM Procedure or an obesity DRG code 37,588 Continuous Enrollment (min 6mo pre-index, min 14mo post-index), 17,717 16,991 Age (>18) 17,656 61 No missing gender 17,655 1 ICD-9 codes from obese (BMI 30.0-39.9) and morbidly obese (BMI >= 40) 16,691 964 Exclude patients with prior open, revision, unknown or mixed BM surgery 15,393 1,298 Had contraindicated gastric diagnosis 15,378 15 No delivery / pregnancy records during study period 14,172 1,206 No malign / benign neoplasm in gastrointestinal tract 13,658 514 Had 1 of the 5 comorbidities (T2DM, Hypertension, Dyslipidemia, 11,318 Depression, Sleep Apnea) 2,340 With any other major procedures during index BM surgery 10,623 695 # of Lost 25
Study Design Timeframe Surgery Cohort Cost match period Index Period 1 2 Post Index Periods 3 4 5 6 7 8 9 10 Control Cohort 360 days -90 day Index day 60 day Surgery admission day 60d+ 0.5yr Pts in control assigned an index date equivalent to the date 60+ 1yr 60d+ 1.5yr 60d+ 2yr 60d+ 2.5yr 60d+ 3yr 60d+ 3.5yr 60d+ 4yr 60d+ 4.5yr 60d+ 5yr Study Periods: Index Day (day 0): Surgery date Index Period: Surgery preparation period (Day -90 ~Day -1): Recovery Period: Day 1 to 60 Cost match period (Day -91 to min 90: to match a patient in Control with similar cost Post Surgery Periods: every 6 months until 5 years after the Recovery period Cost assessment is part of the objective. Index Period??? 26
Matching Process Patients were matched based on characteristics at the index date Exact matches were required for: Gender (M/F) Insurance Type (FFS, HMO, Other) BMI category (Obese, Morbid Obesity) Comorbidity Profile (HTN, Dyslipidemia, T2DM, Depression, Sleep Apnea) Preferential matching was conducted based on: Average Monthly Healthcare Costs (within 15% or $150) Age (within 5 years) Enrollment (at least 2 follow up periods) Minimum DELTA = Abs( Cost Difference) + 10*(Age Difference) 2 + 900*Max(Case Periods Control Periods, 0) 27
Summary of Matching Results Matched rate 85%; The surgery patients and their matched cohort had similar distributions on pre-study cost, age, and Number of study periods Unmatched patients may have had: Restricted surgery patients index date after 10/1/2006 (~53% of unmatched population) Highest pre-index costs for bariatric surgery cohort Strata with fewer patients Each strata represents a set of characteristics used in matching (index year (7) X gender (2) X insurance plan (3) X comorbidity combination (31)=1302 potential strata) Over 4,000 matched pairs 28
Challenges in Conducting a Retrospective Comparative Study Treatments are not randomly assigned Inherent causal inference challenges: Associations vs causal relationships Confounding and selection bias Requires careful planning and often sophisticated statistical modeling techniques Hence the need for careful design and statistical control of confounding The Clinical variables for evaluation may not be available; surrogates are utilized; Deep understanding of Medical Practice and how it is translated into RWD Major reasons for Dropout Implications RWD Discontinued participation of the insurance plan Dropout may be assumed at random RCT Lack efficacy Safety concern Dropout is often associated with treatment effectiveness or safety 29
Overview Summary of results Preparing for publication submission Preliminary results show associated reduction in medications, labs, and complication diagnoses (renal, liver, cardiovascular) Real world data supplements STAMPEDE clinical (controlled) results and points to similar directional improvement of metabolic comorbidities Confirms and quantifies expected surgical complications (hernia, malnutrition) in real world data 30
RWD evolving very quickly in Medical Device New data will enrich clinical information 1. Surgical Technique (digital cameras utilized for MIS surgery) 2. Electronic Medical Record (EMR) linkage Natural Language Processing to extract data from free text 3. Unique Device Identifiers (UDI) will allow for device specific evaluations (comparative effectiveness) 4. Linked Data Sets Cost of Care Reimbursement EMR data 5. Captures as close to real time clinical practice changes New Drugs, Procedures, Devices And be utilized for: Safety Surveillance, Value Based Technology Evaluations, Quality of Care Assessments 31
In Sum RWD research involving medical devices has many challenges With UDI, RWD in the Medical Device Sector is approaching a tipping point for substantial growth Demand for observational/rwd researchers in the Sector is rapidly growing RWD as is today without UDI can add value and insight to many important questions involving medical devices 32