PCORnet Use Cases Observational study: Dabigatran vs warfarin and ischemic and hemorrhagic stroke / extra-cranial bleeding 1
Observational study: Dabigatran vs warfarin and stroke / bleeding Goal: Compare ischemic and hemorrhagic stroke and extracranial bleeding rates among new users of dabigatran or warfarin therapy who have atrial fibrillation/flutter
Dabigatran vs warfarin: Funding / leadership An FDA sponsored Mini-Sentinel evaluation Lead team: Alan S. Go, MD, Kaiser Permanente Northern California, Oakland, CA; Joshua Gagne, PharmD, ScD, Brigham and Women's Hospital and Harvard Medical School, Boston, MA; Marsha Reichman, PhD, FDA; David Graham, MD, MPH, FDA www.mini-sentinel.org/work_products/assessments/mini-sentinel_protocol-for-assessment-of-dabigatran.pdf 3
Dabigatran vs warfarin: Participating Mini-Sentinel sites Kaiser Permanente N. California Northwest Colorado Hawaii Mid-Atlantic Georgia Group Health Cooperative HealthPartners Marshfield Clinic, Meyers Primary Care Inst, Harvard Pilgrim Health Care, Henry Ford Health System Aetna Humana HealthCore (WellPoint) Optum 4
Dabigatran vs warfarin: Overview Design: New user parallel cohort Eligible patients: >21y.o. with atrial fibrillation/flutter without warfarin or dabigatran in prior 180 days, plus others Drug exposure: Pharmacy dispensing data with days supply Primary outcomes: Ischemic stroke, hemorrhagic stroke, extracranial hemorrhage Effect estimation: Stratified Cox proportional hazards regression; Stratified rate differences Both use marginal structural models with inverse probability weighting Confounder adjustment: propensity score matching
Dabigatran vs warfarin: Inclusion criteria First dispensing of dabigatran or warfarin from November 1, 2010 to most recent data Age >21 years at first dispensing One or more diagnoses of atrial fibrillation or atrial flutter based on inpatient or outpatient ICD-9-CM codes (427.31, 427.32) on or before the first dabigatran or warfarin therapy 6
Dabigatran vs warfarin: Exclusion criteria <180 days complete pharmacy and clinical data immediately preceding first exposure to dabigatran/warfarin (index date) Any dispensing for warfarin, dabigatran, rivaroxaban or apixaban during the 180 days before index date Known mechanical heart valve or diagnosed mitral stenosis at index date based on diagnosis or procedure codes Chronic hemodialysis or peritoneal dialysis at index date based on corresponding administrative diagnosis and/or procedure codes History of kidney transplant at index date based on diagnosis and/or procedure codes Resident in a skilled nursing facility or nursing home at index date 7
Dabigatran vs warfarin: Drug exposure Assessed from pharmacy dispensing records, including days supply Exposure begins the day after the first dispensing (index) date. Gaps <7 days are ignored (gaps <14 days will also be evaluated) INR testing data will also be used to assess warfarin exposure when available 8
Dabigatran vs warfarin: Primary Outcomes Ischemic stroke: hospital primary discharge diagnoses Intracranial hemorrhage: hospital primary and secondary discharge diagnoses Sensitivity analysis will exclude intracranial hemorrhage associated with major trauma Major extracranial hemorrhage: hospital primary discharge diagnoses of extracranial hemorrhage (gastrointestinal and all other sites combined) 9
Dabigatran vs warfarin: Secondary Outcomes Hospitalized acute myocardial infarction: Primary discharge diagnosis for myocardial infarction or Death occurring within one day of an emergency department visit for acute ischemic heart disease 10
Dabigatran vs warfarin: Data sources Administrative files: Demographic data Eligible person time Dispensing data: Outpatient medications, including dosage form and days supply Claims: Diagnoses Procedures, including INR obtained and selected inpatient medications EHR (available for a minority) Lab test results (INR, renal function, etc.) 11
Dabigatran vs warfarin: Effect estimation Time to event models: Proportional hazards regression models stratified by matched sets within locations Test for interactions between treatment and time Rate differences: Stratified inverse variance weighted summary rate differences for person time data 12
Dabigatran vs warfarin: Follow up Begins on the day after the index date Continues while they are continuously exposed to the index medication Eligible to enter the analysis only once 13
Dabigatran vs warfarin: Confounders Risk factors for bleeding Risk factors for ischemic stroke, MI, thromboembolism Overall health status, including measures of frailness Medications 14
Dabigatran vs warfarin: Confounder adjustment Propensity score (PS) matching Dabigatran and warfarin (1:1 and variable ratio) PS models will be developed in each site 15
Dabigatran vs warfarin: Sample size Projected 26,000 41,000 matched pairs Excellent power to find 50% difference for every outcome Variable power to find 25% difference 16
Dabigatran vs warfarin: Data sharing Distributed NO individual level data shared Identification of included/excluded new users of dabigatran and warfarin Creation of site-specific propensity score models Identification of propensity score matched pairs of dabigatran/warfarin new users Centralized NO protected health information (PHI) is shared Regression modeling Marginal structural models to assess robustness to time varying confounding Requires one record per person w/ exposure, outcome, subgroup (age <65 or 65+, etc) and confounder status (propensity score, or inverse probability weight) in each 30 day period after index date, censoring in interval, arbitrary ID of matched pair 17
An Analysis of Approaches to Privacy Protecting Data Sharing www.mini-sentinel.org/work_products/statistical_methods/mini-sentinel_methods_evaluating-strategies-for-data-sharing-and-analyses.pdf 18