Population-Based Epidemiologic Safety Studies: Overview and Challenges

1
Population-Based Epidemiologic Safety Studies
Overview and Challenges

• Drug Safety and Risk Management Advisory
  Committee
• Silver Spring, Maryland
• June 19, 2005
• David J. Graham, MD, MPH
• Office of Drug Safety

Center for Drug Evaluation and Research
2
Why Postmarketing Safety Studies?

• Residual uncertainty at approval
• New safety signal post-approval
• Beyond case reports
• Serious
• Potentially large exposure
• Potentially large excess risk
• Potentially safer alternatives
• Less serious indication
• Inappropriate off-label use

3
Mechanisms for Performing Postmarketing Drug
Safety Studies

• Phase 4
• Ad hoc postmarketing studies
• Ad hoc postmarketing studies
• Existing national data
• Cooperative Agreement Program
• GPRD
• Special projects
• Kaiser Permanente, California
• Veterans Administration database

Performed by Companies
Performed by ODS
4
National Data Resources

• NHANES
• QT prolongation in general population
• NHDS
• Pancreatitis in children
• AMI in children
• NAMCS
• Prevalence of drug therapy
• Prevalence of diagnosed diseases
• NCHS statistics
• Age-specific denominators
• Cause of death

5
FDA Population-Based Data Resources

• Cooperative Agreements
• HMO network 1.8M, gt5 yrs
• TN Medicaid 1.5M, gt15 yrs
• UnitedHealth 2M, gt5 yrs
• GPRD 3M, gt5 yrs
• Special projects
• Kaiser California 6M, gt5 yrs
• VA Medical System 3.4M, gt5 yrs
• ? Medicare 36M

6
Basic Database Features

• Population-based
• Large
• Longitudinal
• Automated claims
• Pharmacy
• Provider-encounters
• Procedures
• Record-linkage capacity

7
FDA Funding Costs of Resources

• National resources 0
• Cooperative Agreements 900 K per yr
• GPRD 500 K per yr
• Special projects
• VA medical system 10 K
• Kaiser Permanente 60 K
• Medicare Could be

8
Cooperative Agreement Program

• Access to 3 population-based data resources with
  research expertise
• Longitudinal
• Outpatient prescriptions
• Diagnosis procedure claims
• Medical record linkage
• No routine death ascertainment
• Ability to study
• Patterns of use w/in database
• Exposure-outcome
• Effects of regulatory interventions

9
Claims Data Quality

• Pharmacy claims
• High validity
• ? 95 complete within 2 mos
• Diagnosis claims
• Outpatient low validity
• Inpatient it depends
• ? 95 complete within 6 mos
• Procedure claims
• High validity
• ? 95 complete within 6 mos

10
Proposed Changes in Cooperative Agreement

• Shift from grant to contract
• Growing pains
• Intention to fund multiple databases
• Focus on safety-related issues important to FDA
• Retain collaborative relationship

11
General Practice Research Database

• UK-based electronic medical record
• GP-centered, longitudinal
• Low turnover
• GP visits, health measures
• Consultant referrals and hospitalizations
• Labs, procedures results
• Computer-generated outpatient prescriptions
• Death ascertainment
• Complex relational file structure
• In-house access via Internet

12
Unique Limitations

• UK population - not US
• National formulary - cost containment
• Different health care standards and practice
• Different prescribing patterns
• Very large data files
• In-house resource requirements

13
Other Population-Based Resources

• VA medical system
• Moving to EMR
• Males, older, sicker
• Some acute hospitalizations missed
• Some lab data
• Kaiser Permanente
• Large HMO
• Closed system, integrated
• Some formulary restrictions
• Computerized lab data
• Death ascertainment

14
Types of Studies

• Patterns of drug use
• Persistency
• Co-prescribing
• Case series
• Prevalence cohort
• Inception cohort
• Case-control
• Nested case-control
• Patient surveys

15
Study Examples

• Factors associated with QTc prolongation
• Predictors of COX-2 inhibitor use
• Birth defect incidence with an antibiotic
• Incidence of severe liver injury with a diabetes
  drug
• Chronic use of corticosteroids in children
• Incidence of rhabdomyolysis with lipid-lowering
  drugs
• Compliance with contraindicated use labeling
• Compliance with liver enzyme monitoring
  recommendations

16
Potential Limitations of Databases
Specific to some
Common to most or all

• Outpatient Rxs only
• Market penetration
• Sample size
• No OTC, herbal, alternative
• Data time-lag
• Special populations
• Completion time
• Privacy issues

• Employee-insurance based
• Formulary issues
• Co-pays
• Patient turnover
• Lab results
• Death ascertainment

17
Challenges (1)

• Budgetary
• Databases
• Operation
• Infrastructure
• Personnel
• Training Hardware/software
• Methodologic
• Study design
• Proper covariates
• Power

18
Challenges (2)

• Topic identification selection
• Matching question data resource
• Prioritization
• Use for regulatory purposes