Download presentation
Presentation is loading. Please wait.
Published byTheodore Martin Modified over 9 years ago
1
Active Surveillance Using Longitudinal Data: A Pilot Project Drug Safety and Risk Management Advisory Committee Meeting Silver Spring, Maryland May 18, 2005 David J. Graham, MD, MPH Office of Drug Safety Drug Safety and Risk Management Advisory Committee Meeting Silver Spring, Maryland May 18, 2005 David J. Graham, MD, MPH Office of Drug Safety Center for Drug Evaluation and Research
2
Drug Safety and Risk Management Advisory Committee May 18, 2005 2 BackgroundBackground Multiple potential approaches Drug based Outcome based Setting based Variety of statistical methods Proportional reporting ratios Bayesian approaches Generally non-longitudinal Multiple potential approaches Drug based Outcome based Setting based Variety of statistical methods Proportional reporting ratios Bayesian approaches Generally non-longitudinal
3
Drug Safety and Risk Management Advisory Committee May 18, 2005 3 Project Concept Most pharmacovigilance - spontaneous reports No denominator Underreported numerator Datamining for signals Develop population-based ADR screening Longitudinal Prospective Real-time signal detection in longitudinal data Value of “+” signal Value of “-” signal Two complementary statistical techniques Most pharmacovigilance - spontaneous reports No denominator Underreported numerator Datamining for signals Develop population-based ADR screening Longitudinal Prospective Real-time signal detection in longitudinal data Value of “+” signal Value of “-” signal Two complementary statistical techniques
4
Drug Safety and Risk Management Advisory Committee May 18, 2005 4 HMO Research Network Consortium of 10 HMOs ~11 M enrollees CERT funded by AHRQ Record-linked Hypothesis testing 3 HMOs also FDA funded Consortium of 10 HMOs ~11 M enrollees CERT funded by AHRQ Record-linked Hypothesis testing 3 HMOs also FDA funded
5
Drug Safety and Risk Management Advisory Committee May 18, 2005 5 Sequential Probability Ratio Testing Periodic analysis as data accumulates Sequential measurement of signal strength Statistically based Can adjust for automated covariates Periodic analysis as data accumulates Sequential measurement of signal strength Statistically based Can adjust for automated covariates
6
Drug Safety and Risk Management Advisory Committee May 18, 2005 6 Weekly cohorts of exposed individuals, each cohort followed for 30 days
7
Rotavirus vaccine and intussusception (10x) Licensed 8/98 15 VAERS reports through 7/7/99 Immunization haltWithdrawn (1) (2) (1)
8
Drug Safety and Risk Management Advisory Committee May 18, 2005 8 Tree-Based Scan Statistic Use with hierarchical data ICD-9, MeDDRA, drug classification systems No a priori assumptions re: associations Sequentially moves up tree At each “cut”, evaluates all combinations Adjusts for multiple comparisons Identifies combination with maximum likelihood Two approaches All drugs c/w specific ADR All ADRs c/w specific drug Use with hierarchical data ICD-9, MeDDRA, drug classification systems No a priori assumptions re: associations Sequentially moves up tree At each “cut”, evaluates all combinations Adjusts for multiple comparisons Identifies combination with maximum likelihood Two approaches All drugs c/w specific ADR All ADRs c/w specific drug
9
Drug Safety and Risk Management Advisory Committee May 18, 2005 9 Hierarchical Data Drug All drugs Analgesics Non-narcotic NSAIDs Class of NSAIDs Specific NSAID ADR event All events Gastrointestinal Hepatobiliary Liver Hepatitis Drug All drugs Analgesics Non-narcotic NSAIDs Class of NSAIDs Specific NSAID ADR event All events Gastrointestinal Hepatobiliary Liver Hepatitis
10
Drug Safety and Risk Management Advisory Committee May 18, 2005 10 Tree-Based Scan Statistic
11
Drug Safety and Risk Management Advisory Committee May 18, 2005 11 Current and Future Goals Pilot test candidate drugs and ADR events Using retrospective data Known and unknown relationships Internal consistency, external validity Sensitivity, specificity Modify for prospective pilot surveillance Pilot test candidate drugs and ADR events Using retrospective data Known and unknown relationships Internal consistency, external validity Sensitivity, specificity Modify for prospective pilot surveillance
Similar presentations
© 2025 SlidePlayer.com. Inc.
All rights reserved.