Related Experiment Videos
Evaluating short-term drug effects using a physician-specific prescribing preference as an instrumental variable
M Alan Brookhart1, Philip S Wang, Daniel H Solomon
1Division of Pharmacoepidemiology and Pharmacoeconomics, Department of Medicine, Brigham and Women's Hospital/Harvard Medical School, Boston, MA 02120, USA. abrookhart@rics.bwh.harvard.edu
Epidemiology (Cambridge, Mass.)
|April 18, 2006
Summary
A new instrumental variable method using physician preference shows a protective effect for COX-2 inhibitors in preventing gastrointestinal complications, unlike traditional regression. This approach helps reduce bias in observational drug safety studies.
Area of Science:
- Pharmacoepidemiology
- Biostatistics
- Health Services Research
Background:
- Postmarketing observational studies are crucial for drug safety but face methodological challenges, particularly confounding by indication.
- Existing data sources often lack critical information on drug indications and other confounders.
- Instrumental variable (IV) methods offer a potential solution for confounding by indication in nonexperimental studies, but finding suitable instruments is difficult.
Purpose of the Study:
- To propose and illustrate a novel instrumental variable for pharmacoepidemiology.
- To assess the safety of cyclooxygenase-2 (COX-2) inhibitors compared to nonselective nonsteroidal antiinflammatory drugs (NSAIDs) regarding gastrointestinal complications.
Main Methods:
- Developed a time-varying IV based on prescribing physician preference for one drug over a competitor.
- Applied the IV method to a study comparing COX-2 inhibitors with nonselective NSAIDs.
- Compared IV results with conventional multivariable regression and randomized trial data.
Main Results:
- Multivariable regression found no protective effect of COX-2 inhibitors (risk difference = -0.06).
- The proposed IV method revealed a protective effect of COX-2 inhibitors (risk difference = -1.31).
- IV results were consistent with findings from randomized controlled trials (risk difference = -0.65).
Conclusions:
- The proposed physician preference IV method appears to significantly reduce bias from unobserved confounding in observational drug safety studies.
- This method shows promise for improving the reliability of pharmacoepidemiological research.
- Further research is needed to evaluate the sensitivity of this IV approach to potential assumption violations.