- M.D., Ph.D.
- Cancer Epidemiology
- School of Public Health
- UNC-Chapel Hill
- 2105B McGavan-Greenberg
Area of Interest
My career as a medical researcher started with assessing the role of analgesics on kidney function and mortality, which culminated in a paper in the New England Journal of Medicine. After my formal training in epidemiology at Harvard School of Public Health, I continued pharmacoepidemiologic research with analyses of the role of nonsteroidal anti-inflammatory drugs (NSAIDs) and aspirin on cognitive function, risk for colorectal cancer (CRC), and kidney
function. My current research focuses on the potential of NSAIDs to reduce the risk for CRC. Observational data show that NSAIDs users have a reduced risk for CRC. In the Physicians'
Health Study (PHS), however, we found no effect of randomized treatment with aspirin on CRC risk (Annals of Internal Medicine). This study raised important questions regarding the role of
selection bias in non-experimental research on the effects of long-term medication use. Selection bias may arise due to factors influencing the propensity to start medication use,
including indications and contraindications, but also due to factors influencing the propensity for patients to comply with regular, long-term medication use. I am currently studying such factors
in more detail in the PHS and the Womens Health Study. The issue of prevention of CRC by NSAIDs is timely due to the increased risk of cardiovascular disease associated with COX-2 inhibitors. My current research on NSAIDs and CRC is funded by my ongoing RO1 AG023178 Propensity Scores and Preventive Drug Use in the Elderly. This work includes work in progress on the effects of 9 years of randomized treatment with low-dose aspirin on the incidence of diagnostic endoscopies (colonoscopy and sigmoidoscopy) in over 40,000 women using data from the Womens Health Study.
My research in epidemiologic methods has included work on efficient study designs in pharmacoepidemiology, a comparison of different analytic strategies for recurrent events, flexible matching in case-control studies to increase efficiency, and measurement error correction in case-control studies. Currently I am assessing the value of two increasingly used analytic strategies to control for confounding in non-experimental research: propensity scores and disease risk scores. Propensity score techniques are increasingly used but my work has demonstrated little empirical evidence for their use over conventional multivariable outcome modeling in many situations. I am currently assessing the comparative behavior of propensity scores, disease risk scores, and conventional disease models using extensive computer simulations. Based on my prior work on correction for measurement error, I have developed an innovative combination of propensity scores and regression calibration to control for unobserved confounding. This work is being conducted through the Working Group on Methodology that I established at the Brigham and Womens Hospital Division of Pharmacoepidemiology and Pharmacoeconomics, and funded by my RO1 AG023178 Propensity Scores and Preventive Drug Use in the Elderly. This grant is currently transferred to UNC with a subcontract to the Brigham and Womens Hospital to continue this successful collaboration and enhance it with new collaborations within UNC.