Aug 22 2005
It is well known that aspirin, a non-selective, non-steroidal anti-inflammatory drug (NSAID) that inhibits cyclooxygenase (COX), reduces the risk of heart attack and stroke.
Non-aspirin non-steroidal anti-inflammatory drugs (NANSAIDs) such as ibuprofen and naproxen may reduce this same risk, but studies have shown conflicting results. Some have shown no association between NANSAIDs and heart attacks; some have shown an increased risk; and others have suggested a lower risk of heart attack, particularly with naproxen.
A new epidemiological study from the University of Pennsylvania School of Medicine, based on detailed patient surveys rather than administrative databases of patient prescriptions and billing records, suggests that these administrative-database studies may not accurately estimate the risk of heart attack among users of naproxen and ibuprofen. Indeed, results from the Penn study showed a protective relationship between NANSAIDs and heart attack. The study findings are published in the August issue of Pharmacoepidemiology and Drug Safety, and will be presented at the 21st International Conference on Pharmacoepidemiology and Therapeutic Risk Management on August 23 in Nashville, TN.
Previous studies on NANSAIDs used prescription records from billing data or electronic medical records (referred to as "electronic databases"), but not direct interviews with patients about their lifestyle or their over-the-counter use of NANSAIDs or aspirin. However, a February 2005 study by lead author Stephen E. Kimmel, MD, Associate Professor of Medicine in the Cardiovascular Division and Associate Professor of Epidemiology in the Department of Biostatistics and Epidemiology at Penn, suggested a benefit of non-selective NANSAIDs, when data were collected from study participants instead of relying on the limited information from electronic databases.
Although all epidemiological studies have potential limitations, electronic databases have several limitations inherent in the source of data: First, electronic databases record only prescription records, not over-the-counter use, so most use of NANSAIDs like over-the-counter ibuprofen is unaccounted for. "By using prescription databases you don't completely capture the non-steroidal use," says Kimmel. "You are calling people non-users of the drug when they really are. In our survey, 35 percent of participants had taken a non-steroidal, mostly over-the-counter, in the week prior to taking our survey."
This misclassification of users as non-users of NANSAIDs skews interpretation toward finding that NANSAIDs have no effect on the risk of heart attacks. The researchers found that of all the non-steroidal use, 80% was over the counter, and mostly ibuprofen (e.g., Advil).
Second, electronic databases do not capture complete information on nonprescription aspirin use. "Many people use over-the-counter aspirin for everything from headaches to protecting the heart," says Kimmel. "This means you can't separate the aspirin users from the non-users." This lack of complete information makes it difficult to examine the effects of NANSAIDs in the absence of the anti-platelet effects of aspirin.
Finally, electronic databases do not take into account risk factors for heart attacks, such as lower physical activity and higher body mass index, that may be more common in NANSAID users, who tend to have osteoarthritis.
The researchers hypothesized that the lack of these three types of data or distinctions in studies based on electronic databases would bias results toward showing no association between NANSAID use and lower risk of heart attack. In the new study, participants-1,669 first-time heart-attack survivors and 6,604 controls without a heart attack-were asked about their use of both prescription and over-the-counter non-steroidal and aspirin use and about several risk factors for heart attacks that are typically unavailable or incomplete in administrative databases such as weight and level of activity. When each potential category of bias was removed, NANSAIDs showed a stronger protective association with heart attack. Because of this, the researchers concluded that the limitations of electronic databases might be responsible for the lack of association of NANSAIDs with lower risk of heart attack seen in other studies. "As you add into the model more and more useful and relevant information, the association between non-steroidals and heart attacks changes and it changes in the direction of showing more benefit, or less harm," says Kimmel. The researchers caution that their results are not definitive and suggest that randomized trials are needed to more accurately address the possible risk and benefits of NANSAID use.
"Some recent studies have shown an increased risk of heart attack from traditional non-steroidals and most have not shown a lower risk, except for our study," says Kimmel. "The bottom line on this paper is that we are not saying we know the whole answer, but our data suggest there might be beneficial effects of non-steroidals and there are clearly limitations to interpreting the epidemiological studies that are now out there." Kimmel states that "balancing the risks and benefits of both traditional NANSAIDs and COX-2 inhibitors is so critical to proper patient care that we need to put our resources into randomized clinical trials that are designed to address this issue.
Study co-authors are Leonard Ilkhanoff, James D. Lewis, Sean Hennessy, and Jesse A. Berlin, all from Penn. This research was funded by the National Institutes of Health, and some data collection was supported by grants from Searle Pharmaceuticals, Inc. (now Pfizer, Inc.) and Merck & Co. Inc.