More than half of adults 65 and older report taking four or more prescription drugs. But studying how these drugs interact with one another poses a major challenge. Andrew Zullo, associate professor of health services, policy and practice and of epidemiology, recently received a $3.5 million grant from the National Institute on Aging to study the effects of using multiple medications concurrently on older adults’ health.
Zullo, whose background is in pharmacy, focuses on improving medication and vaccine use for older adults, with the aim of optimizing their physical and cognitive health. “As a pharmacist, I was trying to improve medication use for patients one at a time,” Zullo said. “But I ultimately realized that we needed the actual data to do that. I decided, then, that I would try to create some of that data so that we could guide prescribing and improve medication use for older folks. That’s when I went into public health.”
Tell us a little bit about your research area.
I conduct observational research using large datasets, like health-insurance claims and electronic health records, to answer questions about how to improve medication use in older adults. Sometimes this extends to improving vaccine use or interventions that improve medication use. A lot of the time, I’m looking at the effects of the medications themselves in terms of their intended purposes and also the adverse events they might cause.
What will you be studying with this new NIA grant?
With this particular grant, we’re going to identify potentially harmful drug interactions that occur when two medications work together to increase the risk of adverse events. Each drug by itself might be helpful for a person, but when you take two of them together, the beneficial effects may be neutralized or they may actually harm the person. The problem is that this has been hard to study in randomized trials. In this project, we’re going to identify which of those interactions really matter, estimate their effects, and then check that our results are correct using an external dataset.
Give us an example of a combination of medications that may cause an adverse event.
One example is beta-blockers, an anti-hypertensive drug class used to lower blood pressure, and acetylcholinesterase inhibitors, a medication class that’s supposed to improve memory in people with cognitive impairment or dementia. These two drugs may interact to slow the heart rate too much and could cause syncope, which is where people feel lightheaded and dizzy and may even fall. We’re anticipating that this could be an important drug interaction but we have to wait and see what the data show.
Why is it difficult to identify negative reactions between two medications?
One reason is that in randomized clinical trials—which is how we get most of our information, at least initially, about medications—we’re used to randomizing people to one medication class or maybe a couple of medication classes. But we never ever randomize people to use certain combinations of drugs that could be harmful. This is something that you can’t do in a randomized controlled trial because it wouldn’t be ethical or feasible.
The other reason is that many of the adverse events that we want to study are pretty rare. Falls, for example, may not be a particularly common outcome and we may not see enough of them in a randomized controlled trial to understand the effect that the interaction could have. It really requires us to use large datasets from health insurance claims and electronic health records. Otherwise, we wouldn’t have enough people in the study to estimate the effects.
This is a five-year project. What do you hope to accomplish in that time?
I’m hoping that by the end of the project we’re going to know a lot more about which drug interactions actually matter. Using that information, I’m hoping that we can update electronic prescribing systems so that we can make them much more efficient for clinicians and healthcare professionals who are prescribing medications.
Right now, the way a lot of the software works is that it flags every possible drug interaction based on theory and pharmacology, and so many providers will get five to ten warnings every time they prescribe a drug. They don’t know which of these warnings truly matters, and after a long-enough period of time, they start to have alert fatigue. What tends to happen is that providers get inured to the warnings, and when one of the warnings is actually important, it gets ignored.
We really need to figure out a way to improve these software systems. We’re hoping that this project will allow the software vendors to remove the warnings that are unhelpful, identify the ones that actually matter for patients and let the clinicians work without unnecessary interruptions.