Sean Hennessy, Pharm.D., Ph.D., is a pharmacist and pharmacoepidemiologist who feels he has the “best job in the world.” He is a professor of epidemiology in the Department of Biostatistics, Epidemiology, and Informatics at the University of Pennsylvania (UPenn)’s Perelman School of Medicine. His research centers on drug-drug interactions and cardiovascular effects of non-cardiac drugs, and on developing and evaluating novel research methods that are useful for pharmacoepidemiology. One area he’s researched is how weather affects drug effectiveness.

Dr. Hennessy also leads the Center for Pharmacoepidemiology Research and Training (CPeRT) at UPenn, which strives to be the “academic home of pharmacoepidemiology” within the university. With a team of over 100 faculty investigators and investigators-in-training, CPeRT works to advance population health through science, and through training scientists—including members of the U.S. Food and Drug Administration (FDA).

In addition to his UPenn posts, Dr. Hennessy serves on FDA Advisory Committee meetings, and on panels at the National Academies of Science, Engineering, and Medicine.

We sat down with Dr. Hennessy to discuss CPeRT’s FDA training program, trends and priorities for real-world evidence (RWE) adoption, and where he’s most excited to see real-world data (RWD) applied.

Responses have been edited for clarity and length.

Q: How did CPeRT’s training program with the FDA come about, and what are its goals?
A:
The FDA requested applications for an external group to conduct a training in pharmacoepidemiology. We responded, and now we’re in the fourth of a five-year commitment with the FDA, which we hope to renew after next year. Each week for 33 weeks per year, a different UPenn faculty member teaches a group of about 35 FDA scientists at the FDA’s campus, and now virtually.

With 33 sessions within six different modules or courses, it’s a full and wide-ranging curriculum of pharmacoepidemiology. There’s also a wide range of people who take the course—some are Ph.D. epidemiologists and others are physicians or pharmacists with little or no epidemiology background. 

When you think of the opportunity to have impact through teaching, it doesn’t get much better than providing information to scientists at the FDA, who regulate drugs, biologics, and devices in the country. This is an unbelievable opportunity to impact population health through this FDA teaching role, and it’s something that our team of faculty members and I value and enjoy tremendously.

Q: What are key factors to consider when designing an observational study, or when assessing whether a data set is fit for purpose?
A:
The most important step is defining your research question. Is it a question about causal inference, or is it a question of prediction? If you are comparing two products with regard to safety or effectiveness, are people who receive those different products comparable? How, in real life, are treatment decisions made between people getting drug A versus drug B? What kinds of confounding would you expect there to be?

In addition to that, you also need to consider whether the available data matches the research question. Are the exposure and outcome of interest captured in the data? Are all of the potential confounding factors?

We also try to include control analyses, such as negative control outcomes, when comparing treatments. If you're comparing drug A to drug B on an outcome where you expect them to vary, are there other outcomes that you expect to remain consistent between groups? It’s good to include those outcomes as a test to make sure that your data and research methods are producing expected results. 

Q: What are the big shifts in the use of RWE for regulatory purposes today?
A:
Regulators throughout the world have been using real-world evidence for regulatory decision-making for decades. One big shift is in the use of RWE to inform effectiveness claims rather than just safety. And that’s a controversial issue, it’s polarizing. That said, there are some instances now where initial approval have be made without randomized trials.

However, if there is the opportunity to use nonrandomized designs for regulatory studies to demonstrate benefit, we should require more than one study, and they should be based on different assumptions. For example, cohort studies assume no unmeasured confounding. We would need to supplement at least one well-done, believable cohort study with those using a design that relies on different assumptions, other than unmeasured confounding—an instrumental variable design, for example. But even then, there would be plenty of people who say that, without randomization, you simply can't make claims of benefit.

Q: What are the biggest challenges in using RWE to inform effectiveness decisions?
A:
When talking about the use of RWE to demonstrate benefit, I’d say that the biggest roadblock is—and I don’t think it’s necessarily an inappropriate roadblock—that generally speaking, randomized trials provide more convincing evidence of effect than nonrandomized studies. And there are a plenty of people who are highly skeptical of evidence from nonrandomized studies claiming to demonstrate benefit.

I don’t think that those obstacles are going to be overcome quickly or easily. It’s going to take time. And it’s going to require a history of examples in which nonrandomized real-world studies are later replicated in randomized trials and found to have been valid.

Q: How are RWD and RWE supporting the response to COVID-19? Conversely, how do you expect that COVID-19 will impact the use of RWE going forward?
A:
Depending on how broadly you define them, all of the epidemiology involving COVID-19 can be considered RWD or RWE. Almost none of the data on the disease were developed for the purposes of research. And I think that RWD and RWE are playing a big role in assessing different aspects of the COVID-19 pandemic.

In terms of how COVID-19 is affecting our field, I’d say it’s continuing the trend of broadening the group of people who are doing RWD studies. In the past, everybody doing these studies had epidemiology training, or called themselves an epidemiologist or pharmacoepidemiologist. Now, we see a broader group, including data scientists, for example, conducting these real-world data analyses. 

When a vaccine is developed, existing infrastructures for assessing the real-world safety of vaccines will be put into place to ensure and measure safety. We can expect substantial concern, whether it’s justified or not, by members of the public about its safety and effectiveness, and therefore pharmacoepidemiology, RWD, and RWE are going to be important to try to provide assurance.

Q: What are you most excited about for the future of RWE?
A:
It’s an exciting time to be in this field. I’m most interested about how we can apply pharmacoepidemiology methods to address new types of questions—for example, our team has done a couple of studies on whether the effect of a medication is affected by weather, or by temperature in particular.

I’m also excited about the linkage of health care data, including drug consumption and outcome data, with other sources of data like motor vehicle crash or driver’s license data, to allow us to answer broader and novel questions. Or linking inpatient data to outpatient data to allow us to assess long-term effects of what happens to patients when they're in the hospital. I think that’s really interesting.