Increasing clinical trial diversity has been flagged both nationally and within the Harvard Catalyst community as a critical step toward making health research more accessible and relevant for all. But is it realistic to do for every trial? Or even necessary?

Ankur Pandya, PhD, is a health decision scientist who confronts big, tough questions around healthcare cost-effectiveness in the face of finite resources. Now, with a pilot award from our 2024 funding opportunity, Expanding Inclusivity, he is confronting the thorny question whether increasing diversity in clinical trials is worth the extra resources in time, money, and human participants.

Pandya is associate professor of health decision science at Harvard T.H. Chan School of Public Health. We caught up with him as he was digging into his data analysis for the pilot study.

At a time when many are advocating for greater inclusivity in clinical trials, you’re taking a step back and asking whether it is cost-effective to do so. Why?

I think increasing inclusivity in clinical trials–that is, improving representation of historically underrepresented groups, whether defined by race/ethnicity, health status, or other variables–has broad benefits. But increasing representation is going to take time, money, and the participation of people from those groups, who inherently make up small portions of the population. In some settings, there just aren’t as many people who can even participate.

If every trial enrolls the maximum representation from these groups, the same individuals might end up in many trials. This can burden participants, slow study results, and increase trial costs. These are the trade-offs. There’s no one-size-fits-all right answer.

“These tools and methods could help design clinical trials prospectively to get at that trade-off of how large versus how inclusive they should be. That’s the broader goal we’re trying to achieve, and it could affect every trial.”

In my field of health decision science, we weigh the expected benefits and the expected costs of different options — in this case, how inclusive should certain trials be? Where should we allocate our limited resources, in terms of expense, time, and who to enroll? How long are we willing to wait and how much money are we willing to spend on increased inclusivity?

We’re just scratching the surface on this framework, but if you could imagine this played out, these tools and methods could help design clinical trials prospectively to get at that trade-off of how large versus how inclusive they should be. That’s the broader goal we’re trying to achieve, and it could affect every trial. That’s big, we think.

Do you get pushback on the concept of rationing health resources based on what is or isn’t cost-effective?

Oh, yeah. All the time. In the subfield of cost-effectiveness analysis where I apply much of my work and teaching, we deal with this trade-off between health gains and the costs required to achieve them. We are literally providing tools to ration. We make that very explicit. And we get pushback.

My first counter point is: If you don’t want to use cost-effectiveness analysis to ration healthcare spending based on the metrics, how will you know what you are getting for the money? We are always rationing limited resources, whether this is done explicitly or implicitly.

We don’t have unlimited resources to put toward health or clinical trials. We could spend over 100% of our gross domestic product on improving our health if we wanted to. We could have vegan chefs following us around cooking us three meals a day and eliminate high cholesterol. We can’t. We don’t have the money or the resources to do that.

How will you quantify the value of greater inclusivity in clinical trials?

In our pilot, we propose three different ways that increased inclusivity in clinical trials can provide value that can justify the additional costs. The first relates to heterogeneous effects. If an intervention provides different benefits or side effects for different groups of people, then status quo sampling–which tends to oversample folks who usually enroll in these trials and exclude historically underrepresented groups–might miss some of those heterogeneous effects. If the effects are stark enough to change the statistical significance on the primary study outcome, we might get the wrong answer overall, because our trial was not inclusive enough.

“If you don’t want to use cost-effectiveness analysis to ration healthcare spending based on the metrics, how will you know what you are getting for the money?”

Even if those heterogeneous effects don’t change the overall outcomes, they could mask important subgroup differences that might reveal, for example, that the intervention works in one group but not in another. In such cases it would be important to understand the mechanisms underlying the difference. Is it due to social factors? Biological mechanisms?

The third way inclusivity can add value does not depend on heterogeneous effects. Let’s say the intervention has exactly the same effects and side effects across all groups. Yet in real-world practice, physicians and other clinicians and even patients are hesitant to prescribe, recommend, or take a treatment that hasn’t been studied in a group representative of that patient.

A survey by an economist at Harvard Kennedy School found that approximately 72% of physicians report patients asking if a treatment will “work in people like me.” That suggests that greater inclusivity in trials could increase uptake, and we have ways we can quantify that. On the other hand, if effects are homogeneous and uptake would be unaffected by greater inclusivity, maybe we don’t need as much diversity.

These are empirical questions that we can hash out to allocate resources where they’d be most beneficial. That includes the time commitment of folks who would be enrolled in these trials, which is not infinite. Where would the service they’re doing for society be most beneficial?

What’s driving your interest in this field of health decision science?

Yeah, it’s a niche field. You don’t really grow up saying I want to become a health decision scientist. I think intellectually, I’m motivated by problems where we don’t know the right answer. I like problems that present trade-offs and uncertainties, where we don’t already know what to do and we need a logical structure to organize information. We need a little bit of math to help us make this big decision.

In my introductory decision science class, I say to my students that we’re all decision scientists in some ways. If you’ve ever made a pros and cons list, you’re weighing trade-offs. If you’ve ever sat down with a friend and asked what’s important to you? What do you care about? What can you control versus what’s out of your control? These are all factors that underlie the very foundations of the health decision science methods that we use.

In the case of clinical trials, I see trade-offs. I don’t see a one-size-fits-all solution where we have to max out inclusivity across the board, or that we have to stick with the status quo because we throw our hands up and say: Oh well, resources are limited, so we have no choice. I see a big middle ground where many times it makes sense to increase inclusivity and other times, that may not be where we want to spend our limited resources.

These are complex problems, and our models can’t capture every aspect that may be important or provide a comprehensive solution. But the idea of organizing some of this uncertainty to sort through these kinds of difficult decisions appeals to me. I like taking these on.

What is the value of this pilot award to your research goals?

This is such a new idea that it’s hard to find the funding source that will take a risk on it, but doesn’t have a dog in the hunt, you know? If we had pitched this to a funder from industry, they might only have interest if the results would help their bottom line.

The fact that Harvard Catalyst could take a risk on this small ambitious idea and let us be objective, let us find what we find, has been very… what’s the right word? It’s been a catalyst [laughing] for this research that we’re trying to do, just enabling us to try this out. Otherwise I’d be doing it in my free time. This lets me do it while I’m fully awake, instead of at 2 a.m.