The Importance of RWD in Rare Disease Research

ACT: How important is the role of real-world data (RWD) in clinical research for cancer and other rare diseases?

Gliklich: When I think about real-world data, I think both in terms of leveraging the data that exists for retrospective studies as well as using it for prospective clinical research, which I think may be more of your interest. Early in development, real-world data gives a backbone of prevalence data, patient journey, understanding, understanding what usual care is, understanding unmet medical needs, seeing where opportunities might exist for intervention. As you start early trials, like what we see is real-world data can be used to develop digital twins; you can actually add synthetic control arms to Phase II studies. The value of that is that if you’re then planning a Phase III you can make an economic decision as to whether or not to go forward based on what you would project to be results in that way. We use it to generate and test surrogate endpoints, so in cancer, rare diseases, etc., there’s going to be endpoints that may take years to develop, but if you can use a combination of real-world data and sometimes AI (artificial intelligence), and there are now drug development tool pathways under the FDA for this; you can get to surrogate endpoints that can reduce the time period for follow up, and that might be predictions of disease progression, mortality. Another thing we do is phenotyping with it, so if you have a large body of data, apply an AI approach of what we call digital phenotyping, you can look for subtypes. Even rare diseases, there may be important subtypes for certainly in cancer, there’s important subtypes. We do a lot of work chronic diseases. I think we’ve identified 33 subtypes in lupus for example.

Later on, as you move through development, I’m sure you know this, real-world data can help better design protocols, so you apply inclusion/exclusion criteria, you see where the patients fall in, fall out, modify those before you actually get to the site, so that you don’t have a trial that fails because an advisor said, “Put in urine sodium,” and nobody ever measured urine sodium. Then during the trials, AI can help amplify identification for patients, particularly rare diseases, and you can identify potential subjects. By that, you can use a little bit of fuzzy matching so that AI plus real-world data can enable you to look for patients who might not have certain esoteric labs yet, but would be good targets to go and have those labs, and they might be able to be in the trials. We also use it to identify rare disease patients for both trials, the treatments we’ve done that in Fabry disease, abdominal aortic aneurysm, hypertrophic cardiomyopathy, a bunch of diseases. Increasingly for rare diseases, in particular, where it can be difficult or even unethical to have a control arm, real-world data can be collected as historical control arms and compared to treatment arms for regulatory purposes. If you go back 25 years, growth hormone was really approved on the basis of that. At the time, Genentech had been working at growth hormone, and they just compared, I think it’s congenital shortage stature, it was the diagnosis. They were able to compare what they were seeing to what happened to everybody else who wasn’t treating with growth hormone, historically. They’re able to get a very early approval based on real-world data using a registry.

We also increasingly collect data on patients who are enrolled in Phase III studies, meaning real-world data. We get consent to do it, and then we continue to follow them with that real-world data, collect their medical record, collect other data, link that data, and it’s relatively inexpensive to do that, but then you have this companion real-world registry, which is not costly, but can be incredibly valuable down the line if you have additional questions from the agency about representativeness, if you have questions about bias and treatment, if you have questions about long-term safety and so on, you now have this built-in registry that, in some ways, I think that’s going to be the rule of thumb in the future is to use it for that.