If you are a clinician like me, perhaps when you read a news headline about a new FDA approval for a drug that treats an extremely rare disease you think, “Did I study that in medical school?” If you are a clinician who also works in real-world data (RWD) and outcomes research (okay, there are a lot less of us), you might also wonder how companies figure out who has these diseases and how they are treated in the first place. Here, I will talk about strategies for designing real-world evidence (RWE) studies to enhance the evidence portfolio for orphan drugs, some of the challenges I have encountered in gathering RWE for these conditions, and some potential ways that RWE researchers can address the inherent limitations of RWD in this space. I’ll begin with a working definition of rare (orphan) and very rare (“ultra-orphan”) diseases.
Definitions vary from country to country, but in the United States, the FDA defines a rare disease as a condition that affects up to 200,000 people [1]. The FDA estimates that there are 7,000 such diseases. Is there an ultra-rare designation? The FDA doesn’t make further distinctions beyond the definition above. Some industry experts say that ultra-rare diseases should affect no more than 20 persons in a population of 1 million, translating to less than 6,000 patients in the United States.*
The FDA’s Orphan Drug act has been a boon to patients who suffer from rare diseases. Since the Act was passed in 1983, nearly 5,100 orphan drug designations representing nearly 3,300 unique products have been approved through this process [2]. Orphan drugs comprise a large fraction of FDA’s drug review portfolio. In 2022 alone, 20 of the Center for Drug Evaluation and Research’s (CDER) 37 novel drug approvals (54%) were approved to treat rare or “orphan” diseases [3].
Of course, all those new drugs must then be reviewed by commercial and public insurers as part of their coverage and reimbursement policies. This is a critical step for patients, because drugs for rare diseases are generally very expensive and essentially unobtainable without insurance. Manufacturers typically include evidence packages supporting their submissions. I’m seeing more evidence packages for orphan drugs that include RWE alongside the FDA-required clinical trial data. This isn’t surprising, since the FDA applies different evidence requirements for drugs that treat rare diseases, typically allowing small, uncontrolled studies, in stark contrast to what they would require for drugs to treat more common diseases. The more modest evidence requirements pose challenges for insurers that, in theory, would like to apply the same criteria to every drug they review. To reduce the likelihood of restrictive coverage decisions, manufacturers can try to fill the evidence “holes” with real world studies.
This is welcome news for those of us who work in RWE. Unfortunately, our field faces many of the same problems that drugmakers face for rare diseases. In our case, it’s finding data on patients who meet the clinical criteria for the approved orphan drug. In some ways, I would argue that our task is harder, because we can’t go directly to doctors and clinics that treat patients with rare diseases. We must figure out other ways to find them.
“The answer doesn’t always lie in finding the orphan needle in a haystack, but in defining what the haystack is in the first place”
Let’s start with some basics: you need to find patients with the disease, hopefully more than were in the FDA trial. The RWD pools that we readily access to find people with rare conditions include insurance claims, electronic health records (EHR), and disease registries. Apart from public disease registries, it’s not usually feasible to comparison shop by asking multiple data vendors to tell you how many people with rare condition X they have in their database. (They might do that for a big fee, but there are a lot of vendors out there. How much money do you have?) So perhaps you chose to go with a vendor that has a huge number of individuals, playing the small numbers probability game. Problem solved? Well, no.
Depending on the condition that you are focusing on, the database may or may not have enough information to accurately locate that type of patient. Let’s start with insurance claims. The only way to identify patients with a disease is to use international classification of disease (ICD) codes. The newest ICD-10-CM system has codes for 68,000 conditions. The chance that there is a code for your rare disease is higher if it has a name, ideally that of a person who first characterized it about a century ago (Examples: Fabry disease [E75.21] and Rett Syndrome [F84.2]). You will not be so lucky if you are interested in one of the newer genetically characterized diseases or disease subtypes. For example, ICD codes will not help you find ROS1 subtype non-small cell lung cancer, even though there are two FDA approved drugs for this condition, with more on the way.* Perhaps you could search for the use of an existing drug that specifically treats the condition, but this will only identify a subset of those who are eligible, because not every eligible patient will get treated. Finally, even if it’s possible to accurately identify the condition, drugs have specific indications within the disease that are often difficult or impossible to identify with claims. To get that level of specificity would require looking at the medical record. There are many issues with using EHR’s to identify patients with a rare disease that I won’t dwell on here, because the show-stopper is still that the database must include enough people to give you a chance of finding people with the disease. Because most EHR vendors represent regional providers or provider networks, there may not be enough people even to get started! You can ask but I refer you to my previous thoughts about asking vendors to find patients for you. The problem is particularly severe for ultra-rare conditions that primarily affect younger people, since they are likely to be geographically scattered among providers and insurers.
So, is building an RWE portfolio for patients with rare diseases even a solvable problem? I will close on a hopeful note by suggesting that the answer doesn’t always lie in finding the orphan needle in a haystack, but in defining what the haystack is in the first place. First, it’s important to understand that trials of new treatments for rare conditions typically target subpopulations based on characteristics favorable to the new agent (in consultation with the FDA). In most cases, existing treatments for that subpopulation are not different than the larger population. For example, trials of new gene therapies for sickle cell anemia consider a subset of patients who are, in general, managed exactly the same way as patients who don’t meet trial eligibility criteria. In these cases, studying the larger population, perhaps restricting to a more similar subgroup where data are available (eg, age, comorbidity) is a reasonable way to form a comparison group for patterns of care studies. In other words, the haystack is much easier to define and often much larger than what you might think you might need to find persons who are eligible for the orphan drug (as defined by its label). One can use a similar approach for estimating outcomes: if the factors that drive outcome in the usual care group are identifiable and not necessarily related to the selection criteria for the trial, then the general RWD population can be stratified to reasonably approximate the trial. Getting back to the audience for these studies (insurers), this general approach will give a reasonable and much better approximation of the incremental benefits and risks of the new drug than they would have by only looking at small, uncontrolled trial populations.
In summary, generating RWE to support evidence portfolios for rare diseases is an important but unavoidably imprecise task. While this may discourage some companies and analysts from going down this path, I would argue that the alternative — generating no evidence beyond orphan drug trials, which are necessarily less rigorous—puts all decisionmakers at a disadvantage.
*I will refer to rare versus “orphan” diseases. The term orphan became popular in industry after passage of the 1983 Orphan Drug Act, which gave FDA a mandate to encourage the development of drugs for rare conditions.
**Shocking to me is that ICD codes still do not distinguish between small cell and non-small cell lung cancer, even though those designations have been around for decades, and the treatments are very different.
–Scott Ramsey, MD, PhD
Senior Partner and Chief Medical Officer, Curta
ACKNOWLEGEMENTS
Curta team member Elizabeth Brouwer, MPH, PhD contributed to this article.
REFERENCES
- STAT. No magic bullet: For drugmakers and the FDA, clinical trials on ultra-rare diseases pose thorny challenges. https://www.statnews.com/pharmalot/-2022/07/26/ultra-rare-disease-drugs-fda-clinical-trials/
- Miller KL, Kraft S, Ipe A, Fermaglich L. Drugs and biologics receiving FDA orphan drug designation: an analysis of the most frequently designated products and their repositioning strategies. Expert Opin Orphan Drugs. 2022;9(11-12):265-272. doi:10.1080/21678707.2021.2047021.
- US FDA. New Drug Therapy Approvals 2022. https://www.fda.gov/drugs/new-drugs-fda-cders-new-molecular-entities-and-new-therapeutic-biological-products/new-drug-therapy-approvals-2022#