Patient Registries Now Fueling Regulatory Decision-Making

By Deborah Borfitz 

February 23, 2023 | The challenges and opportunities of working with patient registries, one of the most frequently used real-world data (RWD) types supporting pharmaceutical drug development, was the subject of a series of use cases presented at the recent Summit for Clinical Ops Executives (SCOPE). From UBC came examples of registries enriched with integrated RWD from other “anywhere, anytime” sources—including electronic health records (EHRs), claims, patient-reported outcomes (PROs), wearables, and federated registries—and from Regeneron the contrasting experience of working with the U.S. Food and Drug Administration (FDA) and the European Medicines Agency (EMA) in using registries to support safety and effectiveness evaluations. 

The UBC use cases illustrate that integrated RWD reduces burden and creates value for patients, healthcare providers, payers, and pharma companies, according to Ying Tabak, Ph.D., vice president of global real-world evidence (RWE), health economics and outcomes research (HEOR), and epidemiology. She began by describing a universal rare disease registry, representing 1,849 unique cases of a hematological disease, created by linking and deduplicating four existing registries. 

Variables in the universal registry include patient demographics, medical histories, medications, pregnancies, healthcare visits, lab results, and adverse events, Tabak says. Key challenges in creating a unified database were ensuring patients’ data were represented only once and addressing differences in the underlying coding structure and data completeness. “We needed a common data model to translate raw data into standardized data across all registries.” 

The next step was to enrich the universal registry with additional RWD provided by marketplace vendor HealthVerity, enabling an extended view of patients’ journey prior to enrollment or diagnosis and after enrollment and further long-term follow-up, she continues. The exercise also helped fill in missing data.  

Although hematological laboratory test results were represented in the combined universal registry at or above 90%, after linkage with the HealthVerity, medical data completeness was further enhanced by a couple of percentage points. This was critical, says Tabak, since those measures are essential to understanding disease progression and outcomes.  

Linked and enriched RWD expands the scope of the universal registry by enabling researchers to look back and see how the disease progressed and was diagnosed, as well as any prior misdiagnoses, she says. It also opens a view to disease treatment, patterns, and outcomes. The key processes involved in combining multiple registries are identity management, data harmonization, and linking and enriching the universal registry with secondary RWD.  

Federated Model 

The second UBC use case is an ongoing project that links registries for an international consortium around a gene therapy specific to a rare neurological disease, Tabak continues. As defined in the United States, a rare disease is a condition affecting fewer than 200,000 patients and the medical community believes that 80% of them have a genetic component.  

On the other hand, she quickly adds, “gene therapies are so new the long-term clinical effectiveness and safety is unknown.” This creates the need for following up on patients for 10 to 15 years, based on the follow-up period seen with some recently approved gene therapies. The challenge here is the burden of long-term follow-up on patients, healthcare providers, and sponsors, as well as the fact that low-prevalence diseases require sharing data for further phenotyping and evidence generation. 

Among the publicly funded platforms for connecting databases for genomic and clinical research is RD-Connect, Tabak cites as an example. RD-Connect is a global infrastructure project initiated in 2012, and finally established six years later, funded by the EU under the 27-partner International Rare Diseases Research Consortium. It is enabling development of platforms (three thus far) to host and analyze genomic and clinical data, and serves to connect databases, registries, biobanks, and bioinformatics—and to delineate biomarkers and identify disease modifier genes and therapeutic targets. 

On the private sector side, UBC is currently involved in developing a federated model to solve the data fragmentation problem and empower researchers to conduct robust RWD research, says Tabak. For a rare genetic neurological disease where UBC manages a registry with just under 500 patients, the “proprietary statistical representation” technology of SymetryML is being used to grow a larger, centralized database. Data on about 175 patients with the same disease from registries in Germany and France, together with another 25 in UBC’s own centers of excellence, are being brought together “without data leaving the data owner’s site,” she explains.   

Study Pillars 

Reimagined registries enriched with RWD has enabled the continued integration of science, operations, and technology and thus the “breadth and depth of RWE generation,” says Tabak. They are also improving clinical decision support and next-best actions and allowing refinements and innovations in study design and analytical approaches.  

UBC embraces three pillars for real-world HEOR/epidemiology studies, she says, the first of which is an evidence generation strategy devised by a multidisciplinary team comprised of experts in RWE, HEOR, epidemiology, clinical science, and data science. “The value proposition of this strategy is market access, reimbursement, label expansion, regulatory and HTA [health technology assessment] submissions, and publications.” 

The second pillar is the use of purpose-built technologies, which at UBC includes the MOSAIC open technology platform that can integrate with any interoperable software, data exchange partner, and technology, Tabak continues. Among the long list of data exchange partners beyond HealthVerity are Medidata, Cerner, Cure SMA, World Federation of Hemophilia, the Human Growth Foundation, TREAT-NMD, OPTUM Care, THREAD (decentralized clinical trials platform), GE Centricity, IBM, Truven, and Allscripts.  

HealthVerity, she notes, has a large database encompassing over 300 million patients and 50 billion transactions. It allows individual patients to be tokenized with a unique ID that remains constant across data sources, enabling interoperability and an enriched longitudinal view of patient journeys.  

Tabak’s takeaway message is that universal registries enriched with RWD play an important role in developing RWE that is “publishable and regulatory grade,” enabling sponsors to study disease progression as well as the safety and effectiveness of their medical products. Such registries also lessen the burden on stakeholders and are more powerful than a single registry on its own—especially for rare disease research. 

Nuanced Guidance 

A second presentation by Rachel E. Sobel, Ph.D., head of pharmacoepidemiology at Regeneron Pharmaceuticals, highlights one of the “repeating themes” she has personally experienced as well as read in recent guidances issued by both the EMA and FDA. “It is extraordinarily important to clearly specify your scientific objectives, [as] that will help you define and drive virtually every other scientific and operational decision [thereafter].”  

RWD—most often administrative claims, EHRs, and registries—has been used for regulatory decisions in many ways, she says. Registries, the focus of her talk, are “nothing more than an organized system that collects data on people” that in the biopharmaceutical space are typically based on having a disease or condition or exposure to a drug product.  

The FDA has been busy over the last two years issuing guidances on RWD and RWE, driven by passage of the 21st Century Cures Act back in 2016, says Sobel. Although both have been used in safety epidemiology and safety regulatory decisions for decades, the focus here is how the agency can use RWD to allow a label change or grant a new drug approval. 

In “Considerations for Using Registries to Support Regulatory Decision-Making,” the FDA puts a lot of emphasis on relevance and reliability of registry data to the question at hand, she points out. A fair amount of attention is also paid to linking registries either to other registries or to other external sources such as claims, EHR, and wearables to make a database “more robust or more valid.”  

One of her key observations about the guidance document is that the FDA defines registries and registry data “a bit differently” than the EMA, which issued its guidance just a month prior, Sobel says. This could have major implications for companies in the orphan drug space looking for a globally harmonized regulatory decision where there might be only one registry for a disease on the entire planet.  

Overall, a lot of nuanced situational language (e.g., “it depends”) is employed regarding use of registry data to support the evidentiary standards for regulatory decision-making, she continues. But the FDA unequivocally expects protocols and statistical analysis plans to be submitted before the start of a study and that all key decisions be pre-specified. The agency wants to ensure everything is time-stamped and decisions are documented “before you get into the data.” 

Another big takeaway from the guidance is a “clear expectation” that patient-level data used to support a regulatory decision be auditable and submitted to the FDA, says Sobel. “Many registries that we might work with are independent, whether they’re based at an academic medical center, or run by a patient advocacy organization or other scientific society, and the FDA now expects that sponsors have some sort of third-party agreement that allows these entities to be inspected by the FDA.” This will require a lot of “education and hand-holding,” she adds, “and may not make certain registries viable.” 

EMA’s Approach 

One nice feature of EMA guidance is that it draws a clear distinction between a registry and a registry-based study, says Sobel. Like the FDA guidance, it also includes a lot of details about the required rigor and careful thinking needed when linking data sources and defining study populations.  

The EMA guidance, like that of the FDA, also stresses the importance of prioritizing the research plan and protocol and submitting that in a publicly available repository such as the EU PAS (post-authorization studies) Register. But the EMA’s document “takes a more global view in that it talks about some of the international guidances and entities [e.g., EMA’s Committee for Medicinal Products for Human Use, U.S. Agency for Healthcare Research and Quality, and European Reference Networks] that it consulted before constructing the guidance.” It also recognizes the existence of both interventional and non-interventional registries.  

The EMA’s approach to constructing the guidance document contrasts sharply with that of the FDA, Sobel continues. It was a multi-year process involving a series of public workshops and collaboration with the European Network of Centers for Pharmacoepidemiology and Pharmacovigilance. 

“One of the other big distinguishing factors from the FDA is that the EMA actually allows for qualification of a registry for regulatory use,” says Sobel, adding that two registries had been so qualified as of late 2022. Qualification means the EMA has reviewed the quality of the registry and its governance and determined that it is of “regulatory grade” and in theory could be used to support a regulatory decision.  

Registry guidance appears to be based off learnings that the EMA has had for decades with other types of studies using registries, including post-authorization safety and efficacy studies, she adds. This shows up in language around the importance of data quality and integrity, and potential sources of bias and confounding.  

Challenges To Overcome 

A primary reason that registries fail or are unsuccessful when applied to regulatory decision-making is either because scientific objectives were not well specified or the registry was not optimal for supporting research questions, says Sobel. Specifying the scientific objectives up front “helps drive what is ideal versus the actual study design... you can do.” 

The main new decision that needs to be made is whether to use a disease or exposure (product) registry, she continues. Beyond that, the choices are the same as when designing any other trial—notably, what’s the study population and how generalizable is it to the broader target population, the sample size, the feasibility (notably with rare diseases) of achieving the desired statistical power and plans around recruitment and retention and loss to follow up. 

Operational challenges are the most important set of considerations for a registry trial because they can prevent even well-specified scientific objectives from being achieved, says Sobel. Among the issues to be addressed are how many patients are needed and how willing they are to participate long term. No matter how “enthusiastic and genuine” study investigators may be, “their projections are almost never able to match what they are actually able to do.” 

Much of that has to do with a mismatch between data collection methods and the study population—e.g., a smartphone-based data capture system for a disease that tends to affect the elderly. “I also believe that... keeping it simple is really important,” Sobel says. “Only collect the variables that you absolutely need to address that research question.” 

In general, she says, multi-country, multi-sponsor registries make a study “more generalizable.” It also introduces its own set of operational challenges due to cultural, legal, and ethical approval process differences and the need for language translations.  

In terms of the environmental challenges, Sobel specifically mentions that changing standards of care, clinical guidelines, regulatory actions, and insurance reimbursement policies can change who gets into a registry or how long they stay. Another major consideration with rare disease and pregnancy registries is that studies may well be competing for the same patients and investigators.  

Three Case Studies 

Sobel presented one case study where RWE was used to support both a post-marketing safety and effectiveness assessment for an antibody treatment (Evinacumab) for a rare disease known as homozygous familial hypercholesterolemia (HoFH). The condition afflicts less than 600,000 people on earth and is typically diagnosed in teenagers and young adults. 

Understandably, Regeneron’s clinical trial program consisted of a single clinical trial that was fairly short and small, says Sobel, so the EMA reasonably requested a post-approval commitment to conduct another clinical trial to establish its long-term safety and effectiveness. “We quickly realized this wouldn’t be feasible... from an ethics and IRB standpoint, as well as [that of] investigator willingness and patient willingness.” 

Since the drug was going to be available on the market, patient interest in an interventional trial was expected to be low, she explains, particularly since the patient pool was so small. Development of a long-term registry was next considered, until it was discovered that a well-established, multi-country registry of this type already existed. 

That familial hypercholesterolemia registry, run out of the UK by the European Atherosclerosis Society, has the participation of more than 64 countries. “We realized we’d be competing with them for the same patients and investigators and in fact they already collect... the vast majority of the information that we would need to support the EMA’s concerns,” says Sobel. Regeneron is now collaborating with the registry on the post-approval commitment tied to the 2021 approval of Evinacumab.  

For the same drug, the FDA “did not express explicit concerns about the long-term safety and effectiveness... [but it was] concerned because patients of childbearing potential would likely be exposed,” Sobel says. The agency originally asked us for a product registry specific to pregnant patients, “but we realized again that it would essentially be infeasible since so few patients would likely be pregnant and exposed to this drug.” 

Regeneron went back to the FDA and successfully proposed that it instead do “enhanced pharmacovigilance” using formal follow-up questionnaires that would be collected at multiple points during a patient’s pregnancy and post-partum. “[We] essentially nested a study where we have a formal protocol using [real-world] data from the pharmacovigilance database,” she says. 

Sobel’s third example was a case study the FDA frequently uses about the recent approval of Prograf (Tacrolimus), made by Astellas, for preventing lung transplant rejection. Prograf was previously approved to prevent organ rejection in liver, kidney, and heart transplants, and was being used off-label for lung transplants. 

What’s notable here is that the U.S. maintains a national transplant registry that includes every person in the country who has had a transplant, and the registry data links to the Social Security Death Master File to enable complete capture of all deaths. Based on information in the registry link, Astellas succeeded in getting the lung transplant indication approved by the agency.