The rise of Bayesian statistics in rare disease evidence gathering
Regulators are becoming more open to alternative data sources in the context of rare disease research. Now biotechs need the skills to capitalize on the opportunity, says Brad Carlin, Senior Advisor, Data Science & Statistics, at PharmaLex
At Advanced Therapies Week this month, I will have the privilege of speaking in an important session on the best way to facilitate rare disease research and regulatory evidence gathering, in the company of a senior regulator from FDA and leading patient advocates for two devastating rare diseases.
For a statistician like me, this will be an invaluable opportunity to drill down into what’s possible, and distil the tangible difference we as an industry could make. That’s by incorporating alternative data using novel statistical models, to fill in some of the gaps left when traditional randomized or single-arm studies lack the scope and rigor required for regulatory approval.
The session will look at the evolving role of statistical models in drug approvals linked to rare diseases, and how the regulatory environment is adapting to allow a broader range of data, permitting groundbreaking treatments to be approved more swiftly.
The Bayesian statistical approach, the specific topic I’ll be discussing, assumes that there is already an idea about what’s going on in the specific scientific field, also known as ‘prior information’. This knowledge informs an experiment, facilitating the collection of new data. Those two sources of insight are then combined to form the so-called ‘posterior distribution’ for the unknown quantity of interest (say, the average increase in survival time provided by the treatment).
When graphed, this distribution would show a roughly bell-shaped curve, with its peak corresponding to the most likely value of the unknown quantity. As more and more data are collected, this is pinpointed more clearly as the posterior distribution becomes more tightly concentrated around this quantity.
This accumulation and updating of knowledge, favored by Bayesian statisticians, has particular relevance in rare diseases – where no one can afford to ignore what’s already known, because the inherently small sample sizes preclude us from powering a sufficiently large study to achieve traditional measures of statistical significance.
Breaking new ground
Historically, regulators haven’t supported the mixing of information sources; they have preferred that sponsors collate all the data they need from a single, predesignated clinical trial to determine whether a drug works.
But this rigidity is easing now. Instead of continuing to insist that researchers conduct whatever small study they can and see where this gets them, they are adapting their approaches so that developers can incorporate auxiliary information. This ensures that rare and pediatric disease treatments gain approval more swiftly, while still maintaining patient safety.
That’s even if the prior information comes from less rigorous sources, such as a historical disease registry, often collected by rare disease advocacy groups. In other words, regulators are now offering researchers the opportunity to use data recorded about patients’ ongoing real-world experiences of their condition and any treatments they are taking. Following someone over time and seeing how they do would not previously have been considered sufficient, since patients in such databases were not randomized to treatment or control medications. But in the absence of a strong alternative, regulators see the value of harnessing every scrap of data there is – as long as this is done properly.
Leaning more heavily on real-world evidence in rare disease is becoming increasingly acceptable to regulators for ethical reasons, too. When patient numbers are very small and the diseases are having a profound effect on a person’s lifespan and/or quality of life, a randomized trial in which they might be assigned a placebo ceases to feel right. Indeed, many parents or guardians refuse to allow their children with rare diseases to participate in a study where there is a significant probability the child will receive a placebo.
Exploiting the emerging opportunities
The overriding challenge posed by our panel-based session, to all stakeholders, will be to update their view of what data is viable and acceptable in supporting the approval of a potentially important and life-changing new therapy.
Specifically, we will look at the potential of Bayesian methods for combining diverse information and allowing it to be analyzed in a meaningful and reliable way.
Armed with this knowledge about what’s now possible and increasingly acceptable, life sciences companies then need to know how to move forward with these alternative data and analytic options.
Many pharma players, especially smaller companies, operating in the rare disease space do not have statisticians on their teams. This in turn may be limiting their vision and their room to maneuver, either because they don’t know what’s possible as regulators modernize their own approaches, or because they do not know where to start in exploring new statistical techniques, and alternative data vehicles, that could permit their new therapy to be brought to market sooner.
Innovation in action
In the US, the FDA has come some way in adapting its ‘acceptable evidence’ parameters, with some formal guidance documents now available, and pilot programs underway (such as the complex innovative designs program) to promote the use of more novel statistical models such as Bayesian approaches. While dialogue with FDA regulators is still very much required to gain approval to apply these methods, the move to broaden what is considered scientifically acceptable evidence of rare disease treatment benefit is certainly expanding rapidly.
We’ll be discussing some of the emerging success stories in our session. One area seeing a lot of activity, for instance, is in pediatric extensions – using some portion of adjusted adult data to supplement small pediatric studies, through agreement with the regulator, to get to a more robust evidence base to support pediatric drug approvals. Recent FDA approvals in the rare and pediatric domains include treatments for systemic lupus erythematosus (SLE), transfusion-dependent beta-thalassemia, and Duchenne muscular dystrophy. [i]
Ultimately, the point we’re getting to is the idea of clinical evidence as a continuum, in place of the historic, rigid, ‘randomized trial or bust’ way of thinking which only accepts one very special type of evidence (that, while scientifically rigorous, may have little to do with how drugs are used in the real world).
As well as overcoming the challenge of small trial populations, and addressing ethical issues associated with giving a placebo (or taking too long to approve a drug that’s badly needed), the continuum approach is not as expensive as trying to recruit patients for randomized clinical trials.
Join us to find out more
This promises to be an enlightening and passionate panel session, given what’s at stake and the early signs of new innovative paths now opening up.
For my own part, as someone who normally attends purely scientific meetings, filled primarily with other statisticians, I expect the debate to drive home the real-world impact of the work we do in a profound way. Hearing from patient advocates, whose represent those who are sick and dying right now and thus need solutions as soon as possible, will be very humbling.
Above all it is these patients that need all of us to be more willing to review the boundaries of what is acceptable scientific evidence, so that more life-changing drugs can be approved without delay.
The session, How Do We Best Serve Rare and Small Patient Populations, scheduled for Friday, January 20, 2023 from 11 am to 12:30 pm ET, will be chaired by Christian Chabannon, Head of Center for Cell Therapy, Institut Paoli-Calmettes, Marseille, EBMT & GoCART. I’ll be speaking alongside:
· Wilson Bryan, Director, Office of Tissues and Advanced Therapies, FDA
· Allyson Berent, Chief Science Officer, Foundation for Angelman Syndrome Therapeutics (FAST)
· Sharon Hesterlee, Chief Research Officer, Muscular Dystrophy Association USA
The session will end with a Q&A for all participants with the addition of Tom Whitehead, Co-Founder, Emily Whitehead Foundation.
Phacilitate Advanced Therapies Week will take place at the Miami Beach Convention Center in Miami Beach, Florida, beginning on January 18, when my colleague at Biopharma Excellence will be involved in the opening plenary session at 9am ET.
[i] Garczarek U, Muehlemann N, Richard F, Yajnik P and Russek-Cohen E (2022). Bayesian Strategies in Rare Diseases. To appear in Therapeutic Innovation & Regulatory Science, https://doi.org/10.1007/s43441-022-00485-y.