Bringing the Bayesian Method to Clinical Trials: Q&A with Dr. Stacy Lindborg

In January of this year, FDA issued draft guidance facilitating that use of Bayesian methodologies in clinical trials for drugs and biologics.¹ Using this method, researchers are allowed to combine data collected from a study with relevant existing data.

In the announcement, FDA Commissioner Mackary said, “Bayesian methodologies help address two of the biggest problems of drug development: high costs and long timelines. Providing clarity around modern statistical methods will help sponsors bring more cures and meaningful treatments to patients faster and more affordably.”

Pharmaceutical Executive spoke with Dr. Stacy Lindborg, CEO of Imunon, about the significance of this change. According to her, this draft guidance has the potential to increase innovation and how this method impacted the development of Imunon’s advanced ovarian cancer therapy IL-12.

Pharmaceutical Executive: Can you describe the Bayesian method?
Dr. Stacy Lindborg: Unlike frequentist methods, which base inferences exclusively on the data generated within the current trial, Bayesian approaches in clinical trials treat probability as a formal measure of belief or uncertainty about patient outcomes. They begin with a prior distribution that represents our best initial assessment of the treatment effect, informed by historical data, earlier studies, or external evidence. As new trial data are collected, this prior is formally updated via Bayes’ theorem to produce a posterior distribution that reflects our revised beliefs about an investigational therapy.

This updating process can, in certain situations, support more flexible trial designs—for example, adapting protocols based on interim results, borrowing information from external sources when justified, or making decisions with smaller sample sizes when appropriate. In essence, it encourages us to refine our understanding continuously as evidence accumulates, rather than waiting until the end of a study to perform a single, final analysis.

The methodology is conceptually similar to how we naturally process information in everyday life—interpreting new observations in light of what we already know—only here it is done with explicit, mathematically rigorous rules. I first encountered Bayes’ theorem as a graduate student while completing my Ph.D. in statistics, and I was surprised, early in my industry career, to see how slowly the drug-development world adopted these ideas. The barriers were real and well-founded: limited familiarity among many statisticians, clinicians, and regulators; legitimate concerns about the subjectivity of priors and the risk of undue influence; the computational demands of thorough simulation work; and a strong institutional preference for the long-established frequentist paradigm that provides clear control of type I error rates.

That said, it has been encouraging to see growing regulatory interest over time—including the FDA’s January 2026 draft guidance on Bayesian methods in drug and biologic trials. Where the science and the specific development context align, these approaches have the potential to improve efficiency and help address some of the rising costs and timelines that challenge our industry. The key, in my view, is always rigorous pre-specification, transparent justification, and careful evaluation of operating characteristics—never treating Bayesian methods as a default or universal solution, but as one additional tool in the statistician’s toolkit when they genuinely add value.

PE: How does using this method impact innovation?
Lindborg: Bayesian methods have the potential to support greater innovation in drug development by enabling continuous learning and more agile decision-making under uncertainty. Rather than committing to a completely rigid protocol and deferring all major analysis until the end of a trial, these approaches allow us to update our understanding of a treatment effect dynamically as data emerge — formally incorporating prior information from earlier studies or real-world evidence when that borrowing is scientifically justified.

This framework can open the door to adaptive trial designs that permit pre-specified mid-course adjustments—such as optimizing doses, dropping arms that are not performing, enriching for responsive subgroups, or stopping early for efficacy or futility. The Bayesian framework also allows us to calculate predictive probabilities directly from the posterior distribution, enabling earlier and more informed assessments of futility or success while the trial is still ongoing. In fields like oncology, where patient populations are often small and heterogeneous, this flexibility can, in carefully selected settings, help teams identify promising signals more efficiently while reducing unnecessary patient exposure to treatments that are unlikely to succeed.

When implemented with the necessary rigor—strong pre-specification, extensive simulations, and early regulatory alignment—these methods can contribute to a broader evolution toward more adaptive and learning-oriented clinical development.

PE: Can you discuss your experience leading through uncertainty?
Lindborg: Leading through uncertainty is one of the defining challenges of biopharmaceutical leadership. In drug development, we constantly face heterogeneous patient responses, rapidly evolving science, shifting regulatory expectations, and the possibility that today’s promising signal may not hold. I have found Bayesian thinking to be a useful practical framework for managing this reality: it treats our beliefs as probability distributions that can be quantified and formally updated as new evidence emerges.

At IMUNON, we applied this mindset during the development of our immunotherapy IMNN-001 for advanced ovarian cancer. As we read out the Phase 2 trial and designed the Phase 3 OVATION 3 study — as is typical when advancing a novel therapeutic — we faced layered uncertainties around the optimal size of the trial, the magnitude of effect we might observe in a future trial in the intent-to-treat population and key subgroups, and whether to formally incorporate historical data versus relying solely on our own accumulating results. Given the long-standing lack of progress in overall survival in front-line treatment for newly diagnosed ovarian cancer, this setting initially appeared well-suited for borrowing from historical trials—particularly those in which the standard-of-care arm had failed to show a treatment effect. Such borrowing had the potential to either reduce the overall sample size (thereby accelerating the trial timeline) or to increase the number of patients assigned to our novel investigational therapy rather than standard of care—a more patient-centric design. However, advances in later-line therapies and maintenance regimens over time meant that even trials with similar inclusion criteria could differ in important ways from our current study. We also evaluated borrowing from our own Phase 2 data (112 patients) to supplement the planned Phase 3 cohort of approximately 500 patients—an approach the FDA guidance itself explicitly highlights as promising. The expected gain, however, was modest—roughly a couple of percentage points in statistical power—and did not justify the additional regulatory and operational effort required. After thorough evaluation, we concluded that borrowing in this setting did not pass the bar.

We addressed the remaining uncertainties by running extensive clinical trial simulations across hundreds of scenarios, testing weak, moderate, and strong treatment-effect profiles. The simulations guided our final design decisions, informed the timing of interim analyses, and clarified the probability of meeting clinically meaningful thresholds—including the potential for early stopping and an immediate BLA filing if the data warranted it.

Bayesian posterior predictive probabilities have also proven valuable for patient protection. Our Chief Medical Officer began formally using Bayesian methods years ago for continuous safety monitoring in trials he designs. This allows us to generate early alerts and potentially stop or modify a study before a safety threshold is actually breached. The FDA was very enthusiastic about the approach and has since recommended it to other trialists.

For me, leading through uncertainty requires intellectual humility, disciplined analytics, and strategic agility. It means building teams that can quantify risk, run transparent simulations, maintain open dialogue with regulators, and pivot when the data warrant it—all while upholding the highest standards of scientific integrity and patient safety.

PE: How does analytical leadership shape market vision?
Lindborg: Analytical leadership is about turning complex scientific data into a realistic, actionable view of commercial potential. I approach this by starting with the best available evidence — preclinical data, earlier clinical results, competitor landscapes, and unmet patient needs — and then continuously updating those assessments as new information emerges. This disciplined process yields more dynamic and precise forecasting than static assumptions ever could.

A key part of the work is aligning the clinical data and planned registration trials against the target product profile (TPP). Using a Bayesian framework, we can directly calculate the likelihood of achieving the target or maximum target profile, quantify the impact of the areas where we have the greatest uncertainty, and translate those probabilities into realistic financial projections for the product.

For IMUNON, when considering the performance of other therapeutics for ovarian cancer, the urgent need for novel treatment approaches, and the current oncology treatment market, our current predictions indicate that our novel DNA-based immunotherapy IMNN-001 represents a multi-billion-dollar market opportunity. Demonstrated by our therapy’s unique non-viral DNA-based delivery system, which has resulted in a favorable safety profile, and its ability to result in a median 14.7-month increase in overall survival when combined with standard of care chemotherapy, IMNN-001 has the potential to not only redefine treatment expectations for women with advanced ovarian cancer and address their unmet needs but also to establish a new commercial benchmark for next-generation immunotherapies in oncology overall.

PE: What caused FDA to update its guidance to formally recognize Bayesian statistics?
Lindborg: The FDA’s January 2026 draft guidance on the “Use of Bayesian Methodology in Clinical Trials of Drug and Biological Products” is the direct result of a deliberate, multi-year effort by the agency to modernize clinical-trial frameworks and provide sponsors with clearer pathways for more flexible designs, particularly in areas such as oncology and rare diseases where patient populations are often small and heterogeneous.

This is not a sudden change in philosophy. The Center for Devices and Radiological Health (CDRH) has operated under its own Bayesian guidance since 2010, and the new document essentially extends similar principles to the drug and biologic space under CDER. Advances in computing power have also removed many of the practical barriers that once made large-scale Bayesian simulations difficult, while well-documented adaptive trials have demonstrated that these methods can be implemented rigorously.

Importantly, the guidance does not position Bayesian approaches as superior to or a replacement for traditional frequentist methods, which remain the cornerstone of regulatory decision-making. Instead, it offers sponsors a structured, transparent framework for when and how to incorporate prior knowledge, borrow external data, or use real-time probabilistic updating—provided the design is fully pre-specified, thoroughly simulated, and shown to maintain scientific integrity.

From my perspective as CEO of IMUNON with formal training in statistics, this represents pragmatic regulatory evolution rather than a blanket endorsement of one statistical paradigm. We have evaluated Bayesian tools in our own IMNN-001 program and continue to use them judiciously alongside frequentist methods. In the end, the guidance simply gives development teams one additional, well-defined option when the specific scientific and operational context makes it the right tool for the job.

Sources

1. FDA Issues Guidance on Modernizing Statistical Methods for Clinical Trials. FDA. January 12, 2026. Accessed April 2, 2026. https://www.fda.gov/news-events/press-announcements/fda-issues-guidance-modernizing-statistical-methods-clinical-trials