The FDA issued a new draft guidance on the “Use of Bayesian Methodology in Clinical Trials of Drugs and Biologics”, paving the way for the Bayesian modelling pioneered by Concr to directly support primary evidence for safety and effectiveness
The guidance explicitly endorses Bayesian methods as tools for inference, decision-making, and evidence synthesis in settings where uncertainty, limited sample sizes, or external data sources are unavoidable. It also signals a shift away from rigid, data-heavy frequentist models toward the flexible, probability-based approaches that underpin Concr’s proprietary FarrSight® platform.
The document specifically highlights the utility of Bayesian methods for paediatric populations and rare indications, where recruiting large control arms is often impossible. FarrSight® is uniquely positioned to practically implement this new framework:
Bayesian is favoured where traditional trials are constrained
The FDA’s thinking reflects Bayesian methods as suitable in situations such as small or specialised populations, oncology studies, and cases where evidence must be combined from prior studies or external data. These approaches are described as established tools that can be used when conventional trial designs alone may be impractical or inefficient.
Accounting for uncertainty is essential for interpretation
The guidance emphasises clearly communicating uncertainty, including probability-based outputs and sensitivity to assumptions. The FDA highlights the importance of understanding how conclusions are drawn and changed under different assumptions, settings, and with further data.
Application across development phases
Bayesian methods are increasingly being integrated across drug development stages. Translational teams are recognising that early predictive modelling directly shapes development timelines, probability of success, interpretability, and future regulatory outcomes. The FDA’s draft guidance aligns with this broader shift toward treating Bayesian thinking as a connective layer across the development chain.
Dr Matthew Griffiths, Concr’s CTO and Co-founder, commented: “What we’ve been seeing in practice is that sponsors are increasingly interested in using new AI methods well before trials are run – to select assets, expand indications, explore designs, test assumptions, and interpret emerging signals under uncertainty.”
A leading expert in Bayesian and advanced computational modelling, Dr Griffiths added: “At Concr, we’ve been long-standing champions of Bayesian methods because they provide a transparent framework; one that directly addresses uncertainty and explains biological drivers rather than relying on opaque, “black-box” predictions.
A core part of this is inferring across fragmented, imperfect data, allowing valuable information to be used that would otherwise be discarded. We believe this approach is critical because the outputs from our Bayesian models can be used to simulate different choices of trial design and the potential impact of analysis choices across the full chain, so uncertainty is handled consistently rather than addressed late.”
Why timing is now
With over half of Phase II oncology failures driven by inadequate efficacy, the financial consequences of getting development decisions wrong have never been higher.
For biotechs, where programmes are often driven by a small number of assets and limited capital, late-stage failure can be existential, erasing years of value in a single trial readout.
And for pharma, as the industry faces a patent cliff threatening over $200 billion in revenue, the ability to de-risk assets in silico using FDA-aligned methodologies is no longer a luxury, it is an economic necessity.
Get in touch to assess your trial’s feasibility with FarrSight®.
Click here to access the full FDA draft guidance report.
Back
LinkTree