Regulatory setback or scientific signal? The case for human-relevant evidence

The FDA’s recent rejection of a high-profile gene therapy application from Regenxbio has reignited a debate that has been quietly building across the industry. The therapy had generated considerable optimism, supported by preclinical data and regulatory engagement. Yet, when it came time to demonstrate sufficient evidence for approval, the data was not strong enough to move forward.

This was not an isolated incident. Across gene and cell therapy, we are seeing a widening divergence: some programmes receive expedited designations such as RMAT, while others stall or fail in late-stage development despite promising early signals. The pattern raises an uncomfortable question. Is the bottleneck regulatory friction or is it something deeper in how we define translational readiness?

At its core, this moment forces us to confront a persistent structural problem: our ability to predict human outcomes remains limited. And in complex modalities like gene therapy, those limitations are becoming harder to ignore.

The translational gap: A persistent structural problem

Drug development has always been high risk. Roughly 90% of therapies that enter the clinic ultimately fail. That statistic alone suggests that the issue is not confined to any single modality or regulatory decision. It reflects a systemic challenge in translating biological promise into reliable human efficacy and safety.

For gene therapies, that challenge is amplified. Traditional preclinical development still relies heavily on animal models and simplified in vitro systems to generate evidence of safety and efficacy. However, when it comes to modelling human efficacy, durability, and long-term toxicity, these systems are not just imperfect, but in many cases, they are fundamentally incapable.

There are some mechanisms of human disease that are simply not conserved across species. The pathways active in mice or non-human primates may differ in subtle, but decisive ways from those in humans. We cannot model certain aspects of human biology response because the mechanisms are different. Studying mouse disease does not equate to understanding human disease.

Yet, much of our translational infrastructure is built on precisely that assumption. The result is predictable: therapies that appear robust in preclinical settings enter the clinic with incomplete insight into how they will behave in real patients. When the data falls short, whether on efficacy, safety, or durability, it is often interpreted as a regulatory hurdle. But, in many cases, it reflects uncertainty embedded much earlier in the development process.

And the consequences are significant. Development timelines lengthen. Costs escalate. Regulatory scrutiny intensifies. Investor expectations rise. Most importantly, patients wait. Or worse, they are exposed to risk in trials that lack sufficient predictive grounding.

Towards earlier, human-relevant evidence

Thankfully, the industry is beginning to respond.

There is growing momentum around integrating human biological data earlier in development. Advances in organoids, organ-on-a-chip systems, and ex vivo human tissue models are expanding our ability to study disease in human-relevant contexts. Whole-organ datasets, improved biomarker strategies, and increasingly sophisticated AI-driven discovery platforms are helping translate human derived biological signals into predictive insights.

But a single model or dataset will not solve the translational gap. What is emerging instead is the concept of a “human data stack” – a layered, integrated framework that combines multiple sources of human-derived evidence. Organoid systems. Organ-level functional data. Clinical biomarker patterns. Genomic and phenotypic datasets. Real-world evidence. Computational modelling.

The power doesn’t lie in any one modality, but in integration. To create better therapies, we need not only more human data, but better harmonised human data. We must connect molecular insight to organ-level function and population variability. Precision medicines require precision evidence.

This shift does not eliminate risk. But it can reduce the likelihood of late-stage surprises by anchoring development decisions in data that more closely reflects human biology.

Implications for regulation, investment, and patients

Regulators are not arbitrarily slowing innovation. They have the patients in mind. They are responding to uncertainty. As therapies become more powerful and more permanent, the evidentiary bar rises accordingly. Agencies are increasingly emphasising translational rigour and clearer predictive signals before approving or expanding clinical programmes.

Investors are doing the same. Capital allocators now scrutinise translational risk earlier and more intensely. Programmes built on thin predictive foundations face greater scepticism, particularly in capital-constrained environments.

This convergence of regulatory and investor scrutiny is not a sign of stagnation. It is a signal that the ecosystem is demanding more disciplined innovation. And discipline matters because failure in the clinic carries real consequences. Toxicity-related trial failures have led to patient harm in multiple therapeutic areas over the years. Even when patients are not directly harmed, time lost in failed programmes represents delayed access to effective treatments.

Every clinical trial is built on the promise that the risk is justified by the strength of the evidence supporting it. Strengthening that evidence base is not merely a scientific imperative; it is an ethical one. If more predictive human data can prevent even a single avoidable patient risk, the investment is justified.

What comes next

The recent FDA rejection may ultimately prove less a regulatory setback and more of a turning point. As therapeutic complexity increases, our definition of “clinical readiness” must evolve alongside it. Translational science can no longer rely predominantly on cross-species extrapolation. It must incorporate deeper, earlier, and more integrated human biological insight. It is time to enter the clinic with evidence that meaningfully reflects human biology and patient variability.

Gene therapy remains one of the most promising frontiers in medicine. But promise alone is not enough. If we want more durable clinical wins, we must build development strategies grounded in human-relevant evidence from the start. The real bottleneck may not be regulation. It may be how rigorously we define what it means to be ready for patients.

About the authors

Jenna DiRito, PhD, is co-founder and chief operating officer of Revalia Bio and a world-leading expert in human organ research operations. Throughout her career, she has perfused hundreds of organs to develop a better understanding of human disease. A Yale graduate, she earned her PhD in Surgery from the University of Cambridge and completed postdoctoral training at Yale. She was named to the 2025 Forbes 30 Under 30: Science list.

Janeta Nikolovski, PhD, is chief data innovation officer at Revalia Bio, where she leads strategic data and innovation initiatives to accelerate human-centric biopharma advancements. She brings over 25 years of research and product development experience across pharmaceutical, biotech, medical device, digital, and consumer product sectors, having served in senior scientific and leadership roles at Johnson & Johnson.