Can new approach methodologies improve clinical trial success rates?

More than 90% of drug candidates that succeed in preclinical development ultimately fail in clinical trials. The reasons are complex, but the scale of attrition exposes a fundamental weakness in how new medicines are developed: the industry remains heavily dependent on models that do not adequately reflect human biology.

For nearly 80 years, animal studies have been treated as the gold standard in preclinical drug development. Yet mounting evidence suggests that they often provide a poor representation of the human body, in addition to being costly and ethically contentious.

New approach methodologies (NAMs), including human iPSC-derived cells, organoids, organ-on-chip systems, and computational models, can offer a more human-relevant way to predict drug responses and replace, refine, or reduce the use of animal studies. With regulatory bodies around the world recently embracing NAMs in the drug development process, the pressure for change is growing. But can these approaches outperform the traditional models they are intended to complement or replace?

That question formed the basis of a panel discussion at bit.bio’s The Human Cell Forum 2026, which brought together experts from biopharma, contract research, and academia to discuss the growing use of human cell models in biomedical research and drug discovery. One panel session emphasized why drug discovery must become human-centric and discussed the scientific, regulatory, and practical infrastructure needed to get there.

The translational risk of relying on animal models

Animal models have generated an immense body of data and remain deeply embedded in drug discovery. Their familiarity, however, can obscure their limitations.

“I don’t think we can really overlook our over-reliance on animal models,” says Ross Dobie, Founder of the Centre for Human Specific Research. “For many, many years, we’ve become really good at artificially creating diseases in animals and then treating them and curing them. We’ve seen that this doesn’t always translate – or rarely translates – into humans.”

An animal can provide the complexity of a whole biological system, but it cannot fully reproduce human genetics, physiology, or disease progression. These differences are particularly significant in areas such as central nervous system disorders and bone marrow toxicity, where findings in preclinical species are often poorly predictive of outcomes in patients.

Dobie argues that human drug discovery must therefore begin with human evidence. “If we want to understand human diseases and we want to understand how drugs work, then we need to, as much as possible, focus on the use of human cells, human tissue, and human data.”

One way to do this is through models derived from human induced pluripotent stem cells (iPSCs). These cells can be generated from adult cells and differentiated into disease-relevant cell types, giving researchers access to human biology and genetic diversity at a scale that would otherwise be difficult to achieve.

Introducing human cells, however, is only one part of creating a human-relevant model. Many in vitro workflows still rely on animal-derived biomaterials, including fetal bovine serum, Matrigel, and animal-derived antibodies. Although widely used, these materials can be poorly characterized, vary between batches, and introduce uncertainty into assay performance and experimental outcomes.

Replacing them is not straightforward. “I think scientists are sometimes quite set in their ways and won’t necessarily change something because it means that they have to revalidate everything they do,” explains Dr. Eric Hill of Loughborough University. But he urges researchers to weigh the upfront cost of switching against the hidden cost of irreproducibility.

“One of the benefits of moving away from Matrigel and FBS is that they’re not well characterized, so you get lot-to-lot variation,” he continues. “Think about the cost of buying that bottle of serum but also think about how many experiments you had to repeat because your data didn’t align with itself.”

As models become more biologically and technologically complex, removing avoidable sources of variability becomes increasingly important. “This is another driver for the industry to move towards animal-free, synthetic consumables, because they are more controllable products to use.”

The road to standardizing and validating human-specific models

As the limitations of animal-to-human translation become clearer, human-relevant models are moving from an experimental alternative towards an essential part of preclinical research. Wider adoption, however, will require more than increasingly sophisticated technology.

“We don’t need another 100 new NAMs; we need confidence,” says Dr. Pelin Candarlioglu Deacon, Founder and Director of 3D and 3Rs. “Industry doesn’t buy technology. They buy anything that will save them from uncertainty in a decision. They buy into confidence.”

Building that confidence starts with being clear about what a model is expected to achieve. Rather than asking whether a NAM is simply better than a traditional model, Dr. Deacon argues that researchers must define what “better” means in each application.

“Is it predictability? Is it human relevance? Is it speed, throughput, or cost? All of these could be entirely valid,” she says. “At the end, it comes specifically to the context of use.”

A model may perform well when answering one narrowly defined question but be unsuitable for another. Its validation criteria must therefore reflect its intended context of use: the scientific question it addresses and the development decision its data will support.

This means that one universal validation standard cannot be applied to every system. A liver model designed to assess metabolism, for example, cannot be evaluated in the same way as a blood-brain barrier model, an immune activation system, or a tumor model.

What can be standardized, however, is the process used to evaluate them.

“We can create a framework for defining context of use, generating performance criteria, determining whether a model is fit for purpose and deciding what the final data should be compared against,” Dr. Deacon says. “The global qualification pathway will look the same. However, the evidence package and criteria for each context of use will look different.”

International initiatives are now developing these frameworks for microphysiological systems and organ-on-chip technologies. “It’s ongoing and doable,” Dr. Deacon enthuses.

What can biopharma learn from cosmetics?

Agreeing on a validation framework will not be easy, but biopharma does not have to start from scratch. The cosmetics industry already uses NAMs within animal-free testing strategies to assess toxicity. “If you look to the chemical safety and cosmetics safety spaces, you can learn a lot from what they’ve done and from their frameworks,” says Professor Julie Frearson, Ph.D., Global Executive Lead, Discovery Sciences, IQVIA.

In vitro models are routinely used to investigate endpoints including tissue permeability, immune activation, and genotoxicity. This is not the same as testing oral or injected drugs, but it shows that in vitro systems can generate sufficient evidence to support real safety decisions.

“That should give us confidence that these in vitro models are capable,” Prof. Frearson continues. “The way they’re validated and the way regulators use them can be picked up and used to accelerate the path for biopharma as well.”

Turning human-specific models into the new standard

The future of preclinical research is unlikely to hinge on a single moment when NAMs replace animal models. The transition will be gradual, and in each stage of drug development, the right approach will depend on the question being asked. Human-specific models will form part of a wider weight of evidence, alongside other assays, clinical data, and computational tools.

What matters now is building confidence in where and how they can be used. Animal models benefit from decades of accumulated experience: researchers understand their behavior and regulators are accustomed to interpreting their results, despite their recognized limitations. NAMs are often expected to establish an equivalent level of confidence through formal validation over a much shorter period.

Greater transparency from industry could help accelerate that process. Prof. Frearson calls on pharmaceutical companies to share more about how NAMs are already influencing development decisions. “They are using them for decision-making across the board every day,” she says. “But it’s behind closed doors, and the rest of the world doesn’t see it.”

Funding must also extend beyond developing new technologies to the less visible work of validation, qualification, reproducibility testing, and protocol standardization.

The question is no longer whether human-specific approaches can contribute. In several fields, they already provide information that animal studies cannot. The challenge now is to generate the evidence, standards, consistency, and skills needed to make that contribution routine.