Why toxicology is still the toughest test for NAM adoption

Toxicology remains the most challenging field for adopting new approach methodologies (NAMs) as it requires predicting systemic, long-term human health effects that are inherently complex to replicate outside a living organism. While NAMs offer human-relevant data, the industry faces significant hurdles in validating these methods to the same level of trust as traditional animal models.

DDN spoke with Justin Boyd, Product Manager at Sartorius, to explore how NAMs are being applied in practice across drug discovery and safety assessment, and what ultimately determines whether they transition from scientifically compelling tools into routine components of toxicology workflows.

You’ve spent much of your career building biologically relevant cellular models of disease. How does that emphasis on relevance shape how you think about NAMs in toxicology, compared with more traditional animal-based approaches?

I recently joined the vendor side of NAMs. For nearly two decades before that, as a drug hunter, I was less focused on building models and more on applying them. In that context, I thought of NAMs as fit-for-purpose tools to rapidly explore the effects of experimental drugs on the proximal human biology I care about.

Now, as Product Manager of a NAMs portfolio, I still strongly believe in that utility. The strengths of NAMs lie in: (1) conservation of human biology, (2) speed to data-driven decision-making, and (3) cost to execute study. That said, I don’t see NAMs as replacing the value of a whole organism — whether mouse, rat, or non-human primate. A preclinical toxicity study in animals provides a more comprehensive view of how a compound behaves in the context of an intact organism, including systemic interactions that are still not well captured in vitro.

However, NAMs create an opportunity to rank and/or differentiate compounds with higher molecular resolution while remaining “in human.” That kind of insight can meaningfully inform decisions about which compounds are worth advancing into more expensive and time-consuming animal studies.

Ultimately, I think of NAMs for toxicity as key complementary models for evaluating tissue-specific risk to drive decision to go into the animal models, leading to better stewardship of resources for drug discovery and animal welfare.

NAMs are often discussed as ethical or regulatory advances, but from your perspective, where do they most clearly outperform legacy toxicology methods scientifically?

With respect to performance, there are two clear areas where NAMs excel. First, NAMs can recapitulate aspects of human biology more faithfully than preclinical species. This becomes especially important when studying the proximal biology engaged by an experimental drug, where species differences can significantly limit interpretability.

Second, NAMs substantially reduce the time and cost required to reach a decision. From a project or program management perspective, the ability to make informed and confident stage-gate decisions is where the highest value lies. In this context, NAMs enable a more expedient and cost-effective approach to predicting toxicity in the pre-Investigational New Drug (IND) to IND space.

Although, it’s likely that animals will be used at this point, NAMs can and should be deployed to derisk the Good Laboratory Practice (GLP) toxicity studies in animals and potentially reduce the numbers of cohorts and time for treatments.

Many toxicology assays still rely on relatively reductionist systems. How close are we to NAMs that genuinely capture the complexity of chronic diseases like Alzheimer’s or Parkinson’s when it comes to assessing safety?

I think this is a tricky question, and I would start by noting that the complexity of Alzheimer’s (AD) and Parkinson’s disease (PD) pathobiology is part of what limits our ability to clearly distinguish mechanisms that cause disease from those that simply exacerbate progression. As such, “who, when, and how” these diseases are treated and the potential toxicity from treatment remain controversial.

In some cases, NAMs, particularly complex in vitro models with multiple cell types and structures, can recapitulate complex non-cell autonomous biology, such as the impact of inflammation on neuronal health. Moreover, computation-based NAM tools can help predict the trajectory of biology and stratify at-risk populations for toxicity outcomes.

So, when asking how close we are to NAMs that genuinely capture the complexity of chronic diseases like AD and PD, I would say they are, in many ways, as close to recapitulating that complexity as our current understanding allows us to define it.

Drug-induced nephrotoxicity remains a major clinical challenge. From your experience working with human kidney microtissues, why has traditional animal toxicology struggled to predict renal risk in humans?

It sounds cliché, but animals are not humans. In the case of the kidney, there are two key drivers of translational gaps.

First, the expression of key kidney genes and their protein products — particularly those governing transport and metabolism — differs significantly between preclinical species and humans. Second, baseline renal metabolism itself varies across species, further compounding these differences.

Given that the primary function of the kidney is to clear waste, toxins, and excess fluids from the blood, these species-specific differences directly impact our ability to predict nephrotoxicity using traditional animal models.

You’ve worked extensively with 3D human epithelial tissue models. What does moving from 2D cultures to 3D systems fundamentally change in how we understand toxicity mechanisms?

The difference between traditional 2D cultures and 3D systems, in the context of toxicity, is relatively straightforward. By recapitulating tissue structure, 3D models allow us to move beyond simply asking whether a compound is toxic, to understanding where that toxicity occurs and to what extent.

Understanding the relationship between exposure (where a polarized, functional cell sees a compound) and response is uniquely addressed in our systems compared to 2D. This is particularly important in epithelial tissues, where basolateral versus apical exposure can lead to very different toxicity outcomes. In skin, intestine, and lung, for example, cells may be exposed either from the basolateral side via systemic circulation or from the apical side through local administration or environmental contact. That distinction is fundamentally lost in 2D systems.

Do you see NAMs primarily as screening tools, or are they mature enough to inform dose selection, risk stratification, and IND-enabling decisions?

I believe NAMs have always been able to inform dose selection, risk stratification, and IND-enabling decisions. In fact, screening may not be the best deployment of NAMs due to scalability challenges and cost. The appropriateness of a NAM’s utility is dependent upon the limitations of the human biology you can explore within the NAM and the modality of the therapeutic. If the NAM contains the biology that you are targeting and the therapeutic modality is compatible with the model, then the NAM should be appropriate for dose selection, risk stratification and IND decisions.

One advantage you’ve previously highlighted is integrating human tissue models with live-cell analysis. Why is temporal resolution — seeing toxicity unfold in real time — so important?

There is both a practical and a biologically relevant dimension to the importance of temporal resolution in toxicity responses. From a practical standpoint, when developing any assay, identifying the time point at which the signal is maximal is essential for ensuring robustness and is a key part of assay optimization. In the context of toxicity, being able to observe the behavior and toxicity signals over time will enable you to identify the most appropriate time of incubation for maximal signal response.

Biologically, however, toxicity is not a single event — it manifests in different ways depending on mechanism. If you use tool compounds that induce toxicity through different mechanisms, knowing the kinetics of the toxicity response can help resolve whether your assay can distinguish direct and indirect mechanisms leading to toxicity.

In that sense, time to toxicity signal can be as informative as the signal itself, particularly when evaluating unknown compounds. In the context of advanced cell models for toxicity, often the exposure times can be prolonged (days to weeks) to predict clinical outcome.

NAMs can be scientifically compelling but still fail to gain traction. From a product and commercialization standpoint, what determines whether a NAM actually gets embedded into routine toxicology workflows?

This is the $100+ million question. Adoption of any platform is influenced by a range of factors — cost, fit-for-purpose utility, biological relevance, format, and ease of use among them. In practice, different players in the field tend to emphasize the aspects they value most, often based on their own balance of biological relevance versus scalability.

At the moment, traction tends to emerge organically through a “let’s try it and see if it works” approach. This is not unique to NAMs. However, toxicology is a particularly high-bar area, where established gold standards inherently challenge any new model system more than exploratory or discovery settings do. That makes sense: Toxicology groups are ultimately responsible for generating a weight of evidence that supports progression to the clinic.

In that context, NAMs introduce both opportunity and friction. While they offer potentially better predictive insight, they also require additional effort to validate against established approaches — often more effort than is required to continue using what is already accepted. Because of this, I would argue that regulators are the key gatekeepers of NAM adoption in toxicology. Ultimately, they define what is essential versus optional in the data package required to advance into the clinic. In my view, the biggest lever for accelerating adoption is therefore not customer preference, but regulatory acceptance.

What is the incentive to explore better models of toxicology if existing ones are “good enough” to reach regulatory endpoints? We could discuss the ethics and scientific rationale around choosing better, more predictive models. But if NAMs remain encouraged rather than required, it is difficult to expect meaningful acceleration in their uptake. I really hope that regulators recognize that there’s a big difference between accepting NAMs and requiring them. Making NAMs essential for toxicity studies for IND filing would catalyze their adoption far more effectively than incremental product refinement alone.