More Than Half a Century of Drug Development Attrition: Why Better Human-Relevant Models Are Needed

Drug development has delivered life-changing therapies to countless patients. Yet one challenge has remained remarkably consistent for decades: most drug candidates fail before reaching the market. This persistent clinical attrition highlights the need for a more predictive drug development strategy built around human-relevant evidence.  

A recent paper, Need for NAMs: A systematic evidence synthesis revealing over half a century of drug development failure, examined 32 publications reporting drug development outcomes across more than six decades.  The review included 60 clinical-attrition datasets covering 1963-2017 and found that mean clinical attrition ranged from approximately 80% to 91%. Mean attrition was approximately 91% in both the 2000s and the 2010s, showing no detectable improvement over the levels reported in the 1960s. 

Approximately 77% of reported clinical trial failures were attributed to biological factors, primarily insufficient efficacy and safety concerns, as well as unfavourable pharmacokinetic or pharmacodynamic properties. These are outcomes that preclinical studies are intended to anticipate. 

Although the underlying studies varied in their methods, reporting, time periods and data sources, the overall pattern is difficult to ignore. Despite major advances in molecular biology, disease research and drug discovery technologies, clinical attrition remains persistently high. The findings add urgency to an important question:  how can preclinical testing more accurately predict what will happen in patients?

The Translation Gap Remains a Major Bottleneck

Successful drug development depends on translating laboratory findings into meaningful clinical outcomes. Animal models have traditionally played a central role in this process, helping researchers investigate disease mechanisms and evaluate the safety, efficacy and pharmacokinetic properties of potential therapies before human testing begins.

However, biological and pharmacological differences between animals and humans can limit the predictive value of animal-derived data. A therapy may produce encouraging preclinical results in animals yet behave very differently in patients. The systematic evidence review does not establish that animal models are responsible for every clinical failure. Drug development is influenced by numerous scientific, regulatory, operational and commercial factors. Nevertheless, the finding that approximately 77% of reported failures involve biological factors raises important questions about how effectively prevailing preclinical approaches predict human safety and efficacy.

The result is a costly and time-consuming process in which many drug candidates fail only after substantial resources have already been invested. This persistent pattern suggests that human biology must be incorporated earlier and more effectively into preclinical decision-making.

The Growing Importance of New Approach Methodologies

In response to these challenges, New Approach Methodologies (NAMs) have gained significant momentum. NAMs are broadly defined as non-animal, human-relevant methods and strategies intended to improve the prediction of human biological responses. These non-animal methodologies provide alternative methods for generating evidence without relying solely on conventional animal models. These approaches include organoids, organ-on-chip systems, advanced in vitro assays, computational modeling, physiologically based pharmacokinetic modeling, and other emerging technologies.

These approaches include organoids, organ-on-chip systems, advanced in vitro assays, computational modelling, physiologically based pharmacokinetic modelling and other emerging technologies. They offer new ways to generate human relevant evidence without relying solely on conventional animal models.

Depending on the scientific question, NAMs may help researchers identify potential efficacy or safety concerns earlier, improve dose selection, investigate mechanisms of action and prioritize the most promising candidates for further development.

Regulatory agencies, academic researchers, and pharmaceutical companies are increasingly supporting the development and adoption of these technologies. To contribute meaningfully to drug development, models must be well characterized, reproducible, fit for purpose and evaluated against the biological or clinical outcomes they are intended to predict. The ultimate goal is not simply to shorten development timelines or reduce costs. More importantly, it is to improve preclinical decisions and increase the likelihood that therapies entering clinical trials will provide meaningful benefits to patients.

Human Tissues in New Approach Methodologies

Discussions on NAMs frequently emphasize emerging technologies such as organoids and organ-on-chip platforms. Functional human tissues donated by patients also have an important and sometimes underrepresented role in translational research.

Functional human tissues provide access to native, patient-derived biological systems. Depending on how tissues are collected, prepared and studied, they can retain aspects of tissue architecture, cellular interactions, disease-associated phenotypes and functional responses that may be difficult to reproduce fully in engineered models. Unlike systems designed to recreate selected features of human physiology, human tissues allow researchers to study drug responses directly in native human biology. They capture the biological complexity, disease heterogeneity and patient to patient variability present in real human populations.

This makes functional human tissues particularly valuable for investigating disease mechanisms, evaluating drug responses and identifying biomarkers that may support patient stratification and precision medicine.

Like all experimental approaches, functional human tissue models require careful study design. Tissue availability, viability, experimental duration, and study standardization must be appropriately managed. Their translational value depends on rigorous procurement and characterization, scientifically appropriate experimental design and the selection of tissues that address the research question. When these factors are carefully controlled, functional human tissues provide a highly relevant platform for evaluating drug activity directly in native, patient-derived biological systems. By preserving aspects of human biology and variability that are difficult to reproduce in animal or engineered models, they can generate evidence that is more closely aligned with clinical reality.

Better Translation Through Complementary Models

The future of translational research is unlikely to depend on any single technology.

Organoids can provide sophisticated three-dimensional models of tissue organization and disease-associated phenotypes. Organ-on-chip systems can incorporate flow, mechanical forces and multicellular interactions. Computational approaches can integrate complex datasets and generate testable predictions. Functional human tissues provide direct evidence of pharmacological responses in native, patient-derived biological systems.

Each approach captures different aspects of human biology and has its own strengths and limitations. The objective should be to combine fit-for-purpose New Approach Methodologies based on the question being addressed while incorporating direct evidence from functional human tissue whenever possible. Different tools may be required to evaluate target engagement, efficacy, toxicity, pharmacokinetics, disease mechanisms or variability among patient populations. Evidence generated across complementary models can provide a more complete understanding of drug behaviour, help researchers investigate discrepancies between systems, and support more robust decisions before a candidate advances into clinical trials.

Moving Toward a More Predictive Drug Development Strategy

The systematic evidence synthesis provides a sobering view of long-term drug development outcomes. Across six decades, clinical attrition has remained persistently high, while biological factors continue to account for most reported failures.

The authors acknowledge important limitations, including heterogeneity among the underlying studies, potential overlap between datasets and differences in reporting methods. Nevertheless, the broader pattern reinforces the need to evaluate and modernize preclinical testing strategies.

Reducing clinical attrition will require preclinical strategies that integrate evidence from complementary systems while incorporating human biology as early and effectively as possible. Functional human tissues can play a central role in this strategy by providing direct evidence of drug activity in complex, patient –derived biological systems.

When combined with organoids, organ on chip platforms, computational methods and other advanced approaches, functional human tissues can help researchers make better informed decisions before therapies enter clinical trials. After more than half a century of persistently high attrition, the need for more predictive preclinical research is clear.

Ultimately, the value of any model lies not in its novelty or complexity, but in the quality of the decisions it enables and in how well these decisions translate into better outcomes for patients.