FDA Reviewers Describe Why Some NAM Submissions Remained Supportive

Reviewer experience separates assay function from decision utility

A NAM can work as designed and still fail to answer the question posed in a drug-development program. That distinction runs through an article by pharmacology and toxicology reviewers at FDA's Center for Drug Evaluation and Research and Office of New Drugs. Published online in November 2025, the paper reviews established uses and anonymized submissions, explaining which evidence informed decisions and which gaps limited its use.

This is the authors’ perspective, rather than a new FDA guidance or agency determination. Its disclaimer explicitly states that the findings and conclusions are their own. The case examples describe the evidence available for particular submissions and do not establish a universal acceptance or rejection of a model family.

Established applications have defined evidential pathways

The article describes NAMs already incorporated into nonclinical assessment, including integrated carcinogenicity evaluation, cardiac risk and computational assessment of potentially mutagenic impurities. These examples have different scopes. A weight-of-evidence assessment can address whether a 2-year rat carcinogenicity study would add information for a particular pharmaceutical, while other products still require that study.

For cardiac risk, the Comprehensive in Vitro Proarrhythmia Assay (CiPA) combines ion-channel measurements, computational modeling, stem-cell-derived cardiomyocytes and clinical electrocardiographic evidence. The authors report that human induced pluripotent stem cell (iPSC)-derived cardiomyocyte assessments had contributed to submitted development programs and some labeling decisions, while noting that not every package implemented the complete CiPA framework.

For potentially mutagenic impurities, complementary expert-rule and statistical structure–activity methods address a defined hazard under ICH M7. These established contexts differ from expecting a new complex tissue model to replace an entire repeat-dose toxicology assessment. The article uses that contrast to connect the intended regulatory question with the evidence needed to answer it.

A liver case supplied mechanistic support

One anonymized program involved serious clinical liver injury in a rare-disease population receiving a previously approved drug at higher doses and for longer use. The sponsor combined primary-human-hepatocyte micropatterned coculture with computational liver-injury modeling to investigate susceptibility pathways and possible biomarkers.

The in vitro work used hepatocytes from 3 adult donors with mouse fibroblasts. It found mitochondrial dysfunction, oxidative stress and impaired function with increased apoptosis, while membrane integrity was not affected as assessed through alanine aminotransferase (ALT) release. The computational analysis predicted greater toxicity under the rare-disease dosing conditions than under the earlier-use conditions.

Reviewers identified limitations in disease representation, donor justification, controls and exposure relevance. Nevertheless, the findings supported biological plausibility and suggested unvalidated clinical biomarkers. The serious clinical liver findings ultimately determined labeling and surveillance actions. This separates a useful mechanistic contribution from the evidence that drove the final safety decision.

A kidney-chip result preserved an unsuccessful therapeutic hypothesis

Another example tested complement inhibition in a human kidney microphysiological system with proximal tubular epithelial and endothelial cells. Ischemic conditions activated complement, and the investigational product inhibited that activation. However, active versus inactive complement did not change the measured injury, and the product did not alleviate cellular damage or barrier impairment.

The reviewers therefore found no demonstrated connection between complement inhibition and the intended protective outcome in that model. Missing components, including relevant immune cells, constrained its biological representation. The study provided mechanistic information but was not used for regulatory decision-making. Target engagement and functional rescue remained separate findings.

Extensive respiratory testing still left an exposure gap

A further submission sought to use human upper- and lower-airway cultures in place of a non-rodent inhalation study. The package included barrier, cytotoxicity, metabolic and histological measurements, interference testing and documented study deviations. Reviewers considered the information sufficient to assess how the experiments were conducted.

In the upper-airway model, the investigational product did not damage epithelial cell membranes or induce cytotoxicity, but tight-junction barrier integrity was reduced after 24-hour exposure. The lower-airway model showed no membrane damage, cytotoxicity or reduction in barrier integrity.

The assays nevertheless lacked a demonstrated relationship to clinically meaningful inhalation toxicity and to the proposed human exposure. The tested concentrations were 100–15,000 times the estimated tissue concentrations, and the computational deposition approach had not established valid in vitro-to-in vivo extrapolation. Immune, mechanical and organ-interaction features were also missing, while an excessively toxic positive control complicated interpretation. The data were not accepted as a replacement supporting the proposed clinical safety decision.

The scope of use organizes the scientific questions

Across the cases, the authors discuss five recurring areas: a specific context of use, relevant biology and endpoints, technical characterization, study and data integrity, and transparency. An exploratory study conducted outside Good Laboratory Practice (GLP) could still provide useful evidence when its conduct and data were adequately documented. Conversely, a thorough report could remain insufficient if the exposure or functional relationship needed for the intended decision was unestablished.

The paper also describes developmental-toxicity and liver-injury approaches under investigation. Its liver-chip example records acceptance of a letter of intent into FDA’s ISTAND drug-development-tool qualification pilot in September 2024, with development of a qualification plan as a subsequent step.