Experts propose safety framework for improving predictability, interpretation of CNS toxicity
The use and implementation of new approach methodologies (NAMs) and digital technologies such as artificial intelligence (AI) may help improve the predictability and interpretation of central nervous system (CNS) toxicity through non-clinical safety assessments, according to a recent paper proposing a new safety framework for predicting neurological risk in CNS drug development.
“CNS safety remains a major contributor to clinical reflecting limitations in the predictive resolution and translational alignment of conventional non-clinical paradigms,” Mamta Behl, of Neurocrine Biosciences Inc. in San Diego, and colleagues wrote in their paper recently published in Toxicological Sciences.
Behl and colleagues said that because CNS toxicities are serious, difficult to clinically monitor, and are potentially irreversible, there is a greater need for non-clinical characterization of these adverse effects.
“[T]he current nonclinical toolbox, particularly qualitative neurological screens and terminal histopathology, can miss time-dependent, rare, or higher-order functional liabilities, and may not capture clinically meaningful endpoints, and is limited in its ability to assess CNS adverse effects that can affect patient compliance,” they explained.
However, some new non-clinical approaches have shown promise in case studies, such as Integrated Approach to Testing and Assessment (IATA)-driven integration of NAMs, improvements in electroencephalography (EEG), and the use of AI in non-clinical animal behavior monitoring.
Elements of safety framework
The authors proposed the lessons from the case studies could serve as a safety framework for the assessment of CNS toxicities for drug development. Specifically, they recommended a number of actionable approaches, including changing study designs to more closely align with histopathology and neuropharmacology, using EEG as an objective quantitative tool to confirm neurological signs, considering developmental neurotoxicity and pediatric risks as unique contexts, implementing NAMs in mechanistic and exposure data using IATA, and using continuous digital monitoring to identify observational blind spots.
They noted that early design discussions with regulators can improve the decision relevance of non-clinical studies and make sure they remain fit-for-purpose.
“[S]tudy designs should be informed by mechanistic understanding of target biology, anticipated exposure profiles, and known patterns of CNS vulnerability and manifestations of CNS toxicity,” Behl and colleagues said. “This includes more holistic behavioral monitoring, tailoring sacrifice time-points, expanding regional brain sampling when warranted, and incorporating adjunctive endpoints when standard screening approaches may lack sensitivity.”
Greater use of quantitative tools such as in vitro methods and EEG can help identify true seizure activity when neurological signs are more ambiguous clinically, the authors said.
“Importantly, these tools not only improve diagnostic confidence but also allow for quantitative characterization of seizure onset, duration, frequency, and exposure dependence,” they explained. “Integration of EEG into dose range-finding or repeat-dose toxicology studies, when scientifically justified, can reduce animal use while substantially enhancing interpretability and translational relevance.”
Since excipients in adults could still be a risk to pediatric populations, using adult-focused non-clinical data for developmental neurotoxicity and pediatric safety may not be enough, Behl and colleagues said.
“Nonclinical strategies addressing pediatric risk should therefore explicitly consider developmental windows of susceptibility, incorporate age-appropriate animal models, and include both structural and functional endpoints,” they added.
CNS toxicity assessments could also benefit from the implementations of NAMs into mechanistic and exposure data using IATAs, which “provide a transparent and systematic framework for combining NAMs, traditional in vivo data, mechanistic insights, and exposure modeling.” The approach helps provide context to NAMs, identifies concordant evidence from disparate sources, and highlights sensitive endpoints.
“Importantly, IATA does not replace expert judgment but rather structures it, making assumptions and uncertainties explicit,” the authors explained. “When applied prospectively, IATA can guide targeted data generation, reduce unnecessary testing, and build scientific and regulatory confidence in NAM-based approaches.”
Digital technologies
The use of continuous digital monitoring could help improve observational blind spots seen in traditional clinical assessments, which “rely heavily on episodic, observer-dependent measurements that may miss rare events, circadian-dependent effects, or subtle functional changes,” the authors said. This can be accomplished with AI analytics, wearable sensors, and computer vision-based sensors to detect behaviors associated with seizures, properly contextualize pharmacologic effects, and identify functional changes early in progressive neurodegenerative models.
“These approaches enhance objectivity, support longitudinal analysis, and generate quantitative data that can be integrated with EEG, pathology, NAMs and biomarker readouts,” they explained. “While challenges remain related to data volume, validation, and regulatory qualification, collaborative efforts and clearly defined contexts of use are paving the way for broader adoption.”
The goal of this new approach would not be to replace the status quo but improve the efficiency of animal studies using these tools in a weight-of-evidence framework.
“Ultimately, advancing this integrated CNS safety framework will require continued cross-sector collaboration, methodological standardization, clearly defined contexts of use, and prospective evidence that emerging technologies improve decision quality and patient safety,” the authors said. “Together, these advances should foster a more predictive and translationally aligned approach to CNS safety assessment, enabling earlier identification of neurological risk, reduced animal use, and result in safer development of neurological therapeutics.”
