Testing developmental toxicity in a second species: are the differences due to species or replication error?

Evaluation of rat and rabbit embryofetal development studies with pharmaceuticals: the added value of a second species

Embryofetal development (EFD) studies are performed to characterize risk of drugs in pregnant women and on embryofetal development. In line with the ICH S5(R3) guideline, these studies are generally conducted in one rodent and one non-rodent species, commonly rats and rabbits. However, the added value of conducting EFD studies in two species to risk assessment is debatable. In this study, rat and rabbit EFD studies were evaluated to analyze the added value of a second species. Information on rat and rabbit EFD studies conducted for human pharmaceuticals submitted for marketing authorization to the European Medicines Agency between 2004 and 2022 was collected from the database of the Dutch Medicines Evaluation Board, along with EFD studies conducted for known human teratogens. In total, 369 compounds were included in the database. For 55.6% of the compounds similar effects were observed in rat and rabbit EFD studies. Discordance was observed for 44.6% of compounds. Discordance could often be explained based on occurrence of maternal toxicity or the compound's mechanism of action. For other compounds, discordance was considered of limited clinical relevance due to high exposure margins or less concerning EFD toxicity. For 6.2%, discordance could not be explained and was considered clinically relevant. Furthermore, for specific therapeutic classes, concordance between rat and rabbit could vary. In conclusion, in many cases the added value of conducting EFD studies in two species is limited. These data could help identify scenarios in which (additional) EFD studies could be waived or create a weight-of-evidence model to determine the need for (additional) EFD studies.

Missing Genomic Resources for the Next Generation of Environmental Risk Assessment

Environmental risk assessment traditionally relies on a wide range of in vivo testing to assess the potential hazards of chemicals in the environment. These tests are often time-consuming and costly and can cause test organisms' suffering. Recent developments of reliable low-cost alternatives, both in vivo- and in silico-based, opened the door to reconsider current toxicity assessment. However, many of these new approach methodologies (NAMs) rely on high-quality annotated genomes for surrogate species of regulatory risk assessment. Currently, a lack of genomic information slows the process of NAM development. Here, we present a phylogenetically resolved overview of missing genomic resources for surrogate species within a regulatory ecotoxicological risk assessment. We call for an organized and systematic effort within the (regulatory) ecotoxicological community to provide these missing genomic resources. Further, we discuss the potential of a standardized genomic surrogate species landscape to enable a robust and nonanimal-reliant ecotoxicological risk assessment in the systems ecotoxicology era.

What is the ecotoxicity of a given chemical for a given aquatic species? Predicting interactions between species and chemicals using recommender system techniques

Ecotoxicological safety assessment of chemicals requires toxicity data on multiple species, despite the general desire of minimizing animal testing. Predictive models, specifically machine learning (ML) methods, are one of the tools capable of solving this apparent contradiction as they allow to generalize toxicity patterns across chemicals and species. However, despite the availability of large public toxicity datasets, the data is highly sparse, complicating model development. The aim of this study is to provide insights into how ML can predict toxicity using a large but sparse dataset. We developed models to predict LC50-values, based on experimental LC50-data covering 2431 organic chemicals and 1506 aquatic species from the ECOTOX-database. Several well-known ML techniques were evaluated and a new ML model was developed, inspired by recommender systems. This new model involves a simple linear model that learns low-rank interactions between species and chemicals using factorization machines. We evaluated the predictive performances of the developed models based on two validation settings: 1) predicting unseen chemical-species pairs, and 2) predicting unseen chemicals. The results of this study show that ML models can accurately predict LC50-values in both validation settings. Moreover, we show that the novel factorization machine approach can match well-tuned, complex, ML approaches.

An assessment of the reliability of 52 enhanced preliminary embryofetal development studies to detect developmental toxicity

BACKGROUND

Preliminary embryofetal development (pEFD) data from two species are currently recommended before inclusion of women of child-bearing potential (WOCBP) in clinical trials in Europe or Japan, but not before trials in the United States. The ICH S5(R3) guideline advises an "enhanced" study design for this purpose.

METHODS

The reliability of pEFD studies was assessed by comparing the outcome of 52 enhanced pEFD studies (25 rat, 23 rabbit, and 4 mouse) with the results of the definitive nonclinical EFD assessment.

RESULTS

Four pEFD studies revealed severe developmental hazard without the need for a main EFD study. Only one pEFD study failed to detect drug-related teratogenicity or pregnancy failure subsequently detected in the main study. There were, however, some false positive and some equivocal pEFD study results. Of the 48 pEFD studies for which a main EFD study was performed, 16 (33%) failed to accurately predict (within two-fold) the no adverse effect level (NOAEL) for developmental toxicity subsequently defined in the main EFD study. Skeletal examination of fetuses in the pEFD study was necessary was to detect drug-induced malformations. One quarter (23%) of EFD investigations revealed malformations and/or pregnancy failure at one dose level or more.

CONCLUSIONS

pEFD studies are effective for the detection of serious or irreversible effects on embryofetal development, provided that full fetal examinations are completed. They are not, however, sufficiently powered to reliably define the NOAEL for developmental toxicity. The regulatory impact of pEFD studies remains obscure since maximum pregnancy prevention precautions are required in clinical trials in all regions until the results of the main EFD studies are available.

In vitro models of human development and their potential application in developmental toxicity testing

Recent publications describe the development of in vitro models of human development, for which applications in developmental toxicity testing can be envisaged. To date, these regulatory assessments have exclusively been performed in animal studies, the relevance of which to adverse reactions in humans may be questioned. Recently developed cell culture-based models of embryo-fetal development, however, do not yet exhibit sufficient levels of standardisation and reproducibility. Here, the advantages and shortcomings of both in vivo and in vitro developmental toxicity testing are addressed, as well as the possibility of integrated testing strategies as a viable option in the near future.

Review of embryo-fetal developmental toxicity studies performed for pharmaceuticals approved by FDA in 2018 and 2019

Details of embryo-fetal development (EFD) studies were compiled for all FDA drug approvals in 2018-19. EFD studies were performed for 82 % of approvals (84 % of small molecules and 70 % of biopharmaceuticals). Rats and rabbits were used for 84 % of small molecule (SM) drugs for which EFD studies were submitted. There was at least a 2-fold difference in sensitivity between the rat and the rabbit relative to the human exposure for the majority of drugs (62 %, small molecules and biopharmaceuticals combined) tested in both species. On average, however, the rat and rabbit were equally sensitive to developmental toxicity. Over the last 2 years, the use of non-human primates (NHP) for the developmental toxicity testing of biopharmaceuticals has fallen (26 % of biologics license applications), with many more biopharmaceuticals now tested in rodents (44 % of BLAs). EFD studies were not required for oncology drugs when the mode of action was associated with known developmental risk. One-third of SM non-oncology drugs and two-thirds of SM oncology drugs induced dysmorphogenesis in at least one species. The newly revised ICH S5(R3) guideline will bring about changes to the design of future EFD studies, particularly with respect to high dose selection. The revised guideline will also influence the interpretation of the findings in EFD studies (e.g. fetal morphological variations) and risk assessment.