Joint poster with Lundbeck: Case study on Occupational Health and Safety

Assessing Skin Sensitizing Hazard of Drug Products and Active Pharmaceutical Ingredients using GARD®: New Approach Methodology in Occupational Health and Safety

Presented at 2025 SOT

2025 SOT joint poster with Lundbeck on API and OHS

Download a copy in PDF

 

Conclusion

In conclusion, the findings highlight NAMs like GARD®skin Dose-Response as useful tools for enhancing OHS and safety protocols in pharmaceutical manufacturing.

​

Abstract

Background and Purpose:
Skin sensitizers are chemicals capable of inducing hypersensitivity reactions, such as Allergic Contact Dermatitis (ACD). Identifying and characterizing these chemicals’ skin sensitizing potential is essential for limiting hazardous exposure. Significant efforts have been made in the scientific community to develop New Approach Methodologies (NAMs) to replace animal testing for skin sensitization. Internationally recognized test guidelines as well as integrated strategies have been developed, combining in vitro, in chemico, and in silico approaches for predicting skin sensitization potential.

While widely accepted in cosmetics and personal care, NAMs’ application in the pharmaceutical sector for product development and Occupational Health and Safety (OHS) is less prominent. The aim of this study is to demonstrate how NAMs, including the in vitro GARD® assay, combined with in silico data, can be used by the pharmaceutical industry to assess the skin sensitization potential of drug products and Active Pharmaceutical Ingredients (APIs) to take protective measures and improve occupational safety in production environments.

In this study, the drug product is an oral medicine used to treat mental health problems, and the two APIs are compounds that affects how dopamine works in the brain, helping with symptoms like hallucinations or delusions, and helps boost certain chemicals in the brain to improve mood and reduce feelings of depression, respectively.

Methods:
Following reports of allergic reactions in production and quality control, a drug product and two APIs A and B were assessed. The drug product contained approximately 0.5% of API A and 9.4% of API B. In silico predictions using Nexus Derek software identified all compounds as potential sensitizers. The in vitro assay GARDskin Dose-Response (GARDskin DR) was conducted to confirm these predictions.

The GARDskin method (OECD 442E) is an in vitro assay for assessment of chemical skin sensitizers. The method provides binary hazard identification of skin sensitizers by evaluation of transcriptional patterns of an endpoint-specific genomic biomarker signature, comprising 196 genes, referred to as the GARDskin Genomic Prediction Signature (GPS), in the SenzaCell cell line. Final classifications are provided by a machine-learning prediction algorithm in the form of Decision Values (DV), the sign of which is evaluated by the prediction model. Any test chemical with a positive mean DV is classified as a skin sensitizer. Conversely, any test chemical with a negative mean DV is classified as non-skin sensitizer.

GARDskin DR is an adaptation of the conventional GARDskin method, in which test chemicals are evaluated by the GARDskin prediction algorithm in an extended range of concentrations, to investigate the dose-response relationship between GARDskin DVs and test chemical concentration. It provides a quantitative estimation of sensitizing potency, referred to as cDV0, which corresponds to the lowest required dose able to generate a positive mean DV. The readouts can predict correlating LLNA EC3 values, which are traditionally used to measure the skin sensitizing potency of chemicals. Furthermore, it can predict human skin sensitizing potency NOEL and GHS/CLP classification (1A or 1B), all with high statistical significance.

Results:

All test items were identified as sensitizers by GARDskin DR with the following results:
• Drug product: cDV0 = 12.8 µg/ml. Predictions: LLNA EC3 = 7.63%, NOEL = 3108 µg/cm², GHS/CLP 1B.
• API A: cDV0 = 1.51 µg/ml. Predictions: LLNA EC3 = 1.08%, NOEL = 254 µg/cm², GHS/CLP 1A.
• API B: cDV0 = 1.77 µg/ml. Predictions: LLNA EC3 = 1.19%, NOEL = 315 µg/cm², GHS/CLP 1A.

Conclusions:
Both APIs were classified as strong sensitizers, while the drug product was classified as a weak sensitizer, reflecting the dilution effect of excipients. The GARDskin DR assay demonstrated a clear dose-dependent increase in DVs for all three test items, confirming their classification as skin sensitizers, with varying potencies.
Based on the test outcomes, recommendations were made to improve occupational safety in handling these sensitizers:
• Avoiding dust inhalation and skin/eye contact.
• Minimizing prolonged and/or repeated exposure.
• Using appropriate personal protective equipment (PPE).
• Removing contaminated clothing and washing it before reuse.
• Ensuring thorough hand washing after handling, during breaks, and at the end of each shift.
• Handling substances in closed systems with proper ventilation.

These findings highlight NAMs like GARDskin DR as useful tools for enhancing OHS and safety protocols in pharmaceutical manufacturing.

Joint poster with Cargill: Case study on Hydrophobic Esters

The applicability of GARD®skin for assessing skin sensitization potential of hydrophobic esters during product development


Presented at Eurotox 2024

Download a copy in PDF

 

Conclusion

In conclusion, the inclusion of GARD®skin in the OECD test guidelines has expanded the applicability of NAMs for skin sensitization assessment.
This study demonstrates the applicability of the GARDskin assay to assess skin-sensitizing hazard of hydrophobic esters, which provides an ethical alternative to animal methods for safety assessment during product development.

​

Abstract

The field of skin sensitization assessment is rapidly evolving and the recent advancements in New Approach Methodologies (NAMs) has made it possible for the industry to perform in vitro skin ssensitization testing with good predictivity across a large chemical space. However, challenges remain for “difficult-to-test” chemicals, those with challenging physical/chemical properties or of Unknown or Variable composition, Complex reaction products or Biological materials (UVCBs), which are often outside the applicability domain of conventional cell-based assays. GARDskin (OECD TG 442E) is a genomic-based assay with demonstrated applicability to “difficult-to-test” substances.

The aim of this study is to assess the skin sensitization potential of two ester substances of biological origin, substance A and B, using the GARDskin assay. These substances are very hydrophobic and fall outside of the applicability domain of the conventional in vitro assays.

Both substances were successfully solubilised in cell media by utilizing a combination of heating, sonication, and selection of appropriate solvent vehicles (ethanol or acetone). No cytotoxicity was observed for either substance, thus 500 μM was chosen as the input concentration for cellular stimulations. GARDskin combines a genomic readout with machine learning to predict skin sensitizing hazard, where values above the threshold (DV=0) is predicted as skin sensitizing and below as non-sensitizing. Both substances resulted in negative mean Decisions Values and thus were classified as non-sensitizers in GARDskin (A: -1.54, B: -0.339).

In conclusion, with the inclusion of GARDskin into the OECD test guidelines, the range to where NAMs are appropriate has been increased. This study demonstrates the applicability of the GARDskin assay to assess skin sensitizing hazard of hydrophobic ethyl esters, which provides an ethical alternative to animal methods for safety assessment during product development.

Keywords: NAM, skin sensitization, difficult-to-test, UVCB, Applicability domain

Joint publication with Corteva Agriscience: GARD™skin and GARD™potency: A proof-of-concept study investigating applicability domain for agrochemical formulations

New joint publication with Corteva Agriscience.

SenzaGen scientists, alongside the toxicology team at Corteva Agriscience, have recently published a joint study in Regulatory Toxicology and Pharmacology, presenting new peer-reviewed evidence on the applicability of GARD® for agrochemical formulations.

The study demonstrates a satisfactory performance of GARD®skin and GARD®potency for skin sensitization hazard and GHS potency categorization of tested agrochemical formulations.

Marco Corvaro, Joseph Henriquez, Raja Settivari, Ulrika Mattson, Andy Forreryd, Robin Gradin, Henrik Johansson, Sean Gehen,
Regulatory Toxicology and Pharmacology, Volume 148, 2024, 105595, ISSN 0273-2300,
https://doi.org/10.1016/j.yrtph.2024.105595.

Keywords

GARD™; In vitro; Skin sensitisation; NAMs; Agrochemical formulations

Highlights

  • Tested 42 agrochemical formulations to expand applicability domain of GARD.

  • GARDskin showed good accuracy (76.2%), sensitivity (85.0%) and specificity (68.2%).
  • GARDpotency correctly subcategorized 14/17, correctly predicted sensitisers.
  • GARD satisfactory for Key Event 3 characterisation of agrochemical formulations.


Abstract

Several New Approach Methodologies (NAMs) for hazard assessment of skin sensitisers have been formally validated. However, data regarding their applicability on certain product classes are limited. The purpose of this project was to provide initial evidence on the applicability domain of GARD™skin and GARD™potency for the product class of agrochemical formulations.

For this proof of concept, 30 liquid and 12 solid agrochemical formulations were tested in GARDskin for hazard predictions. Formulations predicted as sensitisers were further evaluated in the GARDpotency assay to determine GHS skin sensitisation category. The selected formulations were of product types, efficacy groups and sensitisation hazard classes representative of the industry’s products.

The performance of GARDskin was estimated by comparing results to existing in vivo animal data. The overall accuracy, sensitivity, and specificity were 76.2% (32/42), 85.0% (17/20), and 68.2% (15/22), respectively, with the predictivity for liquid formulations being slightly higher compared to the solid formulations. GARDpotency correctly subcategorized 14 out of the 17 correctly predicted sensitisers. Lack of concordance was justifiable by compositional or borderline response analysis. In conclusion, GARDskin and GARDpotency showed satisfactory performance in this initial proof-of-concept study, which supports consideration of agrochemical formulations being within the applicability domain of the test methods.

 

New scientific publication by ExxonMobil: Challenges integrating skin sensitization data for assessment of difficult to test substances

New scientific publication by ExxonMobil.

Check out this newly published article by ExxonMobil focusing on the challenges of determining skin sensitization hazard in the case of difficult-to-test substances with conflicting or low-confidence data, where GARDskin data on UVCBs and hydrophobic substances provides valuable input for the integrated hazard assessment.

The article provides new peer-reviewed evidence for the applicability domain of GARDskin on UVCBs, hydrophobic and highly complex substances.

Greminger A, Frasca J, Goyak K, North C. 
Challenges integrating skin sensitization data for assessment of difficult to test substances. 
ALTEX - Alternatives to animal experimentation, published Oct 12, 2023
doi: 10.14573/altex.2201122. Epub ahead of print. PMID: 37843019.

Keywords

LLNA; ToxPi; new approach methodologies; skin sensitization; weight of evidence.


Abstract

Difficult to test substances including poorly soluble, mildly irritating, or Unknown or Variable Composition Complex reaction products or Biological Materials (UVCBs), producing weak or borderline in vivo results, face additional challenges in in vitro assays that often necessitates data integration in a weight of evidence (WOE) approach to inform skin sensitization potential. Here we present several case studies on difficult to test substances and highlight the utility of Toxicological Prioritization Index (ToxPi) as a data visualization tool to compare skin sensitization biological activity. The case study test substances represent two poorly soluble substances, tetrakis (2-ethylbutyl) orthosilicate and decyl palmitate, and two UVCB substances, alkylated anisole and hydrazinecarboximidamide, 2-[(2-hydroxyphenyl)methylene]-, reaction products with 2 undecanone. Data from key events within the skin sensitization adverse outcome pathway were gathered from publicly available sources or specifically generated. Incorporating the data for these case study test substances as well as on chemicals of a known sensitization class (sensitizer, irritating non-sensitizer, and non-sensitizer) into ToxPi produced biological activity profiles which were grouped using unsupervised hierarchical clustering. Three of the case study test substances concluded to lack skin sensitization potential by traditional WOE produced biological activity profiles most consistent with non-sensitizing substances, whereas the prediction was less definitive for a substance considered positive by traditional WOE. Visualizing the data using bioactivity profiles can provide further support for WOE conclusions in certain circumstances but is unlikely to replace WOE as a stand-alone prediction due to limitations of the method including the impact of missing data points.

 

Joint poster with ExxonMobil: Case study on UVCBs and Formulated Lubricant Products

Assessing the Utility of the Genomic Allergen Rapid Detection (GARDskin) Assay to Detect Dermal Sensitization Potential in UVCBs and Formulated Lubricant Products


Presented at Eurotox 2023

 

Download a copy

 

Summary

  • GARD®skin is considered to provide useful information in an overall weight of evidence assessment for difficult to test materials (mixtures, UVCBs) with challenging physical chemical properties.
  • The accuracy for prediction of skin sensitization hazard ranged from 66% for formulated lubricants/greases to 100% for synthetic base oils compared to expected outcomes based on reference data.

Abstract

Advances in new approach methods and their combinations into defined approaches can provide clarity and confidence in concluding on skin sensitization potential. However, challenges remain in utilizing these approaches for difficult to test materials such as those with challenging physical chemical properties (low water solubility, hydrophobic substances) or complex compositions like Unknown or Variable Composition Complex reaction products or Biological Materials (UVCBs) and formulated mixtures. The previously developed available non-animal test methods for skin sensitization based on key-events of the adverse outcome pathway (AOP) have clearly defined requirements for test material properties that impact feasibility or confound reliance on negative results particularly for difficult to test materials and impedes the application of defined approaches to conclude on skin sensitization hazard. A set of difficult to test materials were evaluated in the recently validated GARDskin assay since it offered advantages such as a broader applicability domain, availability of additional validated test solvents for poorly soluble materials and provides mechanistically relevant information on key events from across the skin sensitization AOP. The aim of the study was to evaluate the accuracy of the GARDskin assay for a set of synthetic base oils (UVCBs), lubricant additives (UVCBs/poorly soluble substances) and fully formulated lubricants/greases (mixtures) as well as to provide additional information to assist in a weight of evidence determination given that several of the test materials had borderline or conflicting data from other key events within the skin sensitization AOP. All test items were adequately solubilized in one of the following solvents, Ethanol (0.1% final), DMSO (0.25% or 0.1% final), or Xylenes (0.1% final). SenzaCells were incubated in triplicate under standard conditions with the test items at a max concentration of 500uM for those with a known molecular weight or 100 ppm (w/v) for those without a known molecular weight. Following cell stimulations, RNA was isolated and endpoint measurements were performed using the GARDskin genomic profile signature. Based on the results of this study, the accuracy for prediction of skin sensitization hazard was 100% for synthetic base oils (n=4), 83% for lubricant additives (n=6), and 66% for formulated lubricants/greases (n=6) compared to expected outcomes based on available reference data. In some cases, the available reference data was borderline or considered to have low confidence due to confounding factors such as irritation, and nonmonotonic dose responses impacting the accuracy determination when compared one to one with either animal or human data. However, the GARDskin assay is considered to provide useful insight into the overall weight of evidence for difficult to test materials with conflicting datasets as it provides an additional profile of bioactivity across the skin sensitization adverse outcome pathway. 

 

Joint poster with Johnson Matthey: Case study on Metals

Expanding the applicability domain of NAMs for skin sensitization testing: A case study using GARDskin for assessment of metals


Presented at the 12th World Congress on Alternatives and Animal Use in the Life Sciences, 2023

 

Download a copy

 

Summary

  • Limited data are currently available to support the inclusion of metals into the applicability domain of the OECD TG 442 series of assays.
  • GARD®skin (OECD TG 442 E) correctly predicted 11/12 metals in this study, including nickel which is a false negative in LLNA.
  • GARD®skin has the potential to reduce the need for animal testing for the endpoint of skin sensitization within the metal production and medical device sectors.

 

Abstract

New Approach Methods (NAMs) for detection of sensitization have been validated and adopted as OECD TGs during the last decade. These assays target different Key Events (KE) in the AOP for skin sensitization and are increasingly being applied to replace animal models within different chemical sectors. However, further characterization of the applicability domain (AD) of these assays is critical to understand limitations and to facilitate regulatory uptake in other industrial sectors. Of particular interest from a scientific and regulatory perspective is the potential to use NAMs for assessment of metals, which have been proposed to act via alternative mechanisms to organic chemicals. The current study describes a joint effort by industry and assay developers to evaluate the AD of the GARDskin assay for metal compounds.  GARDskin is the first harmonised method utilizing a combination of genomics and machine learning for a regulatory endpoint and was recently adopted into OECD TG442E.

A selection of metal salts (n=13) was evaluated and the accuracy, sensitivity, and specificity for prediction of skin sensitizing hazard of metals were estimated to 92% (12/13), 100% (7/7) and 83% (5/6), respectively. Interestingly, transcriptomic analysis revealed almost identical response patterns in dendritic cells for metals and organic compounds, indicating a high similarity in the toxicity pathways driving classifications. In conclusion, the result from this study supports the inclusion of metals into the AD of GARDskin, which is an important step to ensure scientific/regulatory confidence to reduce the need for animal testing within the metal production and medical device sector.

Joint poster with ExxonMobil: Case study on UVCBs and Formulated Lubricant Products

Assessing the Utility of the Genomic Allergen Rapid Detection (GARDskin) Assay to Detect Dermal Sensitization Potential in UVCBs and Formulated Lubricant Products


Presented at SOT 2023

Download a copy

Summary

  • GARD®skin is considered to provide useful information in an overall weight of evidence assessment for difficult to test materials (mixtures, UVCBs) with challenging physical chemical properties.
  • The accuracy for prediction of skin sensitization hazard ranged from 66% for formulated lubricants/greases to 100% for synthetic base oils compared to expected outcomes based on reference data.

Abstract

Advances in new approach methods and their combinations into defined approaches can provide clarity and confidence in concluding on skin sensitization potential. However, challenges remain in utilizing these approaches for difficult to test materials such as those with challenging physical chemical properties (low water solubility, hydrophobic substances) or complex compositions like Unknown or Variable Composition Complex reaction products or Biological Materials (UVCBs) and formulated mixtures. The previously developed available non-animal test methods for skin sensitization based on key-events of the adverse outcome pathway (AOP) have clearly defined requirements for test material properties that impact feasibility or confound reliance on negative results particularly for difficult to test materials and impedes the application of defined approaches to conclude on skin sensitization hazard. A set of difficult to test materials were evaluated in the recently validated GARDskin assay since it offered advantages such as a broader applicability domain, availability of additional validated test solvents for poorly soluble materials and provides mechanistically relevant information on key events from across the skin sensitization AOP. The aim of the study was to evaluate the accuracy of the GARDskin assay for a set of synthetic base oils (UVCBs), lubricant additives (UVCBs/poorly soluble substances) and fully formulated lubricants/greases (mixtures) as well as to provide additional information to assist in a weight of evidence determination given that several of the test materials had borderline or conflicting data from other key events within the skin sensitization AOP. All test items were adequately solubilized in one of the following solvents, Ethanol (0.1% final), DMSO (0.25% or 0.1% final), or Xylenes (0.1% final). SenzaCells were incubated in triplicate under standard conditions with the test items at a max concentration of 500uM for those with a known molecular weight or 100 ppm (w/v) for those without a known molecular weight. Following cell stimulations, RNA was isolated and endpoint measurements were performed using the GARDskin genomic profile signature. Based on the results of this study, the accuracy for prediction of skin sensitization hazard was 100% for synthetic base oils (n=4), 83% for lubricant additives (n=6), and 66% for formulated lubricants/greases (n=6) compared to expected outcomes based on available reference data. In some cases, the available reference data was borderline or considered to have low confidence due to confounding factors such as irritation, and nonmonotonic dose responses impacting the accuracy determination when compared one to one with either animal or human data. However, the GARDskin assay is considered to provide useful insight into the overall weight of evidence for difficult to test materials with conflicting datasets as it provides an additional profile of bioactivity across the skin sensitization adverse outcome pathway. 

 

The GARDskin assay: Investigation of the applicability domain for metals

Joint publication with Johnson Matthey

ALTEX – Alternatives to animal experimentation, published Nov 03, 2022, accepted manuscript

DOI: https://doi.org/10.14573/altex.2203021

Forreryd, A., Gradin, R., Larne, O., Rajapakse, N., Deag, E. and Johansson, H.


Abstract

New approach methods (NAMs) for hazard identification of skin sensitizing chemicals have been adopted as test guidelines by the OECD during the last decade as alternatives to animal models. These models align to individual key events (KE) in the adverse outcome pathway (AOP) for skin sensitization for which the molecular initiating event (MIE) is covalent binding to proteins. As it currently stands, the AOP does not include mechanistic events of sensitization by metals, and limited information is available on whether NAMs accurately the predict sensitization potential of such molecules, which have been proposed to act via alternative mechanisms to organic chemicals.

Methods for assessing the sensitization potential of metals would comprise valuable tools to support risk management within e.g., occupational settings during production of new metal salts or within the medical device industry to evaluate leachables from metal alloys.

This paper describes a systematic evaluation of the applicability domain of the GARD™skin assay for assessment of metals. Hazard classifications were supplemented with an extended analysis of gene expression profiles induced by metal sensitizers to compare the induction of toxicity pathways between metals and organic sensitizers. Based on the results of this study, the accuracy, sensitivity, and specificity of GARD™skin for prediction of skin sensitizing hazard were 92% (12/13), 100% (7/7) and 83% (5/6), respectively.

Thus, the performance of GARD™skin for assessment of metals was found to be similar to what is observed on conventional organic substances, providing support for inclusion of metals within the applicability domain of the test method.

Keywords

skin sensitization, metals, regulatory testing, medical devices

Full article on line with open access

Exploration of the GARDskin applicability domain: Indirectly acting haptens, hydrophobic substances and UVCBs

Joint publication with the Lubrizol Corporation

ALTEX – Alternatives to animal experimentation, published April 21, 2022, accepted manuscript, https://doi.org/10.14573/altex.2201281

Forreryd, A., Gradin, R., Humfrey, C., Sweet, L. and Johansson, H.

Abstract

Hazard assessments of skin sensitizers are increasingly being performed using new approach methodologies (NAMs), with several in chemico, in vitro and most recently also defined approaches (DAs) being accepted for regulatory use. However, keeping track of potential limitations of each method in order to define applicability domains remains a crucial component to ensure adequate predictivity as well as facilitating the appropriate selection of method(s) for each hazard assessment task. The objective of this report is to share test results generated with the GARD™skin assay on chemicals that have traditionally been considered as difficult to test in some of the conventional in vitro and in chemico OECD Test Guidelines for skin sensitization. Such compounds may include, for example, indirectly acting haptens, hydrophobic substances, and substances of unknown variable composition or biological substances (UVCBs). Based on the results of this study, the sensitivity for prediction of skin sensitizing hazard of indirectly acting haptenswas92.4%and 87.5%, when compared with LLNA(n=25)and human data(n=8), respectively. Similarly, the sensitivity for prediction of skin sensitizing hazard of hydrophobic substances was 85.1% and 100%, when compared with LLNA(n=24)and human data(n=9), respectively. Lastly, a case study involving the assessment of a set of hydrophobic UVCBs(n=7) resulted in a sensitivity of 100, as compared to available reference data. Thus, it was concluded that these data provide support for the inclusion of such chemistries in the GARD™skin applicability domain, without an increased risk of false negative classifications.

Key words: GARD, GARDskin, skin sensitization, applicability domain, difficult to test substances, Indirectly acting haptens, hydrophobic substances, UVCBs

 

Full article
Article on line with open access

Applicability of GARD™skin for Accurate Assessment of Challenging Substances in the Context of Skin Sensitization Testing

Poster presented at ACT 2020

J. Schmidt, A. Forreryd, H. Johansson, J. Li, A. Johansson
SenzaGen, Inc., Raleigh, NC, USA, SenzaGen AB, Lund, Sweden

Link to the poster

 

Conclusion

  • GARDskin demonstrated an overall high applicability for the evaluated challenging substances with 80% predictive accuracy compared to existing human data.
  • GARDskin demonstrated excellent applicability for pre/pro-haptens and low water solubility substances, correctly classifying all such compounds in the herein investigated dataset.
  • GARDskin also showed high applicability for assessment of surfactants with 89% predictive accuracy compared to existing human data, correctly classifying 8 out of 9 internally tested surfactants, including well known challenging ones such as Sodium Dodecyl Sulphate (SDS) and Benzalkonium chloride.

Abstract

Current legislations and trends in predictive toxicology advocate a transition from in vivo methods for hazard and risk assessments to non-animal alternatives. However, certain groups of chemicals, including substances with severe membrane-damaging properties, pre- and pro-haptens, and those with high log P ratios, have been shown to be challenging to assess using cell-based assays in the context of skin sensitization testing. The aim of this study was to evaluate the applicability of GARDskin for such challenging substances, using an overlapping subset of chemicals previously tested in an integrated tested strategy (ITS) based on validated, aqueous in vitro assays, as well as in a series of Reconstructed Human Epidermis (RHE)-based assays.

The GARDskin assay (Genomic Allergen Rapid Detection) is a robust in vitro assay for identification of potential chemical skin sensitizers with over 90% prediction accuracy and broad applicability. The assay is included in the OECD Test Guideline Program (OECD TGP 4.106) and has gone through a formal validation study. The assay evaluates the gene expression of endpoint-specific genomic biomarkers in a human dendritic-like cell line following exposure to the test substance. Exposure-induced gene expression patterns are analysed using pattern recognition and machine-learning technology, providing classifications of each test item as a skin sensitizer or a non-sensitizer.

The applicability of GARDskin for a total of twelve challenging substances, including pre- and pro-haptens, low water-soluble substances, two surfactants and three additional substances known to have conflictive results when comparing in vitro and in vivo data were evaluated in this study. All twelve substances were selected from the Mehling et al. 2019 publication which reported results from three OECD validated in vitro methods, the “2 out of 3” Integrated Testing Strategy, three RHE-based models and the murine local lymph node assay (LLNA). Human potency classification was available for ten out of the twelve substances.

The GARDskin prediction results were reported from previously published studies, or from in house validation studies. Predictive accuracies were calculated by comparing skin sensitization classifications from different test methods to the available human data of each substance respectively. (N=10). To further explore and substantiate the GARDskin applicability for surfactants, additional GARDskin data for a total of nine surfactants are presented in order to complement the Mehling dataset with respect to the availability of human data.

The GARDskin assay demonstrated overall high applicability for the evaluated challenging substances, with 80% predictive accuracy compared to existing human data. GARDskin correctly classified all pre-and pro-haptens and low water-soluble substances in the data set. Furthermore, high applicability of GARDskin for severe membrane disruptive substances such as surfactants was demonstrated, with 89% predictive accuracy compared to existing human data.