Defined Approaches to Identify Potential Skin Sensitizers
Potential skin sensitizers can be identified without animal testing using defined approaches to testing and assessment. A defined approach consists of input data generated from identified methods and a data interpretation procedure, such as a machine-learning model, flowchart, or decision tree, through which the data are evaluated.
A defined approach can be developed on a framework provided by an adverse outcome pathway (AOP). An AOP is a conceptual model that links an exposure to a substance to a toxic effect by identifying the sequence of biochemical events required to produce that toxic effect. In vitro methods that observe or measure the individual events within an AOP can be used as elements of a defined approach to generate a prediction of whether a chemical might produce that toxic effect. Read More
Skin sensitizers are substances with the potential to cause allergic contact dermatitis. The key biological events of skin sensitization initiated through covalent binding to skin proteins are well characterized and form the basis for an AOP for skin sensitization. NICEATM, working with NTP, ICCVAM, and industry scientists, has created and evaluated defined approaches to identify potential skin sensitizers based on this AOP.
On April 10, the U.S. Environmental Protection Agency (EPA) released a draft Science Policy to reduce animal use by using defined approaches to identify potential skin sensitizers. The draft policy is the result of national and international collaboration among ICCVAM, NICEATM, Cosmetics Europe, the European Union Reference Laboratory for Alternatives to Animal Testing, and Health Canada’s Pest Management Regulatory Agency. Comments on the draft skin sensitization policy can be submitted to docket EPA-HQ-OPP-2016-0093 at www.regulations.gov through June 9.
Collaboration With Cosmetics Europe to Evaluate Defined Approaches
NICEATM and Cosmetics Europe collaborated to evaluate multiple defined approaches for skin sensitization safety assessment that had been submitted to the Organisation for Economic Co-operation and Development. The collaboration produced two publications.
- The first paper describes a database including data from human, animal, and five non-animal tests for 128 chemicals. The chemicals in the database represent a wide variety of chemical structures and use categories. The database is proposed as a point of reference for the evaluation and development of new testing strategies.
Hoffmann et al. 2018. Non-animal methods to predict skin sensitization (I): the Cosmetics Europe database. Crit Rev Toxicol 48(5):344-358.
- The second paper describes an analysis of multiple defined approaches for skin sensitization safety assessment of cosmetics ingredients using the database described above. Many of these approaches were found to perform as well or better than animal methods to predict human skin sensitization hazard.
Kleinstreuer et al. 2018. Non-animal methods to predict skin sensitization (II): an assessment of defined approaches. Crit Rev Toxicol 48(5):359-374.
Defined Approach Developed by ICCVAM
NICEATM and ICCVAM developed several versions of a defined approach that uses non-animal data to predict skin sensitization hazard and potency. This approach combines data from the direct peptide reactivity assay, the KeratinoSens assay, and the human cell line activation test; a read-across prediction of skin sensitization hazard generated by the QSAR Toolbox software package; and physical property data such as partition coefficients.
- The first version of the defined approach used computer algorithms to integrate data to predict murine local lymph node assay outcomes.
Strickland et al. 2016. Integrated decision strategies for skin sensitization hazard. J Appl Toxicol 36(9):1150-62.
- The second version used data from human exposures to predict human skin sensitization hazard.
Strickland et al. 2017. Multivariate models for prediction of skin sensitization hazard. J Appl Toxicol 37(3):347-360.
- Further development of this approach aimed to predict human or animal skin sensitization potency, enabling the classification of skin sensitizers as “weak” or “strong” without animal tests.
Zang et al. 2017. Prediction of skin sensitization potency using machine learning approaches. J Appl Toxicol 37(7):792-805.
Collaboration With P&G to Develop an Open-source Integrated Testing Strategy
An integrated testing strategy is a type of defined approach that relies on:
- Input data generated from identified methods
- A data interpretation procedure such as a machine-learning model, flowchart, or decision tree, through which the data are evaluated
- A simultaneous assessment of the input data to arrive at either a hazard prediction or a decision that more testing is needed
NICEATM and other NTP scientists collaborated with scientists at Procter & Gamble (P&G) to develop an integrated testing strategy to identify potential skin sensitizers without conducting animal tests. Using data from non-animal tests and other information such as solubility and computational predictions of skin sensitizer activity, the strategy produces a numerical probability that a chemical should be placed in a particular skin sensitization hazard class: strong, moderate, weak, or nonsensitizer. P&G and NTP scientists collaborated to develop this integrated testing strategy using free, publicly available software. The goal is to encourage organizations worldwide to use this approach for identifying potential skin sensitizers and support the elimination of animal testing in this area. Read More
Pirone et al. 2014. Open source software implementation of an integrated testing strategy for skin sensitization potency based on a Bayesian network. ALTEX 31:336-340.
P&G updated the integrated testing strategy in 2015. The updated strategy uses only validated non-animal tests, simplifies the bioavailability inputs, and nearly doubles the size of the database used to derive the previous network. The updated P&G integrated strategy is the basis for a free web tool produced by Douglas Connect that can be used to predict skin sensitization potency of a chemical.
Jaworska et al. 2015. Bayesian integrated testing strategy (ITS) for skin sensitization potency assessment: a decision support system for quantitative weight of evidence and adaptive testing strategy. Arch Toxicol 89: 2355-2383.
Files for Running the Open-source Integrated Testing Strategy Analysis
Files to run the analysis described in Pirone et al., including a script that uses the R programming language, are available below.
- Pirone et al.: Reproducing the ITS-2 Model Using R (updated March 31, 2014: read this document first)
- Supplemental Information on Changes from ITS-2 to ITS-2 Lipid and Moving Toward Open-Source
- Compressed (.ZIP) folder containing files for running the analysis (updated March 31, 2014: save to your hard drive). Folder contains:
- R code necessary for conducting the analysis (ITS2_R_version.R)
- Additional R functions needed for the analysis (ITS2_Supplemental_R_Functions.R)
- Document compiled by Sweave to produce the LaTeX file (ITS2_R_version.Rnw)
- Sweave style file needed by LaTeX (Sweave.sty)
- File produced by running Sweave on "ITS2_R_version.Rnw" (ITS2_R_version.tex)
- Data files (tab-delimited text format)
- Training data (ITS2_Lipid_Train_102313.txt)
- Test data (ITS2_Lipid_Test_102313.txt)
- References in BibTeX format (ITS2_Refs.bib)
A bioavailability calculator is available from the National Institute for Occupational Safety and Health (NIOSH) website
Please note that Java must be enabled in your web browser in order to run the calculator.
The software for running R can be obtained from the R Project website
Refer to the FAQs on the R Project website for system requirements and installation instructions.
The OECD QSAR Toolbox can be obtained from the OECD website
Installation instructions and user documentation are available on this page.
NICEATM Murine Local Lymph Node Assay (LLNA) Database
On behalf of ICCVAM, NICEATM conducted a number of analyses to evaluate the usefulness of the LLNA to identify potential skin sensitizers. Data from these analyses are available to interested stakeholders as a reference for developing and evaluating alternative approaches to testing and assessment that replace, reduce, or refine the use of animals for identification of potential skin sensitizers.
Notes for Database Users
- The Excel spreadsheet contains three pages: (1) data, (2) references, and (3) abbreviations.
- The database includes data published through 2010.
- These data have been extracted from published and unpublished data sources with permission. Users of this database should consult the original data source for questions regarding data quality and/or authenticity.