Table of Contents

Definition

Integrated Approaches to Testing and Assessment (IATA) are structured, science-based frameworks, developed under the auspices of the OECD, for the hazard identification and, where the evidence permits, the hazard characterisation and risk assessment of chemicals¹. An IATA integrates and weighs information from multiple sources — existing data from the scientific literature, (Q)SAR models, read-across, in chemico, in vitro, ex vivo and in vivo methods, and omics technologies such as toxicogenomics — in order to conclude on the toxicity of a chemical within a defined regulatory scenario or decision context². An IATA is not a single test method but an iterative strategy: existing information is assembled first, and new data are generated only where the available evidence is insufficient to answer the assessment question².

The biological basis of an IATA is the Adverse Outcome Pathway (AOP) concept, which organises toxicological knowledge into a causal sequence comprising a molecular initiating event (MIE), intermediate key events (KEs), and an apical adverse outcome, spanning multiple levels of biological organisation from the molecular through the cellular and tissue levels to the organism³. OECD guidance establishes the AOP as the systematic framework for characterising the biological relevance of individual alternative methods and for developing IATA around that mechanistic understanding³. The cellular and structural mechanics of an IATA therefore reside in the key-event-based test systems that it integrates. For skin sensitisation, the archetypal application, the validated information sources address discrete key events of the sensitisation AOP: in chemico peptide reactivity assays such as the Direct Peptide Reactivity Assay (DPRA) model the covalent binding of electrophilic substances to skin proteins, which constitutes the molecular initiating event^{4 5}; in vitro keratinocyte activation assays address the second key event; and dendritic cell activation assays address the third key event⁵.

What an IATA models, accordingly, is the causal chain from an initial molecular interaction to a decision-relevant apical hazard, rather than the intact-organism response of a single in vivo procedure. Its output is a regulatory conclusion — for example sensitiser versus non-sensitiser hazard, UN GHS potency subcategorisation, a prediction of the UN GHS eye hazard category, or a quantitative point of departure — derived from the integrated interpretation of all information sources^{5 6}.

Purpose

The primary value proposition of IATA is to replace the historical reliance on individual animal tests with integrated strategies whose conclusions are at least as protective, and frequently more informative, than those derived from conventional in vivo procedures⁷. The OECD frames these approaches as part of a broader modernisation of toxicity testing in which advances in biotechnology and computational modelling enable assessment that is faster, less expensive, and more relevant to human responses than traditional methods, while reducing the need for animal testing². An IATA achieves this by interrogating multiple, mechanistically anchored key events rather than depending on a single apical animal endpoint, and by requiring that new testing be undertaken only when existing data cannot resolve the assessment question^{3 2}.

The human relevance of the framework is twofold. First, the constituent information sources are frequently human-based systems — human-cell in vitro assays and acellular in chemico reactivity measurements — anchored to mechanistic knowledge rather than to species-specific apical responses⁵. Second, the integrated approaches are benchmarked against curated human reference data, notably human predictive patch-test data for skin sensitisation, rather than against animal data alone⁸. For the skin sensitisation defined approaches adopted in OECD Test Guideline 497, the OECD concluded that they provide either the same level of information or are more informative than the murine local lymph node assay (LLNA) for hazard identification⁵, and an independent evaluation of multiple defined approaches found that most non-animal strategies outperformed the LLNA in predicting human sensitisation hazard and potency⁴.

Beyond predictive performance, IATA were conceived to support harmonised regulatory decision-making: the OECD workshop that established the framework was organised in cooperation with the World Health Organization with the explicit goal of providing guiding principles for interpreting results from alternative approaches for hazard identification, hazard characterisation, and risk assessment across jurisdictions⁷. Standardised defined approaches adopted as OECD Test Guidelines are consequently intended to be accepted within the OECD Mutual Acceptance of Data system, ensuring that data generated in one member country satisfy requirements in the others and thereby reducing duplicate animal testing⁹.

Methodology

The construction of an IATA follows a modular, AOP-anchored methodology agreed at international level. First, the regulatory decision context and the assessment question are formulated, since an IATA is designed around a specific regulatory scenario^{2 7}. Second, the AOP relevant to the endpoint is identified, and individual information sources are selected and mapped to the key events of that AOP, so that each data stream carries defined mechanistic relevance^{3 5}. Third, existing data are collected and evaluated; only where these are inadequate is new information generated, most commonly through in vitro testing or computational models rather than animal testing². Fourth, the results of the selected information sources are combined through an integration procedure, which may be a flexible, expert-driven weight-of-evidence analysis or, in the most standardised subtype, a fixed data interpretation procedure (DIP)¹⁰.

The most prescriptive form, the Defined Approach (DA), consists of a fixed DIP — a mathematical or rule-based model — applied to data generated with a defined set of information sources, such that a result is derived without the need for expert judgement and may be used alone or within a broader IATA^{10 6}. For skin sensitisation, OECD Test Guideline 497 specifies DAs whose information sources include: in chemico reactivity assays addressing the molecular initiating event (DPRA and ARED… the Amino Acid Derivative Reactivity Assay, ADRA); in vitro assays addressing keratinocyte activation and dendritic cell activation (KeratinoSens, GARDskin, h-CLAT, U-SENS, and the IL-8 Luc assay); and in silico predictions from expert systems and category tools (Derek Nexus and the OECD QSAR Toolbox, which embeds the automated defined-approaches workflow)⁵. The guideline’s DIPs include the deterministic “2 out of 3” rule for hazard classification and the Integrated Testing Strategy (ITS) scoring model, which supports hazard and GHS potency conclusions, together with a defined approach (SARA-ICE) for deriving a quantitative point of departure for risk assessment⁵. A wider palette of integration techniques has been demonstrated in the peer-reviewed literature, including consensus methods, artificial neural networks, support vector machines, Bayesian networks, and decision trees, most of which were reproduced with open-source software⁴. For eye hazard, OECD Test Guideline 467 specifies DAs that combine in vitro data with physicochemical properties to predict the full range of UN GHS ocular categories⁶.

As a data-integration framework rather than a culture system, an IATA does not employ a single biological substrate; its substrates are the validated test systems of the underlying OECD Test Guidelines, and its protocols are the standardised guideline procedures, including proficiency testing with reference chemicals to demonstrate laboratory competence⁵. Documentation and quality are addressed through OECD reporting guidance, which provides templates for describing information sources, DIPs, training and optimisation sets, predictive performance, and applicability domains, so that defined approaches can be evaluated and reproduced transparently¹⁰. Weight-of-evidence integration, where flexibility is retained, follows the OECD guiding principles, which require a systematic, transparent assembly and weighing of all relevant evidence¹¹. The endpoint-specific IATA for serious eye damage and eye irritation illustrates the full architecture: Part 1 evaluates existing and non-testing data, Part 2 conducts a structured weight-of-evidence analysis, and Part 3 governs the generation of new testing data only where needed¹².

Performance

Empirical performance is best characterised for the skin sensitisation defined approaches. In the current edition of OECD Test Guideline 497, which permits permutations of alternative validated information sources, the “2 out of 3” defined approaches achieve balanced accuracies of 80–95% against LLNA-based reference data (sensitivities 82–97%) and 72–97% against human reference data (sensitivities 84–97%)⁵. The ITS defined approach achieves balanced accuracies of 78–88% against LLNA-based reference data and 67–76% against human reference data, with GHS potency accuracy of 66–80% against LLNA reference data and 54–75% against human reference data⁵. At the guideline’s initial adoption in 2021, the reference evaluation comprised approximately 170 substances with LLNA data and approximately 66 substances with human predictive patch-test data, and the original “2 out of 3” combination (DPRA, KeratinoSens, and h-CLAT) achieved 84% balanced accuracy against the LLNA reference set and 88% against the human reference set⁸. The same analysis showed that the LLNA itself, when used to predict human outcome, achieved only 58% balanced accuracy (94% sensitivity, 22% specificity) and 60% potency accuracy, providing the quantitative basis for the OECD’s conclusion that the defined approaches are as informative as or more informative than the animal assay for hazard identification^{8 5}.

Independent, open-source reproduction of the defined approaches, performed jointly by the Cosmetics Europe Skin Tolerance Task Force and the United States NTP Interagency Center for the Evaluation of Alternative Toxicological Methods (NICEATM) on a 128-substance dataset with curated LLNA and human reference data, confirmed the robustness of the integration concepts across six defined approaches and eight modelling variants⁴. NICEATM has additionally evaluated the Test Guideline 497 defined approaches on 185 substances nominated by United States agencies and developed a public web application (the DASS App) that automates the guideline’s data interpretation procedures, supporting consistent implementation across laboratories and regulators¹³.

Reproducibility is an intrinsic property of the framework: because a defined approach applies a fixed DIP to guideline-compliant data without expert judgement, its output is deterministic for a given set of inputs^{10 6}. Technical boundaries remain, and OECD Test Guideline 497 devotes a dedicated section to them: the applicability domains of the individual in chemico, in vitro and in silico information sources and of the defined approaches themselves must be respected, and each information source carries documented limitations⁵. Specificity estimates are less certain than sensitivity estimates because the reference datasets contain comparatively few negative substances⁵. In addition, several interpretation models were trained or optimised on LLNA data, which may negatively influence their performance against human potency categories⁴, and the point-of-departure defined approaches of Part III cannot substitute for the hazard and potency procedures of Parts I and II unless specifically stated⁵. For endpoints beyond skin sensitisation and eye irritation, validated defined approaches do not yet exist, and the OECD develops regulatory experience through its IATA Case Studies Project rather than through adopted guidelines¹⁴.

Applications

IATA are applied across the principal regulatory and industrial contexts of chemical safety assessment. The most mature application is skin sensitisation, where the OECD Test Guideline 497 defined approaches are used for hazard identification, UN GHS potency subcategorisation, and derivation of quantitative points of departure for risk assessment, particularly in sectors such as cosmetics and consumer-product ingredient assessment, where the underlying evaluation datasets were largely assembled and where non-animal approaches are the primary route to compliance^{5 4 13}. In the United States, the Environmental Protection Agency issued a science policy under which skin sensitisation data generated with defined approaches that predict human sensitisation with accuracy similar to existing in vivo methods are considered for regulatory use¹³. The second mature application is serious eye damage and eye irritation, where the defined approaches of OECD Test Guideline 467 predict the full range of UN GHS ocular hazard categories (Category 1, Category 2, or not classified) for liquid, solid and surfactant chemicals, within the structured three-part IATA of OECD Guidance Document 263^{6 12}.

The standardised contexts of use are defined by the OECD instruments themselves: data generated with Test Guideline 497 and Test Guideline 467 are eligible for acceptance under the OECD Mutual Acceptance of Data system across member countries, which constitutes the strongest available form of international regulatory standardisation^{9 6}. For endpoints where standardisation has not yet reached that stage — including chronic toxicity, carcinogenicity, and systemic effects — the OECD IATA Case Studies Project develops and peer-reviews worked examples that demonstrate predictions fit for regulatory use, together with consideration documents that distil the resulting guidance; a recent product of this project is the case study on the chronic toxicity and carcinogenicity of agrichemicals^{14 2}.

In terms of deployment stage and readiness, the framework is therefore tiered. Skin sensitisation defined approaches are fully deployed as an adopted Test Guideline, first issued in June 2021 and subsequently extended with additional information sources and new quantitative-risk-assessment defined approaches, most recently in 2026^{13 5}. Ocular defined approaches are similarly deployed through Test Guideline 467, adopted in 2022 and updated in 2025, supported by the third edition of the eye IATA guidance document published in December 2024^{6 12}. Other endpoints remain at the guidance and case-study stage, where the methodology is established but predictive frameworks have not yet completed validation and formal adoption¹⁴.

Variations

• Defined Approach (DA): The most prescriptive subtype, in which a fixed data interpretation procedure is applied to data from a defined set of information sources to derive a result without expert judgement; this is the only IATA subtype standardised as an OECD Test Guideline, for skin sensitisation (TG 497) and serious eye damage and eye irritation (TG 467)^{10 5 6}.
• Iterative Modular IATA: A flexible, expert-driven strategy in which existing data are weighed first and new information is generated stepwise only to resolve residual uncertainty; the three-part eye IATA of OECD Guidance Document 263, with its dedicated weight-of-evidence analysis, exemplifies this architecture^{12 2}.
• Read-Across IATA: A data-gap-filling application in which the toxicity of a target chemical is inferred from structurally or mechanistically analogous source chemicals within the integrated assessment; this approach is applied in the OECD case studies for systemic endpoints such as the chronic toxicity and carcinogenicity of agrichemicals^{2 14}.
• Computational IATA: An in-silico-led application in which (Q)SAR expert systems and category tools provide the decisive information streams; for example, Derek Nexus and the OECD QSAR Toolbox supply the in silico input of the ITS defined approach of OECD Test Guideline 497⁵.
• Weight-of-Evidence IATA: An application in which heterogeneous existing evidence is assembled and weighed systematically, without new testing, following the OECD guiding principles for weight-of-evidence assessment; this mode is formalised as Part 2 of the eye IATA guidance document^{11 12}.