Do FDA and ICH accept fish embryo assays in place of embryo-fetal toxicity studies?

Short answer: no, not as a stand-alone substitute. As of mid-2026, neither FDA nor ICH recognizes a fish embryo assay (or any non-mammalian in vivo model) as a validated drop-in replacement for the definitive mammalian embryo-fetal development (EFD) study. What does exist is a formal, conditional pathway: ICH S5(R3) allows "properly qualified" alternative assays, including non-mammalian in vivo assays, to defer or, in certain narrow circumstances, replace conventional in vivo studies, and FDA's draft new approach methodologies (NAMs) framework signals openness to fit-for-purpose alternatives when they are scientifically justified for a defined context of use ⁴⁷⁷⁹. The distinction that matters for a regulatory strategy is between a fully qualified replacement (which does not yet exist for fish/zebrafish EFD) and a qualified adjunct used within a weight-of-evidence or tiered testing strategy (which is where these assays actually sit today).

What ICH S5(R3) actually permits

ICH S5(R3) (Detection of Reproductive and Developmental Toxicity for Human Pharmaceuticals, Step 4, 18 February 2020) is the governing document, and it explicitly contemplates "alternative assays" spanning in vitro, ex vivo, and non-mammalian in vivo models for EFD risk assessment ⁴⁷. The guideline does not name zebrafish or fish embryo assays specifically, but the class of non-mammalian in vivo assays it describes would encompass them if they are appropriately qualified ⁴⁷.

The operative language is deliberately conditional. Qualified alternative assays "have the potential to defer or replace (in certain circumstances) conventional in vivo studies," and qualification is tied to a defined context of use: the assay's chemical applicability domain and the biological mechanisms it covers ⁴⁷. S5(R3) sets the bar at providing human safety assurance at least equivalent to the current testing paradigm, and it expects alternative assays to be deployed in a tiered or battery approach rather than as a single unqualified test ⁴⁷. Assays used only to explore mechanism of action, and not to substitute for in vivo EFD endpoints, do not require this rigorous qualification ⁹⁷.

Where a qualified assay can move the timeline is well defined. Building on ICH M3, S5(R3) states that preliminary EFD (pEFD) data from two species can support limited inclusion of up to 150 women of childbearing potential (WOCBP) for up to three months before definitive EFD studies. As an option, a qualified alternative assay that predicts the outcome in one species can be combined with a pEFD study from a second species to enable that same limited enrollment, with the assay and the second species generally covering one rodent and one non-rodent ⁴⁷. This is the concrete role a fish/zebrafish assay could play if qualified: it substitutes for one species in an early-enrollment strategy, not for the definitive EFD program.

The qualification bar in S5(R3) Annex 2

Annex 2 sets a high evidentiary standard for any predictive model offered in place of an in vivo endpoint. A sponsor must provide a full description and justification of the model, including which species and endpoints it predicts, the biological plausibility and developmental mechanisms covered, an explicit statement of limitations, evidence that exposure timing and duration support prediction of in vivo malformations/embryo-fetal lethality (MEFL), and an assessment of accuracy against in vivo data ⁴⁸. It also requires clear positive/negative decision criteria and thresholds, the algorithm for integrating assay outputs, the training and test sets with rationale for compound selection, the data sources for in vivo exposure and MEFL outcomes, coverage of appropriate biological and chemical domains, and metrics for sensitivity, specificity, positive and negative predictive value, and reproducibility ⁴⁸. Alternative assays supporting hazard identification or risk assessment are also expected to run under GLP and to be embedded in a justified integrated testing strategy, with metabolite coverage where warranted ⁹⁷. In practice, no fish embryo EFD assay has been shown to clear this bar for general replacement of the mammalian study.

FDA's position: NAMs, the 3Rs, and "fit-for-purpose"

FDA's stated direction is supportive of alternatives in principle but built on validation and context-of-use judgments rather than blanket acceptance. The most directly relevant document is the draft guidance "General Considerations for the Use of New Approach Methodologies in Drug Development" (18 March 2026), which frames NAMs as a way to improve human-predictive toxicology and reduce reliance on animal testing ⁷⁹⁸¹. Notably, the draft states that a test does not necessarily need to be formally validated to be considered in a review, and that a fit-for-purpose NAM may adequately address a specific toxicological concern even without full validation, while still describing validation (accuracy, reliability, relevance for a specific context of use) as critical to regulatory reliability ⁷⁹⁸².

The examples FDA cites of NAMs it has actually accepted as fit-for-purpose are telling: eye irritation, skin irritation, and skin sensitization endpoints, including in vitro reconstructed human epidermis assays that replace animal use in some skin sensitization contexts ⁸². These are comparatively simple, well-characterized endpoints, not developmental toxicity. Within the developmental/reproductive toxicity setting, the only zebrafish reference surfacing in FDA guidance is a study result cited in a residual-solvents context, not an endorsement of zebrafish as a substitute for standard EFD testing ⁶.

For the standard EFD package, FDA's expectations remain conventional and species-driven: rodent plus rabbit when both are pharmacologically relevant, or nonhuman primate studies (including a single enhanced pre- and post-natal development study from gestation day 20 to birth) when the NHP is the only relevant species ⁵¹⁵². FDA references alternatives only in limited, justified circumstances, for example transgenic mice expressing the human target when no relevant species exists, an alternative model in place of NHPs with adequate scientific justification, and "appropriately qualified/validated alternative tests" or weight-of-evidence approaches that may reduce or defer a study in a relevant animal species ⁵²⁵⁴⁵⁵.

Procedurally, FDA anchors alternative-method acceptance in the 3Rs (replace, reduce, refine) and directs sponsors to engage early, typically through the review division or the Q-Submission Program, with a method that is "suitable, adequate, validated, and feasible," after which FDA will assess whether the alternative can be treated as equivalent to the animal test method ^7898101. This same language recurs across device and drug guidances, which underscores that acceptance is made case by case against an equivalency standard, not by a general policy declaring fish embryo assays acceptable ^100102113.

Where fish and zebrafish assays actually stand

The scientific literature is consistent with the regulatory texts. Zebrafish developmental and MEFL assays are described as promising screening tools and components of weight-of-evidence or tiered strategies, not as full replacements for mammalian EFD studies ⁷¹¹. Recent work reports that zebrafish MEFL methods are being further validated for regulatory use under ICH S5(R3), but explicitly "in combination with information from other sources," and reviews position zebrafish embryo screens for the pre-regulatory phases of development as complementary to rodent models pending more validation ⁷⁸¹¹.

It is worth separating the pharmaceutical EFD question from the environmental one, because the terminology overlaps and can mislead. The Fish Embryo Acute Toxicity test (FET) is validated and adopted as OECD Test Guideline 236, but that guideline addresses acute toxicity to fish for chemical/environmental hazard assessment, not human developmental toxicity ¹⁰. Even in that environmental context, regulators have held that FET alone is "currently not sufficient" to satisfy acute fish toxicity information requirements under REACH and may be used only within weight-of-evidence approaches ²⁴. In other words, the one fish embryo assay with a formal OECD guideline behind it is an ecotoxicology tool, and its own regulatory acceptance remains partial; it is not a human EFD replacement.

Practical takeaways for regulatory teams

For a nonclinical or regulatory strategist, the current state can be summarized as follows. A fish or zebrafish embryo assay cannot today be submitted as a stand-alone substitute for the definitive mammalian EFD study and expected to satisfy ICH S5(R3) or FDA on its own ⁴⁷¹¹. Such an assay can, in principle, contribute within a defined and justified strategy: as a qualified alternative that replaces one species in a pEFD-plus-alternative-assay approach to support limited early WOCBP enrollment, or as one input in a weight-of-evidence risk assessment that FDA references as a potential way to reduce or defer animal studies ⁴⁷⁵⁵. Any such use turns entirely on meeting the S5(R3) Annex 2 qualification package (context of use, predictive performance metrics, training/test sets, GLP conduct) and on early alignment with FDA through the review division or Q-Submission Program under the fit-for-purpose and equivalency framework in the 2026 NAMs draft guidance ⁴⁸⁷⁹⁹⁸. The trajectory is toward broader acceptance of NAMs, but for embryo-fetal toxicity specifically, the qualified-adjunct model, rather than outright replacement, is the realistic near-term path.