Nonclinical Safety for ADCs and Biologics When No Pharmacologically Relevant Animal Species Exists

Nonclinical safety when no relevant species exists: how FDA and ICH handle ADCs and other biologics

For most small molecules, a rat and a dog will metabolize and respond to the drug closely enough that their toxicology data transfer to humans. Biologics break that assumption. A monoclonal antibody or an antibody-drug conjugate (ADC) is engineered to engage a specific human epitope, and if the animal ortholog of that target is absent, divergent, or unbound, a conventional toxicology study measures the wrong thing. The regulatory response, built over two decades of ICH harmonization and refined in individual biologics license application (BLA) reviews, is a structured decision tree: establish species relevance on pharmacologic grounds first, and only then decide whether a single relevant species, a homologous or surrogate molecule, a transgenic model, or a weight-of-evidence argument will carry the safety case.

The ICH starting point: relevance is a pharmacology question, not a taxonomy question

ICH S6(R1), the preclinical safety guideline for biotechnology-derived pharmaceuticals, defines a relevant species as one in which the test material is pharmacologically active because the receptor or epitope is expressed ⁵¹. The guideline lays out an explicit hierarchy for demonstrating that: first assess target sequence homology, then use in vitro assays to compare target binding affinity, receptor/ligand occupancy, and kinetics, and then confirm functional activity in species-specific cell systems or in vivo ⁴⁸. For monoclonal antibodies specifically, a relevant species is one that expresses the desired epitope and shows a tissue cross-reactivity profile similar to humans ⁴⁹.

The number of species follows from that relevance analysis, not from a fixed rodent-plus-non-rodent template:

Where two pharmacologically relevant species exist, both are used for short-term general toxicology; if the findings are consistent or mechanistically understood, one species usually suffices for longer-term studies ⁴⁸.
Where only one relevant species exists, a safety program in that single species is acceptable when justified ⁴⁹.
Where no relevant species exists, S6(R1) directs sponsors to consider transgenic animals expressing the human target or homologous proteins, and studies in non-relevant species are discouraged because they can mislead ⁴⁹⁴⁸⁵³.

The guideline is candid about the limits of these workarounds. Homologous proteins and surrogate models are useful for hazard identification and for understanding exaggerated pharmacology, but they are generally not suitable for quantitative risk assessment, and animal disease models may sometimes substitute for normal-animal toxicity studies when scientifically justified ⁴⁹. In the developmental and reproductive setting, S6(R1) allows a homologous product or transgenic model as the only practical means of assessing endpoints such as conception and implantation, and it accepts that where none of these tools is available, in vivo reproductive testing is not meaningful and the risk assessment should rest on a documented scientific rationale ⁵³⁵⁴.

ICH S9 and the specific logic for conjugated products

For oncology products, ICH S9 governs, and it defers to S6 on the number of species while adding that one pharmacologically relevant species is usually sufficient for reproductive toxicity assessment of biopharmaceuticals ². S9's most directly relevant content is its dedicated section on conjugated products, which it defines as pharmaceuticals covalently bound to carriers such as proteins, lipids, or sugars, the category that captures ADCs. The guideline sets three principles that shape every ADC program:

The safety of the conjugated material (the intact ADC) is the primary concern, while the unconjugated components, including the linker, can receive a more limited evaluation ¹.
Stability of the conjugate should be characterized in both the test-species and human plasma ¹.
Toxicokinetic assessment should measure both the conjugated and the unconjugated species after dosing the conjugate ¹.

This framing matters when no relevant species exists, because it tells reviewers that much of an ADC's toxicity is expected to come from the cytotoxic payload rather than from antibody-target engagement, which in turn opens the door to rodent-based and payload-focused approaches even when the antibody does not bind the animal target.

FDA guidance: a weight-of-evidence pathway built specifically for ADCs

FDA has translated the ICH principles into an ADC-specific document, the guidance on Clinical Pharmacology Considerations for Antibody-Drug Conjugates ⁷⁵⁷⁷⁷⁹. Its species-selection recommendations read as a graded scheme keyed to whether a relevant species can be found:

Animal toxicology studies should use species in which the ADC binds the target and elicits the intended pharmacology ⁹⁰.
If there is no pharmacologically relevant species, sponsors should substitute a weight-of-evidence (WoE) risk assessment for animal toxicology studies rather than run studies in an irrelevant species ⁹⁰.
If the ADC is pharmacologically active in both a rodent and a non-rodent, general toxicology may be conducted in a single rodent species, supplemented by a WoE assessment as appropriate ⁹⁰.
For ADCs with cytotoxic payloads whose safety is well characterized and which drive the observed toxicities, the 3-month toxicology study may be run in rodents only, regardless of whether the ADC binds the target antigen; if the target is novel and not bound in rodents, a WoE risk assessment should accompany it ⁹⁰.

Critically, FDA states that species selection for an antibody-drug/toxin conjugate follows the same general principles as for the unconjugated antibody, so the choice turns on whether the antibody is pharmacologically relevant in the species, and it notes that where a targeting moiety adds species specificity, both species selection and any surrogate design need careful consideration ³⁷⁸⁹. Where only one pharmacologically relevant species is available, the ADC should be tested in that species ³⁸.

The ADC guidance sits on top of FDA's adoption of the ICH S6(R1) addendum, which restates the core rules for biologics generally: base species selection on pharmacologic activity and human/animal target-binding relevance; use a single species for all general toxicity studies when the product is active in only one; use one rodent and one non-rodent for short-term work when two relevant species exist (but never two non-rodent species); and, when no species interacts with the orthologous target, consider homologous molecules or transgenic models, with the caveat that homologous proteins detect hazard and characterize exaggerated pharmacology but are generally not useful for quantitative risk assessment ³⁷. Adjacent FDA guidances round out the picture: the immunotoxicology guidance asks sponsors to pick a test species appropriate for the endpoints and to integrate findings through a weight-of-evidence approach that weighs the target's role in immune function, dosing across the relevant developmental window, and structural similarity to known immunomodulators ³¹³³; the MABEL/QSP first-in-human guidance stresses conservative, low-end starting doses for novel targets where animal data are weakest ³⁰; and the S12 gene-therapy guidance carries the same conceptual anchor of a biologically relevant species defined by similarity to human tissue distribution and expression ⁴².

How the framework actually resolves in ADC BLA reviews

FDA pharmacology/toxicology reviews for approved ADCs show a consistent pattern that operationalizes the ICH and FDA guidance. The recurring design is a cross-reactive non-human primate paired with a non-cross-reactive rodent, plus the use of non-binding conjugate controls to separate payload toxicity from target-mediated toxicity.

Kadcyla (ado-trastuzumab emtansine). Studies ran in rats and cynomolgus monkeys, with the monkey judged the more relevant species because trastuzumab bound the cynomolgus HER2 receptor with affinity similar to human but did not bind the rat receptor with high avidity. Rat studies with the ADC or with DM1 alone produced similar toxicities, which FDA read as evidence the toxicity was mediated primarily by the DM1 maytansinoid payload ¹⁵¹.
Adcetris (brentuximab vedotin). Cynomolgus monkey was chosen because pharmacology and tissue cross-reactivity studies showed CD30 binding in monkeys and humans at similar affinities. In monkeys, hematopoietic toxicity was treated as target-mediated because the ADC binds CD30; in rats, which do not bind the target, the same class of toxicity was attributed to the MMAE payload rather than to target engagement ¹⁴⁷¹⁴⁸.
Enhertu (fam-trastuzumab deruxtecan). Cynomolgus monkey served as the cross-reactive relevant species because the ADC bound humanized HER2 in monkeys and humans but not in mice or rats; rats were then used deliberately as the non-cross-reactive species to characterize target-independent effects, with the deruxtecan payload driving findings in intestine, lymphatic/hematopoietic organs, kidney, testes, skin, and teeth ¹⁵⁰.
Besponsa (inotuzumab ozogamicin). The cynomolgus monkey was the relevant species based on CD22 binding. FDA attributed the hepatotoxicity to the calicheamicin payload by pointing to a 9-week monkey study with a non-binding ADC that produced similar liver findings, isolating the payload as the cause rather than CD22 engagement ¹⁵⁴.
Trodelvy (sacituzumab govitecan). A dedicated cross-reactivity study of the antibody in normal cynomolgus tissues supported the monkey as the Trop-2-relevant species for repeat-dose toxicology, while the genotoxic/clastogenic SN-38 payload also informed the toxicity assessment, a both/and outcome rather than reliance on payload data alone ¹⁹⁸²⁰³.
Padcev (enfortumab vedotin). Comparable Nectin-4 binding across human, rat, and cynomolgus monkey made both the rat and the monkey pharmacologically relevant for the antibody component, while MMAE was treated as a target-independent toxic payload evaluated on its own terms ²²⁰²²²²²⁴²²⁶²²⁷.
Blenrep (belantamab mafodotin). Repeat-dose studies ran in rats and monkeys, and the review tied the genotoxic, aneugenic signal to the pharmacology of the MMAF payload binding tubulin and depolymerizing microtubules during cell division ¹⁵⁵¹⁶⁴.

The common thread is that reviewers do not force a study into an irrelevant species. They locate the species where the antibody is genuinely active (almost always the cynomolgus monkey for these targets), use a non-binding rodent or a non-binding conjugate as an internal control to attribute effects to the payload, and lean on the well-characterized payload class to support rodent-based or reduced programs, exactly the logic the ADC guidance and S9's conjugated-products section anticipate.

Surrogate molecules and transgenic models in practice

For monoclonal antibodies where no animal ortholog is bound, FDA reviews document the full menu of S6(R1) tools, along with the limitations reviewers attach to each. These non-ADC examples are the clearest illustration of how "no relevant species" is handled:

Stelara (ustekinumab). Because the human antibody did not bind mouse IL-12 or IL-23, the sponsor used a chimeric surrogate antibody, CNTO 3913, against the mouse cytokines to enable rodent studies, supporting cynomolgus monkey and mouse as toxicology species ¹⁷⁷.
Kevzara (sarilumab). With the mouse not pharmacologically relevant, the sponsor used a surrogate antibody, REGN844, that binds mouse IL-6Rα for the mouse repeat-dose toxicology study ¹⁹².
Denosumab (Prolia/Xgeva). Denosumab is specific to human and non-human primate RANKL and inactive in rodents, so cynomolgus monkey was the only relevant species; the sponsor also built a human-RANKL knock-in mouse for pharmacology and efficacy work, which FDA noted was not appropriate for carcinogenicity studies ¹⁸³¹⁹¹¹⁹⁶.
Infliximab products (Inflectra, Avsola). Infliximab cross-reacts only with human and chimpanzee TNFalpha, so rats were not pharmacologically relevant and rat studies addressed off-target effects only; FDA observed that Tg197 transgenic mice expressing human TNFalpha would have been more informative, while also cautioning that a transgenic rat producing human TNFalpha could generate difficult-to-interpret data from unphysiologic, over-abundant expression of the human cytokine ¹⁷³¹⁹⁵.
Actemra (tocilizumab). Tocilizumab does not bind rat IL-6R, making a rat fertility study uninformative, and the reviewer recommended developing a species-specific anti-IL-6R surrogate antibody for reproductive and carcinogenicity work ¹⁹⁷.
Enjaymo (sutimlimab). The cynomolgus monkey was pharmacologically relevant but no relevant rodent species was identified, so the program was conducted in the monkey alone ¹⁷⁹.
Tezspire (tezepelumab) and Gamifant (emapalumab). Both illustrate the upstream binding screen that drives the decision: tezepelumab did not bind rat or rabbit TSLP and was tested in cynomolgus monkey ¹⁷⁴, and emapalumab bound cynomolgus and rhesus but not dog, rat, or mouse, again leaving the monkey as the sole relevant species ¹⁷⁵.

Two limitations recur across these reviews and are worth carrying into any program planning. First, a surrogate molecule characterizes the biology of the pathway but is not the clinical product, so it supports hazard identification rather than quantitative human risk estimates, consistent with S6(R1)'s own caveat on homologous proteins ³⁷¹⁷⁷¹⁹². Second, transgenic and knock-in models carry interpretive baggage: non-physiologic expression of the human target can distort findings, and such models may be unsuitable for specific study types such as carcinogenicity ¹⁷³¹⁹¹.

What this means for a nonclinical program

The practical sequence a reviewer expects is stable across ICH and FDA guidance and visible in the approval record. Start with a rigorous, documented species-relevance assessment built on sequence homology, in vitro binding and functional data, and tissue cross-reactivity ⁴⁸⁴⁹. If one relevant species exists, use it, and justify the single-species program ⁴⁹³⁸. If two exist, pair a rodent and a non-rodent for short-term work and consolidate to one for longer studies where findings converge ⁴⁸³⁷. If none exists, choose deliberately among a homologous or surrogate molecule, a transgenic or knock-in model, or, for ADCs specifically, a rodent-plus-payload weight-of-evidence package, and state plainly what each approach can and cannot support ⁴⁹⁹⁰³⁷. For ADCs, the S9 conjugated-products logic and FDA's ADC guidance make the payload the analytic center of gravity: characterize conjugate stability and toxicokinetics of both bound and free species, use non-binding conjugate controls to attribute effects, and lean on a well-understood payload class to justify a reduced or rodent-only program when target binding is absent ¹⁹⁰. Where the animal package is weakest, as with wholly novel targets and no relevant species, that gap is expected to be met on the clinical side through conservative, MABEL-anchored first-in-human dosing ³⁰.