FDA Preclinical Testing and Toxicity Endpoints for Large-Molecule Drugs, IND to BLA

ICH S6(R1) nonclinical safety expectations for biologics, IND to BLA

FDA evaluates the nonclinical safety of monoclonal antibodies (mAbs), Fc-fusion proteins, and other biotechnology-derived pharmaceuticals under ICH S6 (1997) and its 2012 addendum, S6(R1). ICH M3(R2) states that the appropriate nonclinical studies for biotechnology-derived products are set by S6, while M3(R2) governs when those studies are needed relative to clinical development ⁵¹. The result is a flexible, case-by-case framework. Program design depends on pharmacological relevance, target biology, and clinical context rather than a fixed small-molecule checklist.

This article summarizes the core expectations: species selection, general (repeat-dose) toxicity, safety pharmacology, immunogenicity in animals, immunotoxicity and cytokine release, reproductive and developmental toxicity (DART), genotoxicity and carcinogenicity, first-in-human (FIH) dose selection, and the oncology exceptions under ICH S9.

At a glance

Species selection: the foundation of the program

Pharmacological relevance

Toxicity testing should use a relevant species. That is a species in which the test article is pharmacologically active because the receptor or, for an mAb, the target epitope is expressed ¹⁹. Conventional rat and dog testing may be inappropriate because biologics often show species- and tissue-specific activity. Studies in nonrelevant species can be misleading and are discouraged ¹⁹.

S6(R1) calls for a weight-of-evidence assessment of relevance, including ²⁰:

• target sequence homology as an initial screen;
• cross-species comparison of binding affinity, receptor or ligand occupancy, and binding kinetics;
• functional activity in species-specific cell systems and/or in vivo pharmacology;
• modulation of a known biological response or pharmacodynamic (PD) marker; and
• whether the species, at the intended regimen, can reveal adverse consequences of target modulation.

When the target is expressed at very low levels in healthy animals, cell-based binding and functional data can be enough to guide selection ²⁰.

One species versus two

The default is two relevant species ¹⁹. If both a rodent and a nonrodent are relevant, both should be used in short-term general toxicology studies of up to one month ¹⁶⁴. When the findings are similar, or can be explained by the mechanism of action, longer-term studies in one species are usually sufficient. The rodent should be the default unless there is a scientific reason to use a nonrodent ¹⁶⁴. S6(R1) states that two nonrodent species are not appropriate. When the clinical candidate is active in only one species, that species alone is justified for all general toxicity studies. A second-species study with a homologous product is not recommended because it adds no value to the risk assessment ¹⁶⁴.

Tissue cross-reactivity

For mAbs, a relevant species should express the desired epitope and show a tissue cross-reactivity (TCR) profile similar to human tissues ¹⁹. A TCR study using a human tissue panel is a recommended part of the package supporting the initial clinical dose for mAbs and antibody-like products. TCR is usually run by immunohistochemistry ²². Tissue binding alone does not show in vivo activity, and binding at sites an antibody cannot reach in vivo, such as cytoplasm, is generally not relevant ²². Animal-tissue TCR has limited value for species selection and is not recommended for that purpose ²⁰. Relevance is established with binding, functional, and pharmacodynamic data instead. For bispecific antibodies, TCR should be run with the clinical candidate, not the individual binding arms ²².

When no relevant species exists

When no relevant species is available, sponsors should consider transgenic animals expressing the human target, or a homologous (surrogate) protein against the animal ortholog ²⁴¹⁶⁴. Homologous proteins can support hazard identification and help explain toxicity from exaggerated pharmacology. They generally do not support quantitative risk assessment ¹⁶⁴. They may also differ from the clinical product in manufacturing process, impurity profile, pharmacokinetics (PK), and mechanism ¹¹³. If neither approach is feasible, S6 notes that a limited single-species evaluation may be prudent. An example is a repeat-dose study of 14 days or less that includes functional endpoints such as cardiovascular and respiratory endpoints ¹¹³.

mAbs against foreign targets (bacterial or viral) are a distinct case. A short-term safety study in one species may be considered, with a justified choice of species, and no further toxicity studies, including reproductive toxicity, are appropriate. Alternatively, safety endpoints can be built into a disease-model proof-of-concept study ¹⁶⁴.

Repeat-dose (general) toxicity

Duration

For chronic-use products, S6(R1) considers 6-month repeat-dose studies in either rodents or nonrodents sufficient, provided the high dose follows S6(R1) dose-selection principles. Longer studies generally have not produced information that changed clinical development ³. Advanced-cancer products follow ICH S9 instead ³.

To align with the clinical program, M3(R2) sets the general-toxicity durations that support marketing. The table below is the general M3(R2) duration table. For biologics it is timing guidance only; species number and study design follow S6(R1), so the two-species and 9-month cells are not requirements for a biologic:

Source: ⁹². For chronic biologics, S6(R1) replaces the 9-month nonrodent study with the 6-month single-species study described above ³.

Dose selection

The rationale for dose selection should draw on the dose-response relationship and PK/PD. The S6(R1) high dose is the higher of ³:

1. the dose producing the maximum intended pharmacological effect in the relevant species; or
2. the dose producing about a 10-fold exposure multiple over the maximum anticipated clinical exposure.

A lower high dose can be justified, for example by the maximum feasible dose (MFD) ³. If no PD endpoints exist, dose selection can rely on PK and in vitro binding and pharmacology data, adjusted for species differences in affinity and potency ³. If no toxicity is seen at doses chosen this way, testing at higher multiples is unlikely to add useful information ³. Where feasible, studies should define a toxic dose and a no-observed-adverse-effect level (NOAEL). For products with little or no toxicity, sponsors should give a scientific rationale for the doses and projected human exposure multiples ².

Recovery

When effects with potential clinical impact occur at clinically relevant exposures, sponsors should understand whether they reverse. This can come from existing scientific knowledge, or from a nondosing recovery period in at least one study at one or more justified dose levels ³. The aim is to assess reversibility, not delayed toxicity, and complete recovery does not need to be shown. A recovery period solely to evaluate immunogenicity is not required ³.

Route, formulation, local tolerance, and exposure

Repeat-dose studies should use a clinically relevant route and regimen and, where feasible, include toxicokinetics ²⁵. Systemic exposure should be monitored in toxicity studies whenever possible ³⁹. Routine mass-balance studies are not useful for biologics. Single- and multiple-dose PK, TK, and tissue-distribution studies in relevant species are the useful approach ³⁹. Immune-mediated clearance can change PK profiles and complicate interpretation ³⁹. Local tolerance should be evaluated with the intended marketing formulation, or a justified representative formulation. It can often be assessed within single- or repeat-dose studies rather than in separate studies ⁴.

Safety pharmacology

ICH S7A applies to biotechnology-derived products as well as small molecules ¹⁰². S6 expects functional assessment of major physiological systems, particularly cardiovascular, respiratory, renal, and central nervous systems. These assessments can be run as separate studies or built into toxicity and/or clinical studies ³⁹. For biotechnology-derived products with highly specific receptor targeting, S7A states that evaluating safety pharmacology endpoints within toxicology and/or PD studies is often sufficient. Dedicated studies can therefore be reduced or eliminated ¹⁰⁴. A more extensive evaluation should be considered for products in a novel therapeutic class and for products without highly specific targeting ¹⁰⁴.

S7B covers nonclinical assessment of delayed ventricular repolarization (QT prolongation) ¹⁰⁰. The guidance reviewed here does not contain protein-specific S7B triggers for hERG or dedicated in vivo QT studies. In practice, cardiovascular endpoints (for example, ECG in nonrodent repeat-dose studies) are normally captured within the integrated S6/S7A approach ³⁹¹⁰⁴.

Immunogenicity in nonclinical studies

Under S6(R1), anti-drug antibody (ADA) measurement in animals helps interpret study findings and plan later studies. It is not meant to predict human immunogenicity ¹⁶⁰. Animal immunogenicity does not predict antibody formation against human or humanized proteins in patients ¹⁶⁰. Positive guinea pig anaphylaxis tests for protein products do not predict human reactions ².

Key expectations:

• When to assess ADA: when PD activity is altered, when exposure changes unexpectedly and no PD marker is available, or when there is evidence of immune-mediated reactions such as immune complex disease, vasculitis, or anaphylaxis ¹⁶⁰. The need may only become clear after the in-life phase, so samples should routinely be collected and banked ¹⁶⁰.
• Characterization: titer, number of responding animals, and neutralizing versus non-neutralizing activity, correlated in time with pharmacological and toxicological changes ². Where there is no PD marker showing sustained activity, neutralizing potential can be assessed indirectly (ex vivo bioactivity or PK/PD assays) or directly with a neutralizing antibody assay ¹⁶⁰.
• Interpretation: consider effects on PK/PD, exposure, the incidence and severity of findings, complement activation, and pathology from immune complex deposition ².
• Study conduct: finding ADA alone should not usually lead to early termination or a change in study duration. Such changes may be justified when neutralization abolishes the pharmacological or toxicological effects in a large proportion of animals ².

Clinical immunogenicity risk is assessed separately. FDA's therapeutic protein immunogenicity guidance, for example, recommends monitoring immunomodulatory properties from the earliest stages of development, including effects on responses to the product itself and the potential to induce autoimmunity ¹⁴⁹.

Immunotoxicity and cytokine release

ICH S8 does not apply to biotechnology-derived products covered by S6 ¹⁵². For biologics, routine tiered immunotoxicity testing or standard test batteries are not recommended. Screening studies followed by mechanistic studies may be appropriate ¹. FDA's immunotoxicity guidance recommends a risk-based, weight-of-evidence approach. General toxicity studies serve as the first-tier screen, and focused immune-function studies follow if a hazard is identified ¹⁵⁰. For immunomodulators, the role of the target in immune function should be considered ¹⁵⁰. If there is a concern for immunosuppression, a focused host resistance model may be considered after discussion with the review division ¹⁴⁷.

For immunostimulatory products (T-cell engagers, checkpoint inhibitors, receptor agonists), in vitro assays should be chosen based on mechanism of action, target expression, and the potential to directly stimulate key cell types ¹⁴⁷. A positive cytokine release assay does not necessarily stop development. The results can inform FIH dose selection, clinical monitoring, dose-escalation rules, and stopping criteria ¹⁴⁸. For bispecific antibodies, homodimers of anti-CD3 or Fc-engaging arms may cause cytokine release, so their relative amounts should be assessed ³¹.

First-in-human starting dose

FDA's MRSD guidance describes the standard approach. The NOAEL from the most appropriate (pharmacologically relevant) species is converted to a human equivalent dose (HED), and a safety factor, generally at least 10, is applied ²⁹⁵⁹. The MRSD should then be compared with the pharmacologically active dose (PAD). For mAbs and growth factors, where toxicity may reflect exaggerated pharmacology, the PAD may be more sensitive and may justify a lower starting dose ⁶². For large humanized mAbs given intravenously, allometric, PK, and PD modeling can help translate safe nonhuman primate exposures to humans. Safety factors are still needed because receptor sensitivity or density may differ between species ⁵⁸.

For high-risk immunostimulatory products, a minimum anticipated biological effect level (MABEL) or PAD-based starting dose may be more appropriate than a NOAEL-based dose, particularly when relevant animal models are limited ¹⁴⁸. FDA recommends considering a MABEL-based initial dose for bispecific antibodies with agonistic properties ⁵⁴. For receptor-targeting mAbs with cytokine release potential, cellular activation and cytokine release should be evaluated in human whole blood or PBMCs. Negative conventional animal findings do not rule out the risk ⁷¹. A 2026 FDA draft guidance describes MABEL estimation informed by quantitative systems pharmacology ⁶⁹⁶⁶.

Reproductive and developmental toxicity

General approach

Under S6(R1), the need for DART studies depends on the product, the indication, and the intended patient population. Study designs can be modified to account for species specificity, immunogenicity, biological activity, and long elimination half-life ¹¹⁴¹.

Fertility

When the nonhuman primate (NHP) is the only relevant species, mating studies are not practical. Fertility effects are instead assessed through reproductive organ weights and histopathology in repeat-dose studies of at least 3 months in sexually mature NHPs ¹⁷¹. If there is a specific concern, menstrual cyclicity, sperm count, morphology and motility, and reproductive hormones can be added. Concerns about conception or implantation may require a homologous product or transgenic model ¹⁷¹. M3(R2) also recognizes that a repeat-dose study of at least 2 weeks with thorough histopathology of the testes and ovaries is as sensitive as fertility studies for detecting effects on reproductive organs ⁸³.

Embryo-fetal development and species choice

If the clinical candidate is active in rodents and rabbits, EFD studies should be conducted in both species, unless embryo-fetal lethality or teratogenicity has already been identified in one ¹⁷¹. NHP developmental studies should be run only when the NHP is the sole relevant species. An alternative model can replace NHPs with appropriate scientific justification ¹⁷¹.

Placental transfer of IgG and the ePPND design

Placental transfer of IgG through FcRn shapes the design of mAb DART studies. In NHPs and humans, IgG transfer is low during organogenesis, rises from the early second trimester, and is highest late in the third trimester ¹³⁷. A conventional NHP EFD study dosed from early pregnancy through gestation day 50 therefore has limited value for detecting direct embryo-fetal effects during organogenesis. It can still show effects secondary to maternal toxicity ¹³⁷. Rodents differ because FcRn-mediated transfer occurs through the yolk sac and can happen earlier in gestation ¹³⁷.

The NHP enhanced pre- and postnatal development (ePPND) study combines EFD and PPND endpoints, with dosing through gestation to parturition ¹³⁸. S6(R1) specifies the following for ePPND ¹³⁷:

• postnatal follow-up of at least one month for early functional endpoints such as growth and behavior;
• offspring immune function testing when general toxicology studies show immune effects, with immunophenotyping possible from postnatal day 28 and follow-up of 3 to 6 months depending on the test;
• neurobehavioral evaluation that can be limited to clinical behavioral observations (instrumental learning testing, which needs at least 9 months of follow-up, is not recommended).

For rodents, dams should be dosed during lactation so that pups are exposed through milk until at least lactation day 9 ¹³⁷. mAbs with prolonged immunological effects may raise developmental immunotoxicity concerns that call for modified designs ¹.

Timing relative to clinical trials

M3(R2) sets out the timing of DART studies relative to clinical development ⁷⁴⁷⁵:

• For mAbs where embryo-fetal exposure during organogenesis is understood to be low in humans, developmental toxicity studies may be conducted during Phase 3, with completed reports submitted in the marketing application.
• In the US, EFD assessment for women of childbearing potential (WOCBP) may be deferred until before Phase 3 when pregnancy precautions are used.
• Male and female fertility studies should be completed before large-scale or long-duration trials (for example, Phase 3).
• PPND data (for biologics, often the NHP ePPND) should be submitted for marketing approval.

Genotoxicity and carcinogenicity

The standard genotoxicity battery generally does not apply to biotechnology-derived products and is not needed. Large quantities of peptide or protein can also produce uninterpretable results ¹. Studies are warranted only when there is cause for concern, such as an organic linker in a conjugated protein ⁴. Standard genotoxicity assays are not considered appropriate for evaluating process contaminants ⁴.

Standard rodent carcinogenicity bioassays are generally inappropriate ⁴. Instead, S6(R1) calls for a product-specific assessment that considers clinical duration, the patient population, and biological activity such as growth factor or immunosuppressive effects ⁴. The assessment can draw on target biology, receptor expression in normal and malignant human cells, and cell proliferation endpoints from chronic toxicity studies ⁴. If the weight of evidence does not suggest carcinogenic potential, no additional nonclinical testing is recommended. If it does, the sponsor may propose studies or address the concern in labeling ⁹. Rodent bioassays or short-term carcinogenicity studies with homologous products are generally of limited value ⁹. The conclusions feed into labeling, clinical monitoring, and postmarketing risk management ⁹.

Oncology exceptions under ICH S9

For biotechnology-derived anticancer products in patients with advanced cancer, including mAbs and antibody-drug conjugates, the program is abbreviated:

• Duration: 3-month repeat-dose studies on the intended clinical schedule should be available before Phase 3 and are generally sufficient to support marketing ⁷⁸. S6(R1) defers to S9 for these products ³.
• Species: testing in one pharmacologically relevant species is generally acceptable ¹²⁶. S9 indicates that TCR studies for mAbs and ADCs generally have little utility absent a specific concern ¹²³.
• DART: an EFD assessment is needed to support the marketing application, but dedicated fertility and PPND studies are generally not warranted. For biopharmaceuticals, EFD can typically be assessed in a single relevant species ¹¹⁹¹²⁶. A weight-of-evidence approach, which may include a surrogate molecule where the animal target biology is relevant to humans, can be used ¹³⁹. Fertility or PPND studies may be considered case by case if development moves into adjuvant, neoadjuvant, or other non-advanced settings ¹²⁵.

What FDA expects in the IND submission itself

Under 21 CFR 312.23(a)(8), the pharmacology/toxicology section of an original IND must contain ¹⁷²:

• A description of pharmacologic effects, mechanism of action, and ADME in animals (a summary report usually suffices; state what is unknown).
• An integrated summary of the toxicologic effects in animals and in vitro, with the specific studies depending on the product and the clinical phase.
• Full data tabulations for each toxicology study supporting the proposed trial, suitable for detailed review.
• A GLP compliance statement for each study subject to GLP, or an explanation of any noncompliance.

The integrated toxicology summary should describe study designs and dates, present findings systematically across body systems, flag signals an informed expert would view as possible human risk (with reasons where a signal is judged not to indicate human risk), identify and give the qualifications of the person who concluded it is reasonably safe to begin, and state where studies were conducted and records are held. Final quality-assured study reports should be available within 120 days of starting the human study ¹⁷².

Practical takeaways for program design

• Establish species relevance early with binding, functional, and PD data. The choice of species drives every downstream study, including DART feasibility ²⁰¹⁶⁴.
• For most chronic-use mAbs and fusion proteins, plan short-term (up to 1 month) studies in two relevant species if available, then a 6-month study in one species, using S6(R1) high-dose logic and a justified recovery arm ³¹⁶⁴.
• Build safety pharmacology endpoints into the repeat-dose studies, and bank ADA samples in every study ¹⁰⁴¹⁶⁰.
• Replace default genotoxicity and carcinogenicity studies with a documented weight-of-evidence assessment ⁴⁹.
• For NHP-only programs, plan fertility endpoints within a 3-month or longer study in sexually mature animals, and an ePPND study timed to M3(R2) milestones ^17113775.
• For immunostimulatory mechanisms, run human in vitro cytokine release assays and consider a MABEL- or PAD-based FIH dose ¹⁴⁸⁵⁴.

Open questions that often need product-specific analysis include pediatric and juvenile toxicity (ICH S11), comparability after manufacturing changes, bispecific-specific designs, and how FDA divisions have applied these principles in specific BLA reviews.