1.3 Test System Selection, Species Models & Animal Welfare (3Rs, IACUC)

Why species choice is a DABT design skill

Handbook I.A.1 E–F pairs principles of the 3Rs with selection of an appropriate species. Wrong species is not a husbandry footnote. It changes metabolites, portal-of-entry dosimetry, background pathology, and whether a finding is interpretable for humans. DABT items often present a “default” protocol that ignores metabolism or animal numbers that ignore welfare rules.

Default models and when to leave them

The rat is the usual default rodent for general toxicity. The mouse is not the default 90-day oral rodent simply because it is smaller and cheaper. Mice earn their keep in carcinogenicity (often rat plus mouse), in knockout or transgenic mechanistic work, and in some host-resistance/immunotox designs.

Pharmaceuticals generally need a rodent plus a nonrodent. The nonrodent is not automatically the dog. If the dog does not express the target, vomits the dose, or makes a metabolite pattern unlike humans, NHP or minipig may be the better second species. If a monoclonal antibody binds chimpanzee and human but not dog or rodent Fc receptors in a relevant way, two irrelevant species do not become relevant by being GLP.

Rabbit is the standard second species for embryo-fetal developmental (EFD) studies under OECD 414 and ICH S5, alongside a rodent (often rat). That is anatomy and history, not a claim that rabbits are “more human.” Minipig is not the default EFD species.

Metabolism and anatomy: why the table is not trivia

Species selection is an ADME and comparative-anatomy problem:

• Biotransformation. Dogs acetylate poorly relative to humans for some arylamines. Rodents may be high-capacity peroxisome proliferator-activated receptor alpha (PPAR-alpha) responders, producing liver tumors with weak human relevance. Extrahepatic CYPs in mouse club cells can drive lung toxicity that does not scale to human bronchioles. If human hepatocytes or 14C-ADME already show a human-unique metabolite, a second species that never makes it will not “cover” that metabolite.
• Gastrointestinal anatomy. The rat has no gallbladder. Rodents have a forestomach lined by squamous epithelium; forestomach irritation after gavage is a recurring human-relevance debate. Dogs empty the stomach and vomit more readily than rats, capping oral exposure.
• Reproductive physiology. Rodents have a short estrous cycle; Old World NHPs have a menstrual cycle closer to humans for some reproductive questions. Rabbit EFD uses a different implantation timing than rat; you cannot copy-paste gestation-day windows.
• Respiratory anatomy. Rats and mice are obligate nasal breathers with a large olfactory epithelium. A nose-only aerosol that injures rodent olfactory mucosa may over-predict human nasal injury or under-predict pulmonary deposition, depending on particle size. That is a species-plus-route problem, not “the rat was wrong” in the abstract.
• Dermal barrier. Rabbit and rodent skin is often more permeable than human skin. A minipig or human explant may be more informative for a topical product than a default rat 90-day oral.

If the finding that will drive the risk assessment is species-specific anatomy (forestomach, olfactory epithelium, PPAR-alpha liver), say so in the protocol objective instead of discovering it at the pathology peer review.

IACUC and the ethics layer

An Institutional Animal Care and Use Committee (IACUC) reviews the protocol before animals are ordered. In the United States, the Animal Welfare Act covers dogs, rabbits, NHPs, minipigs, and other listed species; purpose-bred laboratory rats and mice used in research are not AWA-covered in the same way but are covered by Public Health Service Policy when PHS funds are involved and by institutional AAALAC-type standards in most GLP facilities. Practically, a modern toxicology protocol still goes to IACUC for all vertebrate species.

IACUC will ask:

• Why this species and not a phylogenetically lower or non-animal system (the Replace search).
• Why this number (Reduce: power, OECD/ICH minimums, not “we always use 15/sex”).
• How pain and distress are limited (Refine: analgesia, humane endpoints, restraint time).
• Whether duplication of an existing study is unjustified.

Starting a GLP in-life without an approved protocol is both a welfare violation and a quality-system failure. “The sponsor insisted” is not an IACUC justification.

3Rs with tactics a study director can actually use

The 3Rs are Replacement, Reduction, and Refinement (Russell and Burch). On DABT they are operational, not slogans.

Replace. Use a validated alternative when it answers the question: OECD 471 Ames for bacterial mutagenicity; reconstructed human epidermis for some irritation/corrosion questions; OECD 437 bovine corneal opacity and permeability (BCOP) for certain eye-irritation questions; OECD 442C/D/E defined approaches for skin sensitization; in silico quantitative structure-activity relationship (QSAR) and read-across for data-gap filling. Human hepatocytes for metabolite identification can Replace a non-informative animal ADME species for that metabolite question.

Reduce. Microsampling (capillary volumes on the order of tens of microliters) can collect sparse toxicokinetic (TK) data from main-study animals and shrink or eliminate satellite TK groups. Do not add recovery groups “because the template has them” on a 14-day dose-range-finding study whose only job is to pick doses. Do not inflate N beyond guideline minima without a power calculation. Sharing a vehicle control across two unrelated concurrent studies is rarely valid; that is not a 3Rs win if it wrecks interpretation.

Refine. Better gavage technique and volume limits, social housing when compatible with the design, shorter nose-only restraint, environmental enrichment, and humane endpoints instead of death as the planned endpoint (the reason OECD 420 fixed-dose acute oral replaced older lethality-centered guidelines).

Recovery groups versus satellite TK

These two extra cohorts are not interchangeable:

• Satellite TK animals exist because serial bleeding of rodents can stress animals, lower red-cell mass, and confound clinical pathology or organ weights on the main study. They answer “what was the systemic exposure?”
• Recovery animals are held without dosing after the last dose to ask “did the lesion reverse?” They answer reversibility, not TK.

Reducing recovery is appropriate when reversibility is not a decision-critical question. Reducing satellites is appropriate when microsampling or sparse main-study bleeds will do. Cutting the scientifically indicated second species without a waiver is not Reduction; it is an incomplete package.

Limits of alternatives

Validated alternatives still fail closed-system questions:

• A liver spheroid does not provide 90-day multi-organ histopathology, immune surveillance, or a developing fetus.
• Ames detects point mutations in bacteria, not all clastogens, aneugens, or nongenotoxic carcinogens.
• In vitro skin corrosion does not predict systemic dermal toxicity after repeated application.
• Organ-on-chip platforms are powerful mechanistic tools; they are not OECD MAD replacements for a 408-style study in 2026.
• Metabolic incompetence of a cell line will miss bioactivation. Add S9 or a metabolically competent model, or you have a false negative, not a 3Rs success.

Realistic exam scenario

A biotechnology sponsor proposes to skip all in-life work for a small-molecule IND and submit only human organ-chip cytotoxicity plus Ames. The 3Rs instinct is right; the regulatory design is not. ICH M3(R2) still expects repeat-dose in vivo data in relevant species unless a documented, agency-accepted waiver exists. The ethical protocol is to Reduce and Refine the in vivo work: microsampling instead of large satellite cohorts, recovery only if reversibility will change the clinical plan, the nonrodent chosen because it shares target pharmacology and metabolites, and in vitro/in silico used to Replace lower-value local-tolerance or sensitization add-ons where validated methods exist. IACUC should approve that reasoned package and should reject both “animals because we always use dogs” and “no animals because chips exist.”

Traps

• Using mouse as the default 90-day oral rodent to save money.
• Selecting NHP without showing that dog and minipig are pharmacologically irrelevant.
• Calling extra satellite TK animals a recovery cohort.
• Claiming an unvalidated organ-chip study satisfies OECD MAD.