25.7 — Before Humans: Animal Studies, Toxicology and the IND

On 13 March 2006, at a clinical trials unit in north-west London, six healthy young men were given the first human dose of an experimental antibody called TGN1412. The dose was five hundred times lower than a dose found safe in monkeys. Within about an hour all six were in agony. Within twelve hours all six were in intensive care with multiple organ failure. All survived; one lost fingers and toes.

Nothing had been skipped. The animal studies had been done, in the species considered most relevant, with a wide safety margin applied.

The drug was designed to switch on a receptor on T cells. In the monkeys used, that receptor differs subtly from the human version, and the animals' cells did not respond the way human cells did. In the men, every T cell in the body activated at once and released a flood of inflammatory signals — the cytokine storm of Chapter 13.4.

Everything in this chapter exists because of the gap that disaster demonstrated: animal data is the only bridge we have to the first human dose, and it is a bridge with known holes. The industry's answer was not to abandon animal work but to change how the first dose is calculated for drugs that act on the immune system, and to require the first volunteers to be dosed one at a time with a gap between them rather than together.

What the preclinical package has to answer

"Preclinical", also called nonclinical, means all the laboratory and animal work done to justify giving the compound to a person. It has to answer four questions, and it is easiest to keep them separate.

Does it do what we think it does, in a living animal? This is primary pharmacology — showing the drug hits the target and changes the disease in an animal model.

What else does it do? Secondary pharmacology screens for activity at unrelated receptors, and safety pharmacology specifically examines the three organ systems whose sudden failure kills: the cardiovascular system, the respiratory system and the central nervous system.

What does the body do to it? Animal pharmacokinetics — absorption, distribution, metabolism and excretion (Chapter 22.2) — measured properly, so that the blood levels reached in animals can be compared with the levels expected in people. This comparison is the real currency of safety margins; the dose in milligrams matters far less than the drug exposure achieved.

And at what dose does it cause harm? That is toxicology, and it is the largest and most expensive part.

The toxicology programme, study by study

These studies are standardised worldwide through the ICH guidelines — the International Council for Harmonisation, covered in Chapter 25.14 — so a package accepted in one region is largely accepted in others. The guideline that lays out what is needed before each clinical stage is ICH M3(R2), and it is worth knowing by name because clients plan around it.

Single-dose and repeat-dose toxicity. The compound is given daily to two species — usually a rodent, typically rats, and a non-rodent, typically dogs or monkeys — for a duration that matches or exceeds the intended human treatment. A two-week human study needs shorter animal studies than a year-long treatment for a chronic disease. Everything is measured: body weight, food intake, blood chemistry, blood counts, organ weights, and a full microscopic examination of tissues at the end.

Genotoxicity. Does the compound damage DNA? A bacterial mutation test, a chromosome damage test in cells, and usually a test in animals. A positive result here can stop a programme outright, because DNA damage points towards cancer.

Carcinogenicity. Two-year studies in rodents, required for drugs intended for long-term use. These are slow and costly, which is why they run alongside clinical development rather than before it.

Reproductive and developmental toxicity. Effects on fertility, on the developing embryo, and on offspring after birth. This is the direct legacy of thalidomide (Chapter 25.2), and it is why the labelling of any drug in pregnancy is treated so seriously.

Local tolerance and immunogenicity where relevant — does an injection damage the tissue where it is given, and does the immune system attack the drug.

GLP: the rulebook that makes the data believable

Safety studies are run under Good Laboratory Practice, in the United States codified at 21 CFR Part 58. It is worth understanding what GLP is and is not, because people constantly assume it is about doing good science.

GLP is about traceability and honesty, not about scientific quality. It says: there must be a written protocol approved before the study starts; a named study director responsible for the whole study; standard operating procedures for every routine activity; raw data recorded at the time and never obscured; every instrument calibrated and every calibration recorded; test material characterised so you know exactly what was given; and an independent Quality Assurance unit that inspects the study and reports separately from the people running it.

The reason it exists is a scandal. In the 1970s American inspectors found that a large contract laboratory doing safety testing for many companies had fabricated and manipulated study data on a substantial scale; several executives were later convicted. Regulators discovered they had been approving products on the basis of studies that had not been done as described. GLP followed in 1978.

For a software engineer this is the first place the phrase "the record is the product" becomes literal. A GLP study's value is entirely in whether its records can be shown to be complete, contemporaneous and unaltered. Every requirement you will meet in Chapter 25.20 on data integrity has its ancestor here.

Calculating the first dose in a human

This calculation is one of the most consequential arithmetic exercises in the industry, and it is completely learnable.

Step one: find the NOAEL in each species. NOAEL means the no-observed-adverse-effect level — the highest dose at which no harmful effect was seen. Note the wording: not the highest dose with no effect at all, but with no adverse effect.

Step two: convert to a human equivalent dose. You cannot simply scale by weight, because smaller animals run their metabolism faster per kilogram. The standard method scales by body surface area, using conversion factors published by regulators — dividing the rat dose in milligrams per kilogram by about 6.2, and the dog dose by about 1.8, to get the human equivalent in milligrams per kilogram.

Step three: choose the most sensitive species — the one giving the lowest human equivalent dose — unless there is a strong scientific reason to prefer another.

Step four: divide by a safety factor, normally ten, and increase that factor when there is extra uncertainty: a steep dose–response curve, a target with unclear biology, toxicity that would not be detectable early, or irreversible harm.

The result is the maximum recommended starting dose for a first-in-human study. For drugs where an exaggerated pharmacological effect is the danger rather than classical toxicity — exactly the TGN1412 case — the starting dose is instead set from the minimum anticipated biological effect level, meaning the lowest dose expected to do anything at all in a human, which is a far more conservative anchor.

The IND: the document that opens the door

In the United States, it is illegal to ship an unapproved drug across state lines. An Investigational New Drug application, the IND, is the exemption that allows it for research purposes. Filing one is the formal boundary between preclinical and clinical, and every client plan has an "IND date" on it.

What goes into it comes in three blocks.

Chemistry, manufacturing and controls, universally shortened to CMC. What the substance is, how it is made, how purity is measured, how stable it is, and what the clinical material's specification is. Enough that the regulator believes the material given to volunteers is what the sponsor says it is.

Pharmacology and toxicology. The whole package described above, with full study reports, plus the reasoning for the proposed starting dose.

The clinical plan. The protocol for the first study, the qualifications of the investigators, and the Investigator's Brochure — the document that tells every doctor running the trial everything known about the drug so far. This document is legally important: it defines what counts as an "expected" side effect, which in turn decides which safety reports must be rushed to regulators, as Chapter 25.22 explains.

Then a rule that surprises people: the FDA does not approve an IND. The sponsor may begin dosing 30 days after submission unless the agency objects. If the agency has concerns it issues a clinical hold, and no participant may be dosed until the issues are resolved. In Europe and most other regions the equivalent submission does require an explicit authorisation, which matters when planning a global study.

In parallel, and independently, an ethics committee must approve the study — the IRB of Chapter 25.2. Regulatory permission and ethical approval are two separate gates and neither substitutes for the other.

What is changing about animal testing

The FDA Modernization Act of 2022 removed the statutory requirement that a drug must be tested in animals before human trials, allowing alternatives such as cell-based assays, organ-on-a-chip systems and computational models where they are adequately justified.

Read that carefully, because it is widely misreported. It did not ban animal testing and it did not make animal data optional in practice. It removed a legal requirement and left the scientific judgement to regulators, who still expect a convincing safety package. The practical effect so far is gradual: more use of human-cell-based systems as supporting evidence, and continued animal studies where nothing else predicts systemic toxicity.

The honest position to hold, and the one that survives contact with an expert, is that alternatives are improving quickly in specific areas — liver toxicity, cardiac electrophysiology, barrier tissues — and have not yet replaced a whole-body system for detecting what nobody thought to look for.

Where the work and the systems are

A preclinical programme costs in the low tens of millions and takes one to three years, and most of it is outsourced to specialist contract laboratories.

The systems supporting it are laboratory information management systems handling study designs, animal records and specimen tracking; toxicology data systems producing the tabulated results; document systems holding protocols and reports under signature; and the assembly systems that build the submission itself.

One technical detail comes up on nearly every preclinical data project and is worth knowing in advance. Regulators require study data in standard formats defined by the CDISC consortium — SEND, meaning Standard for Exchange of Nonclinical Data, for animal studies, and SDTM for clinical ones. A client asking for "SEND conversion" is asking you to transform laboratory results into a rigid, validated, regulator-readable structure, and the validation rules for it are published and checkable. This is a well-defined, high-volume, unglamorous problem, which makes it excellent service work.

Next: Chapter 25.8, the clinical phases — what each one is actually asking, and why the failures cluster where they do.