The Drug Safety Lifecycle · Section 3.2
~7 min read · The Drug Safety Coach — Global PV Career Course
Key points
Full text
Before any medicinal compound is administered to a human, it must pass through a rigorous preclinical safety evaluation programme designed to identify potential toxicities, define the dose range within which human studies can safely begin, and predict how the compound will behave in the human body.
Preclinical toxicology follows ICH M3(R2), which defines the minimum non-clinical studies required to support human trials: acute toxicity (single-dose, identifies maximum tolerated dose), repeat-dose/sub-chronic toxicity (28–90 days, dose-response for organ toxicity), genotoxicity (Ames test and others, mandatory before Phase I), reproductive toxicology (required before enrolling women of childbearing potential), carcinogenicity (typically required before large-scale or long-duration Phase III trials), and safety pharmacology (cardiovascular hERG channel, CNS, respiratory — required before Phase I).
For PV professionals, understanding what was tested preclinically — and what was not — is essential context for interpreting unexpected adverse events later. A hepatotoxic signal in Phase II reads very differently if hepatotoxicity was already flagged in animal studies versus if it was entirely unexpected preclinically — the first suggests a known, monitorable risk; the second suggests a genuinely novel finding that may need a faster, more cautious response.
This is a distinction worth internalising early: preclinical data isn’t a box that got checked before your job started — it’s the baseline your entire causality assessment later rests on. A PV professional who can’t answer "was this seen in animals?" is working with half the picture when a genuinely serious signal appears.
2026 Update
In 2025–2026, AI-driven predictive toxicology is increasingly replacing or supplementing traditional animal testing for early-stage compounds. QSAR models and deep learning algorithms trained on large molecular toxicity datasets can predict hepatotoxicity, cardiotoxicity, genotoxicity, and skin sensitisation from molecular structure alone. These predictions are now accepted by some regulators as supporting data for read-across arguments and to justify reduction in animal studies.
Quick check
Test yourself before moving on — no pressure, just click an answer.
1. Which ICH guideline defines the minimum non-clinical studies required before human trials can begin?
2. Why does it matter for causality assessment whether a signal was already seen preclinically?