Most drug candidates that enter clinical development never reach approval, and a large share of those failures trace back to a gap between what a compound did in vitro and what it does in a living system. That gap is not bad luck. It usually comes from a handful of recurring causes, and most of them can be probed early.
1. The drug never reaches the target at the right concentration
Potency in a biochemical or cell assay is a statement about concentration at the target, but in vivo you control only the dose. Between the two sit absorption, distribution, metabolism and excretion. A compound with a nanomolar IC50 can still fail if it is cleared in minutes, poorly absorbed, or concentrated in tissues that are not the site of action.
Plasma protein binding matters as well. It is generally the unbound drug that is free to engage the target, so a total concentration that looks comfortably above the IC50 can hide a free concentration that is well below it.
2. Exposure is there, but target engagement was never shown
Teams often infer that the drug is hitting its target because the exposure looks right. Inference is not evidence. Where possible, measure engagement directly in the relevant tissue: a proximal biomarker, receptor occupancy, or a downstream marker that moves with the target. Without it, a negative efficacy result cannot tell you whether the biology was wrong or the drug simply never did its job.
3. The disease model does not resemble the human disease
An animal model that responds to a compound is not the same as a model that predicts a human response. Models can differ from patients in the species' biology, the timing of dosing relative to disease, the severity and heterogeneity of the condition, and the endpoint being measured. A treatment given before disease is established is a very different test from one given in a patient who has had the illness for years.
4. Species differences
Target expression, receptor pharmacology, metabolic enzymes and transporters can all differ between rodents, larger animals and humans. A metabolite that is harmless in one species can be the problem in another, and a compound that is stable in rat liver can be cleared rapidly by human enzymes. Cross-species comparison of potency, metabolism and protein binding is cheap relative to a failed study.
5. Safety shows up before efficacy does
Sometimes the compound reaches the target and works, but the dose that works is too close to the dose that harms. A narrow margin between efficacious and toxic exposure is a failure mode in its own right. Off-target pharmacology, reactive metabolites and organ-specific toxicity are common reasons a promising profile collapses once repeat dosing begins.
6. The study was not designed to give a clear answer
Underpowered groups, a single dose level, a poorly chosen time point, unblinded scoring or a formulation that changes between batches can all produce a result that looks like failure, or worse, like success that later cannot be reproduced. A weak experiment wastes the animals and the time as surely as a weak compound does.
Questions worth asking before the in vivo study
- What unbound concentration do we expect at the target, and for how long, compared with the concentration needed for the effect?
- How will we know the drug engaged its target in the tissue that matters?
- Which features of the human disease does this model capture, and which does it miss?
- Do potency, metabolism and binding hold up across the species we will use and in human systems?
- What is the margin between the effective exposure and the first signs of toxicity?
- What result would make us stop, and have we written it down in advance?