The Translation Gap: Why So Many Peptides That Work in Mice Fail in Humans

One of the most persistent patterns in drug development is also one of the most humbling: a compound produces a clean, convincing result in mice, moves into human trials, and quietly disappears. This is not the exception. It is close to the rule. Understanding why this happens is one of the best inoculations against overreading any single animal study — including the many that circulate in peptide discussions.

This is a research log entry, not medical advice. The goal here is to explain the machinery behind the "works in mice, fails in humans" problem so you can read preclinical claims with appropriately calibrated skepticism.

The attrition numbers are stark

Across the industry, only roughly 8–10% of drug candidates that enter Phase I human trials are ultimately approved. The steepest drop is at Phase II, where compounds that looked safe in early testing repeatedly fail to show adequate efficacy — this is where the mouse-to-human story most often falls apart. In some fields the gap is even wider: by one accounting, of every 100 neuropsychiatric drugs that reach clinical trials — usually after "working" in mice — only around nine become approved medicines, among the lowest rates of any category.

Those numbers describe drugs that already cleared preclinical hurdles. The failure isn't random noise; it reflects structural reasons a mouse result doesn't carry over.

Reason 1: Dose doesn't scale one-to-one

You cannot take a milligram-per-kilogram dose that worked in a mouse and apply the same ratio to a human. Regulators use allometric scaling precisely because species differ in metabolic rate, body surface area, and clearance. The FDA's 2005 starting-dose guidance uses a "dose-by-factor" approach: take the no-observed-adverse-effect level (NOAEL) in animals and convert it to a human-equivalent dose (HED) using body-surface-area scaling (the classic exponent is ~0.67), not simple bodyweight. Ignoring this math is one of the fastest ways a "safe in mice" dose becomes dangerous — or ineffective — in people.

Reason 2: Peptides are metabolized very differently across species

This one is especially relevant to peptides. Their plasma half-life is dominated by proteolytic degradation, hepatic first-pass metabolism, and renal clearance — and those processes differ sharply between species. Rodents generally metabolize many peptides faster than primates, driven by higher metabolic rate and plasma peptidase activity. Small, unmodified peptides can be cleaved by serum proteases within minutes, while lipidated or albumin-binding analogs persist far longer. A peptide half-life measured in a mouse can be off by several fold in a human, which means both efficacy and safety windows can shift. Receptor sequences and expression patterns can differ too, so even the target may not behave identically.

For how a compound's regulatory and evidentiary status is (or isn't) established in humans, see our FDA status reference and the individual entries in the peptide library.

Reason 3: The disease model may not be the disease

An animal "model" of a human illness is a set of assumptions made concrete. In complex conditions — especially neurological and psychiatric ones, where many genes and circuits are involved — a mouse engineered to show one feature of a disease may not reproduce the disease itself. When the underlying biology is oversimplified, a drug that fixes the model can do nothing for the actual human condition. The model was never the thing you cared about; it was a proxy, and proxies leak.

Reason 4: Genetic uniformity vs. human variability

Laboratory rodents are typically inbred and near-genetically identical, housed in controlled conditions and eating identical diets. Humans are genetically diverse, variably aged, and carry different comorbidities and co-medications. A response that is clean and consistent in a homogeneous colony can fragment across a real population — some people respond, some don't, some react badly. That variance is invisible in the mouse data.

Reason 5: The preclinical literature is biased toward good news

Even setting biology aside, the evidence base is skewed. Systematic reviews of animal research repeatedly find small, underpowered studies, low rates of randomization, allocation concealment, and blinded outcome assessment — plus publication bias, where positive results are far more likely to be written up than negative ones. Pharmaceutical teams attempting to reproduce headline academic findings have often failed to. Guidelines like ARRIVE exist specifically to push back on this, but a lot of the older and informal literature predates them. So the mouse study you're reading may be not only species-limited but also optimistically reported.

What this means for reading peptide claims

None of this says animal research is worthless — it's a necessary early filter. What it says is that a positive rodent result is a hypothesis about humans, not a conclusion. When a peptide is promoted on the strength of mouse or rat data, the honest questions are: Was the dose scaled properly? Is the peptide's metabolism likely to translate? Does the model actually resemble the human condition? Was the study powered, randomized, and blinded — and were negative results reported? Until human trials exist, the gap between the two remains wide open.

We keep tracking these distinctions across the blog and the peptide library, always trying to separate what's been shown in animals from what's been shown in people.

PepStash is a research log and reference tool. This article is educational and is not medical advice — it does not diagnose, treat, or recommend any protocol. Regulatory status and trial data change; always verify against primary sources and consult a licensed physician before making any decisions about your health.