Allometric Scaling Failures

A focused companion to our animal-vs-human evidence page. This page explains the dose-translation pitfalls in detail — why a rodent dose in mg/kg almost never converts cleanly to a human dose, and why peptides are particularly bad at translating.

Educational research-literacy content only. Not medical advice, not dosing guidance, not sourcing advice, and not a protocol for human or animal use. See our responsible information policy.

A standard pattern in grey-market peptide content: a rodent study reports a dose of, for example, 10 µg/kg subcutaneous twice daily. An online source multiplies this by a 70 kg body weight to derive a 700 µg twice-daily “human dose.” This is wrong for several independent reasons. This page enumerates them.

Body weight is the wrong scaling variable

Allometric scaling — relating physiological variables across species — uses body-surface area or metabolic rate, not body weight. Smaller animals have higher per-kg metabolic rate, more rapid clearance, and proportionally smaller plasma volumes. The FDA's published Guidance for Industry on initial human dose selection uses a Human Equivalent Dose (HED) calculated as the animal dose divided by a species-specific factor — for example, rat dose × 0.16 to estimate the HED in mg/kg.

A 10 mg/kg dose in a rat does not equate to 700 mg in a 70 kg human. By HED calculation it equates to approximately 1.6 mg/kg — i.e. ~112 mg, an order of magnitude less than the naive bodyweight extrapolation. This is the conservative no-other-information starting estimate, not a clinical dose.

See our species dose-scaling explainer for an interactive walk-through of the FDA HED calculation across different animal species.

Receptor distribution differs between species

Allometric scaling assumes the receptor target is comparably distributed across species. For many peptide targets, this is false:

• GLP-1 receptor expression differs in density and tissue distribution between rodents and humans, contributing to the different magnitude of weight effects observed across species.
• Melanocortin receptors have substantially different tissue distributions in mice vs humans. Mouse models of obesity, pigmentation, and sexual behaviour driven by MC4R activation do not all translate cleanly.
• Ghrelin receptor (GHSR-1a) central expression differs across species; the rodent appetite response to GHRPs does not predict the human magnitude.
• Thymic-peptide targets — thymic involution patterns differ markedly between species, complicating the age-related modelling.

Pharmacokinetics differ — sometimes by orders of magnitude

Plasma half-life, protein binding, and clearance pathways all diverge across species:

• Native peptide half-life is typically minutes in all mammals — but the per-mg clearance rate scales with metabolic rate, not body weight.
• Fatty-acid-conjugated peptides (semaglutide, tirzepatide) depend on albumin binding for their long half-life. Albumin concentration is broadly comparable across mammals, but binding affinity for any given conjugate can differ.
• DPP-IV cleavage rate differs between species and across the same peptide's sequence variants.
• Renal clearance for peptides eliminated through glomerular filtration scales with kidney size, which is not linear with body weight.

Immunogenicity profiles diverge

A peptide that is non-immunogenic in rodents can be substantially more immunogenic in humans, and vice versa. Mice often do not produce neutralising antibodies to human-sequence peptides on chronic dosing; humans do. This is a particularly important constraint for the chronic-dosing translation of mitochondrial and bioregulator peptides. See immunogenicity explained.

Why peptides translate especially poorly

Several class-specific factors compound the general rodent-to- human translation problem for peptides:

• Route of administration matters more. A peptide delivered subcutaneously in a rat vs intraperitoneally vs orally produces very different exposure profiles; the published literature is inconsistent about which route was studied.
• Tissue penetration is harder to extrapolate.Peptide tissue distribution depends on molecular size, charge, and active transport — all of which vary across species.
• Single-laboratory effects. Many peptide findings are concentrated in one or two laboratories with specific handling and formulation protocols. Translation assumes the formulation reproduces — it often doesn't.
• Disease-model mismatch. A chemically-induced rodent injury model does not match human chronic pathology. Even if dose translation worked, the indication wouldn't.

The first-in-human approach

Properly conducted drug development handles this by:

1. Calculating the Human Equivalent Dose (HED) from the most sensitive species' No Observed Adverse Effect Level (NOAEL).
2. Applying a 10× safety factor to derive a Maximum Recommended Starting Dose (MRSD).
3. Beginning Phase I dosing well below that, with small single-ascending-dose cohorts and intensive monitoring.
4. Iteratively titrating across multiple cohorts to determine a tolerated and pharmacologically active human dose.

This process can take years and millions of dollars per compound. It is what produces a clinically-meaningful human dose. Online “mg/kg × bodyweight” calculations are not a substitute for it.

What this means for reading peptide content

When a source describes “research-protocol dosing” for an unapproved peptide, ask: was this dose derived from a registered human clinical trial, or extrapolated from animal data by bodyweight? Most online doses for unapproved compounds are the latter. They are not safe starting estimates; they are not clinical doses; they have not been through the process above. On PeptideStacks we report doses in study context only, with the species and route of the underlying study made explicit.

Related

• Animal vs human peptide research
• In vitro evidence limitations
• How to read peptide studies
• Human data vs preclinical data
• Immunogenicity explained
• Administration routes compared