Written and maintained by CASRAI Editorial Board
Last updated
The No Observed Adverse Effect Level (NOAEL) is the pivot point between animal toxicology and the first dose a human volunteer ever receives. Every first-in-human (FIH) starting-dose calculation for a systemically administered therapeutic begins with a NOAEL identified from nonclinical safety studies, converts it to a Human Equivalent Dose (HED) through allometric scaling, and then divides that HED by a safety factor to arrive at a Maximum Recommended Starting Dose (MRSD). This is not one sponsor’s convention — it is the algorithm FDA’s Center for Drug Evaluation and Research (CDER) laid out in its 2005 guidance, Estimating the Maximum Safe Starting Dose in Initial Clinical Trials for Therapeutics in Adult Healthy Volunteers, and it remains the standard framework reviewers expect to see reproduced, step by step, in an IND’s nonclinical overview.
Step 1: Identifying the NOAEL From Dose-Ranging Animal Studies
The FDA guidance defines the NOAEL for dose-setting purposes as the highest dose tested in an animal species that does not produce a significant increase in adverse effects compared to the control group — with the explicit instruction that biologically significant effects count even if they fail to reach statistical significance. This benchmark is identified from the dose-ranging and repeat-dose toxicology studies that make up the nonclinical package, typically conducted under Good Laboratory Practice and dosed by whichever route of administration matches the intended clinical route, using several dose groups spanning from clearly non-toxic to frankly toxic.
Three categories of finding can establish where the NOAEL falls: overt toxicity (clinical signs, gross or microscopic lesions), surrogate markers (elevated liver enzymes and similar clinical-pathology signals), and exaggerated pharmacodynamic effects. The guidance is careful to separate the NOAEL from three terms it is routinely confused with:
- NOEL (No Observed Effect Level) — the highest dose with no detected effect of any kind. A NOAEL can sit above the NOEL, because some observed effects may be acceptable pharmacodynamic actions rather than adverse findings.
- LOAEL (Lowest Observed Adverse Effect Level) — the lowest dose that does produce an adverse effect; a bracket, not a substitute, for the NOAEL.
- MTD (Maximum Tolerated Dose) — the highest dose that does not produce unacceptable toxicity, a different threshold typically used in oncology dose-finding rather than healthy-volunteer starting-dose calculations. FDA’s guidance is explicit that MTD is not generally used as the benchmark for a healthy-volunteer starting dose.
One nonclinical wrinkle matters directly for the calculation that follows: if drug absorption saturates at a dose below the highest nontoxic dose tested, the guidance directs sponsors to use the lowest saturating dose, not the highest nontoxic dose, when calculating the HED — because a dose above the saturation point no longer produces a proportionally higher, and therefore no longer reliably comparable, exposure.
Step 2: Converting the Animal NOAEL to a Human Equivalent Dose (HED)
A NOAEL expressed as mg/kg in a mouse and a NOAEL expressed as mg/kg in a dog are not comparable numbers — smaller animals have proportionally larger body-surface-area-to-mass ratios, which drives faster metabolic clearance per kilogram. The FDA guidance resolves this with allometric scaling by body surface area rather than a straight mg/kg carryover. Body-surface-area normalization is grounded in decades-old oncology dose-scaling work (Freireich et al. 1966; Schein et al. 1970) showing that lethal and maximum-tolerated doses correlate across species when expressed as mg/m² rather than mg/kg.
The guidance’s own appendix (Appendix A) acknowledges that later analyses (Travis and White 1988; Watanabe et al. 1992) found an allometric exponent of 0.75 fits some datasets better than the 0.67 implicit in strict body-surface-area scaling — but concludes that 0.67 should still be used as the default for starting-dose calculations specifically because it produces the more conservative (lower) HED, and therefore a safer starting point, absent species- or class-specific evidence to the contrary.
The formula and the Km shortcut
The general conversion is:
HED = Animal NOAEL (mg/kg) × (Animal weight in kg ÷ Human weight in kg)0.33
For the standard laboratory species and body weights, FDA’s guidance pre-computes this into a single conversion factor, Km — the ratio needed to convert an animal’s mg/kg dose to the equivalent mg/kg dose in a 60 kg human. Table 1 of the guidance gives the standard values:
| Species | Km | To convert animal mg/kg to HED, multiply by |
|---|---|---|
| Human | 37 | — |
| Mouse | 3 | 0.08 |
| Hamster | 5 | 0.13 |
| Rat | 6 | 0.16 |
| Ferret | 7 | 0.19 |
| Guinea pig | 8 | 0.22 |
| Rabbit | 12 | 0.32 |
| Dog | 20 | 0.54 |
| Monkey (cynomolgus, rhesus, stumptail) | 12 | 0.32 |
| Micro-pig | 27 | 0.73 |
| Mini-pig | 35 | 0.95 |
So a rat NOAEL of 10 mg/kg converts to an HED of 10 × 0.16 = 1.6 mg/kg — not 10 mg/kg. The guidance also documents when mg/kg-to-mg/kg scaling (skipping the body-surface-area step entirely) is defensible instead: when the available NOAELs already fall at a similar mg/kg dose across two or more tested species, direct mg/kg extrapolation may be more appropriate than the default mg/m² approach — but the sponsor has to justify the deviation, and the guidance flags that mg/kg scaling gives a 12-, 6-, and 2-fold higher HED than the mg/m² default for mice, rats, and dogs respectively, so the default remains the more conservative choice absent that justification.
Step 3: Selecting the Most Appropriate Species
Once every tested species’ NOAEL has been converted to an HED, the species producing the lowest HED is, by default, the “most sensitive species” — and using its HED gives the most conservative starting dose. FDA’s guidance allows departing from the most-sensitive-species default when other evidence identifies a more biologically relevant species instead, which is common for biologics with narrow target specificity: a monoclonal antibody that only binds the human version of its target may show little or no activity in standard toxicology species, in which case ICH S6 calls for in vitro binding and pharmacology data to guide species selection before, not after, the toxicity studies are designed. The guidance’s own worked example is antisense oligonucleotides, where monkeys — not rodents — reproduce the human dose-limiting toxicity (complement activation), making monkey data the relevant basis for the HED even when it is not the numerically lowest.
Step 4: Applying the Safety Factor to Reach the MRSD
The HED is not the starting dose. FDA’s guidance calls for dividing the HED by a safety factor, with a default value of 10 — described as “a historically accepted value” rather than a fixed statutory minimum. The safety factor exists to cover uncertainties the animal-to-human extrapolation cannot resolve on its own: possible heightened human sensitivity to pharmacologic effects, toxicities that are hard to detect in animals (headache, myalgia, mood disturbance), receptor-density or -affinity differences, and interspecies differences in absorption, distribution, metabolism, and excretion.
FDA frames the factor as a sliding scale, not a fixed constant, and lists specific findings that justify moving off 10 in either direction:
Reasons to raise the safety factor above 10
- A steep dose-response curve for significant toxicities
- Severe or organ-specific toxicity (e.g., CNS damage)
- Toxicity that isn’t monitorable by routine clinical pathology
- Toxicity with no premonitory warning signs
- Widely divergent or poor bioavailability across species
- Irreversible toxicity, or unexplained mortality
- Nonlinear pharmacokinetics, or large dose/exposure variability at the toxic threshold
- A novel therapeutic target with no prior clinical experience
- Limited cross-reactivity of the animal models used
Reasons a factor below 10 can be justified
The guidance permits going lower only for well-characterized drug classes with consistent, monitorable, reversible, predictable toxicity across species — including humans, where relevant class experience exists — and it places the burden of explaining the deviation squarely on the sponsor: “It is incumbent on the evaluator to clearly explain the reasoning behind the applied safety factor when it differs from the default value of 10, particularly if it is less than 10.” A longer toxicology-study duration relative to the planned clinical dosing schedule is one specific, guidance-cited basis for a smaller factor, since it implies extra safety margin is already built into the NOAEL.
Step 5: Checking the MRSD Against the Pharmacologically Active Dose (PAD)
Selecting a PAD is outside the scope of the FDA guidance itself, but once an MRSD has been calculated, the guidance recommends comparing it against the PAD — the lowest dose in animal pharmacodynamic models showing the intended pharmacologic activity, converted to its own HED the same way the NOAEL is. If that pharmacologic HED comes out lower than the toxicology-derived MRSD, FDA’s guidance flags this as a reason to consider lowering the clinical starting dose further — particularly for drug classes where exaggerated pharmacology, not off-target toxicity, is the dominant safety concern (vasodilators, anticoagulants, monoclonal antibodies, and growth factors are the guidance’s own examples).
Where This Fits in the IND Pathway
The MRSD calculation is one deliverable inside a much larger nonclinical-to-clinical transition. It draws on the toxicology package’s exposure and margin data, gets discussed (often explicitly) at the pre-IND meeting, and is written up in the nonclinical overview that accompanies the IND submission itself — a submission sponsors should confirm they actually need via the IND application decision tree. Once dosing begins, the MRSD sets only the ceiling on the first cohort; subsequent escalation follows a separate, pre-specified design such as the 3+3 design or a continual reassessment method, and the whole sequence sits inside the protocol structure described in designing a clinical trial protocol and the broader clinical trial phase framework.
FAQ
Is the MRSD the dose a trial actually starts at?
No — it is a ceiling, not a target. FDA’s guidance is explicit that the algorithm generates the maximum recommended starting dose; sponsors are free to start lower, and often do when a wide margin of caution is warranted for other reasons.
Why use body surface area instead of body weight to scale doses?
Because metabolic rate and clearance scale more closely with body surface area (proportional to weight raised to roughly the 0.67 power) than with weight itself, smaller animals clear a given mg/kg dose faster than larger ones. Direct mg/kg carryover would systematically overestimate the safe human dose from small-rodent data; the guidance keeps mg/m² normalization as the default specifically because it is the more conservative choice.
Can the default safety factor of 10 ever be reduced?
Yes, but only with an explicit, documented justification — a well-characterized drug class, consistent and monitorable toxicity across species including any available human data, and often a NOAEL derived from a toxicology study of longer duration than the planned clinical dosing. The guidance places the explanatory burden on the sponsor whenever the factor drops below 10.
Does the most sensitive species always determine the HED used?
By default, yes — the species giving the lowest HED is used, because that produces the most conservative starting dose. Sponsors can use a different, more biologically relevant species instead when supporting evidence exists, which is common for biologics with limited cross-species target reactivity.
How does NOAEL relate to toxicokinetic exposure margins?
They are complementary, not competing, calculations. The NOAEL-to-MRSD algorithm on this page works from administered dose; toxicokinetic exposure margins compare measured systemic exposure (AUC or Cmax) at the NOAEL to anticipated human exposure. See toxicokinetics in nonclinical safety studies for how the two calculations sit alongside each other in a nonclinical package.
