Why a Human-Focused Rationale Is Needed
For some biologics, no tested animal species has sufficiently similar target binding and functional pharmacology to predict the human on-target response. In that situation, an apparently reassuring animal toxicity study does not establish that humans will tolerate a pharmacologically active exposure: the animals may simply not experience the relevant biology. The starting-dose rationale therefore needs a human pharmacology anchor, supported by human in vitro evidence, while treating available toxicology as informative only for the hazards it can address.
The earlier material on defining and deriving MABEL provides methods for turning mechanistic evidence into a minimal anticipated biological effect level. Here the challenge is integration: deciding which evidence is relevant when species translation is weak, how to use toxicity findings without giving them undue weight, and how to explain the remaining uncertainty. A starting dose is a risk-management decision, not a guarantee of safety or a prediction that no pharmacology will occur.
Establish What the Evidence Can Support
Before calculating a dose, establish why no species is considered pharmacologically relevant. Evidence can include comparative target binding, functional assays, target sequence or structure, and the response to the biologic in species-derived cells. A lack of response in an animal assay is not enough if the assay did not express the relevant target or was not capable of detecting the expected response. Conversely, a measurable response in a species does not automatically establish relevance if its potency, pathway, or target distribution differs materially from humans.
The human evidence should be organized around the intended biological action and plausible early hazards. Depending on the biologic, this may include binding and functional potency, target occupancy, downstream biomarkers, and responses in primary human cells or multiple donors. Use a concentration or exposure threshold that is both supported by the assay and relevant to the intended clinical route and exposure metric. The MABEL should reflect the lowest credible effect threshold, not an average chosen to make the resulting dose more convenient.
Assay interpretation requires attention to what was measured. A threshold from a sensitive functional assay may be more informative than a binding value alone; a cytokine-release assay may capture a hazard that a target-specific potency assay misses. Check concentration units, free versus total concentration, assay duration, target expression, donor variability, and whether the observed signal is reproducible. If an assay is noisy or its threshold is uncertain, that uncertainty should lead to a more cautious interpretation or additional evidence—not to false precision.
From an Exposure Threshold to a Starting Dose
A concentration threshold is not yet a dose. It must be mapped to the human exposure expected after administration, using the route, bioavailability, distribution, and pharmacokinetic assumptions appropriate to the molecule and dose range. The earlier material on translating pharmacology into exposure targets and accounting for route and bioavailability supplies the relevant framework. The numerical conversion below uses a deliberately simple linear approximation; it is valid only when exposure is approximately proportional to dose over the range considered.
If the selected MABEL concentration is \(C_M\), and predicted concentration at the relevant time is \(C(D)=kD\), where \(D\) is dose and \(k\) is the dose-to-concentration proportionality, then the dose corresponding to that concentration is \(D_M=C_M/k\). Reducing the dose by a factor \(F\) reduces the predicted concentration by the same factor under this model:
This is the justification for dividing a MABEL-equivalent dose by a safety factor in a linear model. It is not a universal rule for every biologic. If target-mediated disposition, nonlinear binding, delayed effects, or a different exposure metric materially changes the dose-exposure relationship, calculate the dose by inverting the appropriate exposure model rather than assuming proportionality. The safety factor should reflect the uncertainty and consequence of error; it is not a substitute for a defensible exposure model.
Worked Example: Converting a Human In Vitro Threshold to a Dose
Hypothetical data. A human primary-cell assay gives a reproducible minimal-effect threshold of 0.30 ng/mL in free concentration. For an illustrative intravenous first exposure, assume the relevant early exposure is approximated by \(C_{\max}=D/V_c\), with a central volume \(V_c=4.0\) L. Assume the free fraction in the assay is representative of the relevant human plasma concentration, and use a 30-fold reduction from the MABEL-equivalent dose to address uncertainty in translation and the seriousness of the potential response.
First convert the threshold to compatible units: \(0.30\) ng/mL equals \(300\) ng/L. The dose that would produce this concentration in the simplified model is:
Applying the stated factor gives \(D_{\text{start}}=1.2\ \text{µg}/30=0.040\ \text{µg}\), or 40 ng total. For a 70-kg adult, this is \(0.040\ \text{µg}/70\ \text{kg}=0.000571\ \text{µg/kg}\), approximately \(0.571\) ng/kg. Under the assumptions, the predicted concentration is \(0.30/30=0.010\) ng/mL.
This is a worked rationale, not a general dose recommendation. The calculation depends on the assumed volume, free-concentration correspondence, intravenous route, linearity, and selection of the 30-fold factor. If any of those assumptions is unsupported, the numerical result must be revised or described as uncertain.
Use Toxicology Without Treating It as Human Pharmacology
A non-relevant species can still provide useful information. A study may reveal formulation-related toxicity, nonspecific organ effects, immune responses unrelated to the human target, or toxicity at exposures that help contextualize systemic tolerability. It can also show whether the molecule was tolerated at substantial exposure in that species. Those observations are evidence about the tested species, study conditions, and hazards plausibly shared across species; they do not establish a safe human exposure for target-mediated effects.
Interpret the study by asking what caused the findings, whether the relevant target was present and engaged, which endpoints were measured, and what exposure was achieved. A NOAEL is an observed dose or exposure at which adverse effects were not observed under the study conditions. It is not a universal boundary below which all humans are safe. When the species lacks the relevant pharmacology, a NOAEL may not constrain human on-target risk even if it is useful for other risks. A large exposure margin should therefore not be presented as proof that the human starting dose is safe.
Worked Example: Interpreting a Non-Relevant-Species NOAEL
Hypothetical data. In a repeat-dose study, a biologic has a NOAEL-associated AUC of 200 µg·h/mL in a species shown not to bind the human target. The proposed human starting dose is 0.040 µg total. A simple human model predicts clearance of 0.10 L/h and assumes linear pharmacokinetics at this dose.
The predicted human AUC is dose divided by clearance:
The numerical ratio of the animal NOAEL-associated AUC to this predicted human AUC is \(200/0.00040=500{,}000\). This striking ratio is a descriptive comparison of the stated exposures; it does not establish a 500,000-fold safety margin against human on-target effects. The animal did not bind the target, and the exposure comparison does not repair that mechanistic gap. The study may support a conclusion that certain non-target toxicities were not observed at the tested animal exposure, subject to the study's limitations. The human MABEL rationale remains the relevant anchor for on-target risk.
Integrate Evidence and Make Uncertainty Visible
When evidence points in different directions, do not combine it by mechanically averaging dose estimates. First classify each result by the risk it can address. Then identify the most sensitive credible human pharmacology threshold, convert it with an appropriate human exposure model, and consider whether toxicity findings impose an additional constraint for hazards they can plausibly inform. The controlling rationale should be explicit: a toxicity result that cannot address on-target human effects should not override a more cautious, human-relevant MABEL.
Document binding and functional evidence. State which animal findings are interpretable for human pharmacology and which are limited to other hazards.
Review assay quality, biological relevance, donors, endpoints, and concentration units. Select the lowest credible minimal-effect threshold and describe its uncertainty.
Use the intended route and a human exposure model that is appropriate at the proposed dose. State the exposure metric and assumptions, including whether proportionality is reasonable.
Justify the reduction for uncertainty and consequence of error. Use non-relevant-species toxicity findings only for hazards and exposure comparisons they can support.
Identify unresolved human risks and explain how the initial clinical design, monitoring, and escalation decisions will manage them.
Worked Example: Choosing Among Discordant Human Assays
Hypothetical data. Three human assays for the same biologic yield minimal-effect thresholds of 0.30 ng/mL, 0.12 ng/mL, and 0.50 ng/mL. The 0.12 ng/mL result is reproduced in a relevant primary-cell assay across donors, while the other values come from less sensitive systems with lower target expression. A non-relevant animal study has a NOAEL-associated AUC of 200 µg·h/mL. Use the same illustrative intravenous model as above: \(V_c=4.0\) L, linear exposure at low dose, and a 30-fold reduction.
The 0.12 ng/mL threshold is selected because it is the lowest credible result from the more relevant, reproducible assay. Convert it: \(0.12\) ng/mL equals \(120\) ng/L. The MABEL-equivalent dose is:
The illustrative starting dose is \(0.48\ \text{µg}/30=0.016\ \text{µg}\), or 16 ng total. The higher thresholds do not justify using a larger dose when the lower result is credible and more biologically relevant. The non-relevant-species NOAEL does not negate the human assay evidence; it provides separate context for toxicities that the animal study could detect. The rationale should also state that the chosen threshold, volume, model, and factor remain assumptions requiring review as evidence accumulates.
Diagnostics, Alternatives, and Limitations
Several checks can expose a weak rationale. Confirm that assay concentrations and clinical exposure predictions use compatible units and concentration definitions. Check that the assay endpoint corresponds to the intended biological action or a plausible safety signal. Review donor-level responses rather than relying only on a pooled mean, and investigate whether a low threshold is reproducible or an isolated artifact. Verify that the pharmacokinetic model is suitable for the biologic and dose range; target-mediated disposition can make a simple proportional dose conversion unreliable. Finally, make sure the stated toxicology conclusion does not extend beyond the species, endpoints, and mechanisms actually studied.
MABEL is usually more directly relevant than a species-derived NOAEL when no animal species reproduces human pharmacology. A relevant-species NOAEL can be informative when the species shares the important pharmacology, but that condition is absent here. A surrogate molecule, humanized model, or other specialized system may add evidence, yet its relevance depends on how well it represents the clinical molecule and human biology. These alternatives can complement human in vitro evidence; none automatically removes the need to explain remaining uncertainty.
The main limitation is that human in vitro systems do not reproduce the full human organism. They may not capture tissue distribution, immune context, delayed effects, or variation in clinical exposure. A safety factor is a transparent allowance for uncertainty, not a numerical guarantee that all unknown risks have been covered. A defensible rationale therefore combines mechanistic evidence, justified exposure translation, fit-for-purpose toxicology, and a clear account of residual risk. Clinical safeguards such as sentinel administration, close monitoring, and cautious escalation can manage risk but do not make an inadequate starting-dose rationale adequate.
Check Your Understanding
Use the stated assumptions in each question. Solutions follow the questions.
- Why does a high NOAEL in a species that does not bind the human target fail to establish a safe human on-target exposure?
- A human assay threshold is 0.20 ng/mL, and an illustrative IV model uses \(V_c=5.0\) L with a 20-fold reduction. What starting dose follows from the linear model?
- Two assays give thresholds of 0.08 ng/mL and 0.25 ng/mL. The lower threshold is reproducible in relevant human primary cells; the higher threshold is from a less relevant system. Which should anchor the MABEL rationale, and why?
- Name two findings from a non-relevant-species toxicology study that may still inform risk, and one conclusion it cannot establish.
Solutions. (1) The species cannot express the relevant target-mediated pharmacology, so absence of that toxicity does not test the human on-target mechanism. (2) \(0.20\) ng/mL is \(200\) ng/L, so \(D_M=(200\ \text{ng/L})(5.0\ \text{L})=1000\) ng \(=1.0\) µg. Dividing by 20 gives \(0.050\) µg, or 50 ng total, under the stated linear assumptions. (3) The 0.08 ng/mL threshold should anchor the rationale because it is both lower and supported by the more relevant, reproducible assay; the less relevant result does not justify disregarding it. (4) The study may inform formulation-related or non-target organ toxicity and describe exposures tolerated in that species. It cannot establish that a human on-target effect will not occur at the proposed exposure.