Why the Rationale Must Be Traceable
A first-in-human starting dose is a decision made from evidence with limitations, not a guarantee that the dose is safe for every participant. A reviewer should be able to follow the path from the underlying findings, through the calculations and assumptions, to the selected dose and the risks that remain. If a key input changes, the record should also make it possible to see which parts of the rationale need reconsideration.
The earlier material on NOAEL-, HNSTD-, and MABEL-based approaches, human-equivalent-dose conversion, safety factors, exposure margins, and conflicting evidence provides tools for evaluating candidate doses. The focus here is how to document the decision clearly. The record should not blur a measured animal result, a model prediction, a calculation, and a judgment into a single statement such as “the dose is conservative.”
What a Decision Record Should Show
A useful record separates four kinds of statements. Observed findings report what was measured, in which species or assay, at what dose or exposure, and under what conditions. Derived values show transformations such as human-equivalent-dose conversion or exposure ratios, with units and inputs. Predictions identify estimates for humans and the model or evidence supporting them. Decision judgments explain why a candidate dose was selected over alternatives, including which evidence controlled the decision.
For each important input, name its source and meaning. A “NOAEL of 3 mg/kg” is incomplete unless the record identifies the species, study, route, duration, and relevant findings. Similarly, an exposure margin needs both the exposure at the proposed dose and the comparator exposure, with the endpoint stated. If the exposure values are predictions, label them as such rather than presenting them as observed human measurements.
The calculation chain should preserve units at every step. Record any conversion factor, safety factor, body-weight assumption, dose rounding, or formulation constraint. If multiple evidence streams yield different candidate doses, state which one controls and why; do not present only the lowest number without explaining the decision. A brief summary table can keep the evidence basis and its interpretation distinct.
| Record element | What to make explicit |
|---|---|
| Evidence | Source, species or assay, dose or exposure, findings, and limitations |
| Calculation | Equation, inputs, units, conversions, and unrounded result |
| Assumptions | Conditions needed for the calculation or interpretation to apply |
| Decision | Selected dose, controlling evidence, alternatives considered, and reason for selection |
| Residual risk | Unresolved concern, why it matters, and a proportionate management action |
Show the Calculation Path, Not Just the Answer
The calculation should be reproducible from the documented inputs. For example, if a NOAEL-based candidate is obtained by converting an animal dose to a human-equivalent dose (HED) and then applying a safety factor, write the relationship explicitly:
This expression does not establish that the starting dose is appropriate by itself. It shows how one candidate was calculated. The scientific record must still justify the relevance of the source finding, the conversion, and the factor, and explain how this candidate compares with other evidence.
A dose may be rounded to a practical amount, but the record should show whether rounding increased or decreased it and what that did to the margin. For an exposure endpoint, define the margin as the comparator exposure divided by the predicted exposure at the proposed dose:
The numerator and denominator must refer to the same exposure metric and compatible conditions. A Cmax margin does not substitute for an AUC margin. If a margin is based on a prediction, its interpretation depends on the prediction assumptions; reporting more decimal places does not resolve uncertainty in those assumptions.
Worked Example: Documenting a NOAEL-Based Calculation
Hypothetical data. In a repeated-dose study in a relevant animal species, the NOAEL was 3 mg/kg. For this example, the documented body-size conversion factor is 0.32, and the selected safety factor is 10. The proposed adult body weight for expressing the total dose is 60 kg. The record proposes rounding the calculated total dose downward to 5 mg.
Calculation. The HED is \(3\ \text{mg/kg} \times 0.32 = 0.96\ \text{mg/kg}\). Applying the safety factor gives \(0.96\ \text{mg/kg} \div 10 = 0.096\ \text{mg/kg}\). At 60 kg, this corresponds to \(0.096\ \text{mg/kg} \times 60\ \text{kg} = 5.76\ \text{mg}\).
The proposed 5 mg is below the calculated 5.76 mg. Expressed per kilogram for the assumed body weight, it is \(5\ \text{mg} \div 60\ \text{kg} = 0.0833\ \text{mg/kg}\). The ratio of the calculated HED to this proposed dose is \(0.96\ \text{mg/kg} \div 0.0833\ \text{mg/kg} \approx 11.5\). This arithmetic describes the dose reduction relative to the HED; it does not prove that all relevant risks are covered by a factor of 11.5.
What the documentation still needs. It should identify the study and its relevant findings, explain why the conversion factor and safety factor were selected, and say whether other evidence gives a lower candidate dose. It should state that 60 kg is an expression assumption, not a claim about every participant, and record the practical reason for rounding to 5 mg.
Explain the Controlling Evidence and Assumptions
When several evidence streams disagree, the record should make the decision logic visible. State which candidate was lower or more restrictive, which one controlled the proposed dose, and why. The reason may involve the relevance of a pharmacology result, a toxicology finding, an exposure endpoint, or limitations in the evidence. A bare statement that “the most conservative value was used” is not enough if the values refer to different endpoints or have different levels of reliability.
Assumptions should be specific enough that a reviewer can tell what would happen if they were false. Examples include assuming a conversion factor applies to the studied species, assuming exposure scales predictably over the dose range used, or assuming an assay captures the relevant pharmacological response. Distinguish assumptions supported by data from those adopted because direct evidence is unavailable. If the selected dose depends heavily on an uncertain assumption, say so prominently rather than hiding it in a general limitations paragraph.
A short justification can be made auditable by connecting each decision to evidence and an implication:
Name the finding or candidate dose that limits the proposed starting dose, and state why it is relevant.
Provide the source values, equations, units, conversions, and any rounding used.
State assumptions and evidence limitations that could change the meaning of the calculated margin or candidate dose.
Describe what remains uncertain and how the initial clinical plan will detect or limit the associated risk.
Worked Example: Recording Which Margin Controls
Hypothetical data. Two candidate doses are under consideration: 0.05 mg from a toxicology-based approach and 0.02 mg from a pharmacology-based approach. At 0.05 mg, predicted Cmax is 0.10 exposure units and predicted AUC is 75 exposure units. At 0.02 mg, the corresponding predictions are 0.04 and 30 units. The relevant toxicology comparator values are 1.2 for Cmax and 600 for AUC. Assume the predictions are applicable across these doses for this illustration.
At 0.05 mg, the Cmax margin is \(1.2 \div 0.10 = 12\), and the AUC margin is \(600 \div 75 = 8\). At 0.02 mg, the Cmax margin is \(1.2 \div 0.04 = 30\), and the AUC margin is \(600 \div 30 = 20\). These are endpoint-specific ratios; they are not interchangeable.
Suppose the decision is to start at 0.02 mg because the pharmacology-based candidate is lower. The rationale should say that this candidate controls, report both exposure margins at the selected dose, and identify AUC as the smaller of the two toxicology exposure margins in this example. It should not claim that an AUC margin of 20 establishes safety. The conclusion is conditional on the exposure predictions, comparator relevance, and the stated assumption that the predictions apply at the selected dose.
Documentation implication. Record the origin and limitations of the exposure values and the pharmacology evidence. If the exposure predictions are uncertain or do not represent a relevant risk, the numerical ratios may create an appearance of precision without settling the underlying decision.
Residual Risk, Diagnostics, and Limitations
Residual risk is the risk that remains after selecting and justifying the proposed dose. It may arise because no available species fully represents human pharmacology, a safety endpoint is not well characterized, an assay has limited coverage, or a prediction depends on assumptions that cannot yet be checked. Name the uncertainty and the possible consequence in plain scientific terms. Do not imply that a risk is absent merely because it was not observed in a small or limited data set.
For each material residual risk, document a proportionate response. Depending on the concern, this may include an appropriate starting-dose rationale, staged participant dosing, a dosing interval that allows review of emerging information, relevant clinical observations or measurements, and criteria for pausing or reassessing. These actions should be connected to the specific risk and the clinical plan; a generic statement that participants will be monitored is not a substitute for saying what information matters.
Several signs suggest that the rationale needs further review: a unit changes without explanation; a reported margin cannot be reproduced; the comparator endpoint differs from the predicted endpoint; the selected dose is higher than a candidate derived from relevant evidence with no justification; or an important assumption is described as established fact. Sensitivity checks from the earlier material can help identify whether a decision changes under plausible alternative inputs, but the record should describe the assumptions tested and the decision implication, not merely say that “sensitivity was assessed.”
A decision record also has limits. It cannot remove uncertainty in the underlying studies, establish a human response from animal evidence, or guarantee that a rare or unanticipated effect will not occur. Its value is that it makes the reasoning inspectable and supports reconsideration when new evidence, a changed formulation, or a revised clinical plan alters a relevant assumption.
Worked Example: Making an Unresolved Risk Actionable
Hypothetical situation. A biologic has a low proposed starting dose supported by the available pharmacology and toxicology evidence. However, the relevant assay has limited sensitivity at low concentrations, so the degree of target engagement near the proposed dose is uncertain. The documentation should not say “no target-mediated risk is expected” solely because the assay did not show a response.
A more precise record would say that the selected dose is based on the stated evidence and calculation path, while low-level target engagement cannot be excluded because of the assay’s sensitivity limitation. It should identify the potential consequence if engagement occurs and connect that concern to a planned observation or measurement capable of informing the decision. It should also specify what finding would trigger review or a pause, in a manner consistent with the clinical plan.
This does not assign a numerical probability to an unquantified risk. It distinguishes what is known from what is not known and makes the residual uncertainty relevant to the planned safeguards. If the assay limitation makes the dose rationale too uncertain to support the proposed dose, documentation alone cannot fix that scientific gap; the evidence or dose decision may need reassessment.
Common Documentation Pitfalls
- Reporting only the final dose. Without source values, units, conversions, and rounding, the result cannot be independently checked. Show the full calculation path.
- Calling a margin a probability of safety. A ratio compares two exposure values; it is not the probability that harm will or will not occur.
- Mixing evidence and judgment. Label measurements, predictions, and decision choices separately so that a judgment is not mistaken for a study finding.
- Leaving assumptions implicit. State assumptions that could materially change the candidate dose or its interpretation, and explain the consequence if they fail.
- Listing residual risks without a response. Connect each material unresolved concern to an appropriate action or explain why no additional action is warranted.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions.
- A record states “the proposed dose has a 20-fold margin.” What additional information is needed to interpret that statement?
- A hypothetical HED is 0.40 mg/kg and the safety factor is 8. Calculate the resulting candidate dose in mg/kg. If a 50 kg adult is used only to express a total dose, calculate that total.
- At a proposed dose, a relevant comparator Cmax is 2.4 units and predicted Cmax is 0.12 units. Calculate the Cmax margin and state one limitation of its interpretation.
- Name two kinds of information that should be recorded when a material assumption is uncertain, beyond simply naming the assumption.
Solutions. (1) Identify whether the margin is based on Cmax, AUC, or another endpoint; give the comparator and predicted values with units; and state the evidence source and relevant prediction assumptions. (2) \(0.40\ \text{mg/kg} \div 8 = 0.050\ \text{mg/kg}\). At 50 kg, \(0.050\ \text{mg/kg} \times 50\ \text{kg} = 2.50\ \text{mg}\). The body weight is an expression assumption, not a universal participant value. (3) \(2.4 \div 0.12 = 20\), so the Cmax margin is 20. It is a ratio, not a probability of safety, and its meaning depends on the relevance of the comparator and prediction. (4) State how the uncertainty could affect the dose or its interpretation, and identify the risk-management action or evidence that would prompt reassessment.