Why Check Starting-Dose Assumptions?
A proposed first-in-human starting dose is the result of several linked assumptions. A different toxicology species, scaling exponent, exposure measure, or safety factor can change the numerical candidate, sometimes substantially. A sensitivity check makes that dependence visible: it recalculates the candidate while changing a specified assumption and holding the other inputs constant.
The purpose is not to choose whichever assumption produces the lowest or highest dose. Nor does a change in a calculated candidate, by itself, establish which dose is appropriate. The check reveals what the calculation depends on and helps identify inputs that need stronger justification or closer review.
Set Out the Calculation Before Varying It
For a NOAEL-based candidate, a common sequence is to convert an animal dose to a human equivalent dose (HED), then apply a safety factor. If the animal and human weights are \(W_a\) and \(W_h\), respectively, and total dose is assumed to scale with body weight to exponent \(b\), the dose per unit body weight scales with exponent \(b-1\). The candidate is
Here \(D_a\) is the animal NOAEL in mg/kg, \(F\) is the safety factor, and \(D_{\text{start}}\) is the resulting human candidate in mg/kg. This equation assumes that the selected animal dose, weights, and dose units are appropriate for the conversion; that the stated power relationship is the one being tested; and that \(F\) is held fixed in a one-at-a-time scaling check. A change in \(b\) is not a statistical estimate of the correct exponent: it is a deliberate alternative assumption.
For an exposure-based calculation, a simple dose-proportional relationship can be used when it is justified for the relevant range and route. If \(D_a\) produces animal exposure \(E_a\), and \(E_h^*\) is a specified human exposure target for the same metric, the corresponding candidate with safety factor \(F\) is
This expression assumes exposure is proportional to dose over the range being considered, and that the animal exposure and human target refer to the same metric, such as Cmax or AUC. It does not make Cmax and AUC interchangeable. A separate calculation is required for each metric.
Worked Example: Changing the Allometric Exponent
Suppose a hypothetical rat NOAEL is 20 mg/kg, the rat weighs 0.25 kg, the reference human weighs 60 kg, and the safety factor is 10. Compare exponents \(b=0.67\), \(0.75\), and \(0.80\). Assume the same route and dose definition are being compared, and that the body-weight scaling equation above is the specified conversion.
The weight ratio is \(0.25/60=0.004167\). For \(b=0.75\), the exponent on that ratio is \(1-b=0.25\):
Applying the safety factor gives \(5.081/10=0.5081\) mg/kg. Repeating the calculation for the alternatives:
In this hypothetical comparison, the candidate increases as the exponent increases. The calculations isolate the exponent’s effect because the animal NOAEL, weights, and safety factor are unchanged. They do not establish which exponent is best supported for the drug.
Why the Exponent Changes the Candidate
The direction of the effect can be checked algebraically rather than inferred from a table. Let \(r=W_a/W_h\). For the allometric candidate,
Differentiating the logarithm with respect to \(b\) gives
When the animal weighs less than the reference human, \(0<r<1\), so \(\ln r<0\) and \(-\ln r>0\). Thus the candidate increases as \(b\) increases, provided the other inputs are fixed. This is a property of this formula and weight comparison, not a general claim that higher exponents always produce higher doses in every scaling setup.
Compare Species Without Hiding the Inputs
Species sensitivity can be checked by calculating a candidate from each relevant species’ toxicology input under a consistent set of conversion rules. The comparison is meaningful only when the underlying dose endpoints, route, duration, and units are sufficiently comparable for the intended calculation. If they are not, present the candidates as distinct evidence rather than as directly interchangeable estimates.
Worked Example: Comparing Two Species-Based Candidates
Suppose a hypothetical comparison has already converted the rat and dog NOAELs to HEDs of 6.1479 mg/kg and 7.3338 mg/kg, respectively, using the same reference human and stated conversion approach. Use a safety factor of 10 for both candidates. This isolates the effect of selecting the species-based input; it does not assert that the two toxicology findings are otherwise equivalent.
For the rat-based candidate:
For the dog-based candidate:
The dog-based candidate relative to the rat-based candidate is
Thus, the rat-based candidate is about \(1-0.8383=0.1617\), or 16.2%, below the dog-based candidate. This percentage describes the numerical comparison; it does not show that one species is more relevant or that either candidate is more protective. Those judgments require review of the species’ relevance and the evidence supporting each input.
Check Exposure Metrics Separately
An exposure-based candidate can depend on which exposure metric is used. Cmax describes a peak concentration, whereas AUC summarizes exposure over time. A calculation using one metric cannot stand in for a calculation using the other. Before comparing results, confirm that each animal value and human target uses the same metric and compatible units, and justify any assumption that exposure changes proportionally with dose.
Worked Example: Comparing Cmax- and AUC-Based Candidates
In a hypothetical linear, dose-proportional setting, an animal NOAEL of 15 mg/kg is associated with Cmax of 12 mg/L and AUC of 90 mg·h/L. Suppose the specified human exposure targets are Cmax of 0.6 mg/L and AUC of 3.0 mg·h/L, and use a safety factor of 3 in both calculations. Assume the route and exposure definitions match and proportionality is considered appropriate over this range.
Using Cmax, the dose ratio is \(0.6/12=0.05\). Therefore:
Using AUC, the exposure ratio is \(3.0/90=0.03333\). Therefore:
The AUC-based candidate is lower in this example. The result depends on the specified targets and the dose-proportionality assumption; it does not establish which metric should control a starting-dose decision.
Vary the Safety Factor Deliberately
A safety-factor sensitivity check holds the exposure or HED calculation fixed and changes only \(F\). Since the candidate is inversely proportional to \(F\), doubling \(F\) halves the calculated candidate. This arithmetic is straightforward, but it is not a method for choosing the factor: that requires a rationale tied to the evidence and risks under consideration.
Worked Example: Changing the Safety Factor
Suppose the hypothetical HED is 5.081 mg/kg. Calculate candidates using safety factors of 3, 10, and 20, holding the HED fixed.
The factor-20 candidate is half the factor-10 candidate because \(10/20=0.5\). The factor-3 candidate is higher because it applies a smaller divisor. These comparisons describe how the calculation responds to \(F\); they do not say that any particular factor is warranted.
Organise the Checks and Inspect Their Limits
A transparent check starts with a clearly specified baseline, changes inputs in a controlled way, and reports the resulting candidates with units. The following sequence helps keep the interpretation disciplined.
Write down the species, dose endpoint, weights, scaling exponent, exposure metric if used, safety factor, and units.
Change one input at a time to assess its individual numerical effect, or define a combined scenario and label it as such.
Substitute the alternative value into the same stated formula. Check unit compatibility, arithmetic, and whether the assumptions still apply.
Report how the candidate changes and identify the assumption responsible, without treating the result as proof that the alternative is clinically preferable.
A one-at-a-time table is useful when the purpose is to see which assumptions drive the arithmetic. A combined scenario is useful when several alternatives belong together, but the resulting difference cannot be attributed to a single input. More detailed uncertainty analysis is needed to describe uncertainty quantitatively; a small set of sensitivity values is not a probability distribution.
Common diagnostics include recalculating at least one result by hand, checking that all dose values are expressed in compatible units, and confirming that the direction of change agrees with the formula. For example, with \(W_a/W_h<1\), the exponent result above should increase as \(b\) increases. An unexpected direction can signal a misplaced exponent, a reversed weight ratio, or a conversion error.
Exposure checks need additional scrutiny. Nonlinear pharmacokinetics can make exposure fail to scale proportionally with dose. A mismatch in route, sampling schedule, analyte, or exposure definition can also invalidate a direct ratio. Likewise, a species comparison may be misleading if the relevant toxicology endpoint, duration, or pharmacology differs materially. Sensitivity checks reveal dependence on the chosen inputs, but cannot repair inputs or assumptions that are unsuitable.
Check Your Understanding
Answer each question, then compare with the brief solutions.
- If \(W_a/W_h<1\) in the stated HED equation, what is the direction of change in the candidate as \(b\) increases?
- A fixed HED is 4.8 mg/kg. What candidates result from safety factors of 8 and 16?
- Why must a Cmax-based exposure calculation not be substituted for an AUC-based calculation?
- In a one-at-a-time check, the species changes and the safety factor also changes. Can the resulting difference be attributed to species alone? Explain.
Solutions. (1) Since \(d\ln D_{\text{start}}/db=-\ln(W_a/W_h)>0\), the candidate increases as \(b\) increases. (2) \(4.8/8=0.600\) mg/kg and \(4.8/16=0.300\) mg/kg. (3) Cmax and AUC are different exposure metrics with different units and meanings; each requires its own matched animal exposure and human target. (4) No. Two inputs changed, so the comparison is a combined scenario and cannot isolate the species effect.