Why Check the Boundary?
An age-band rule assigns a recommended regimen according to which side of a cutoff a child falls on. If the rule changes abruptly at that cutoff, two children with nearly identical ages and other characteristics can receive different regimens. That may be intentional, but it should be checked against the model-predicted exposure rather than assumed to be harmless or appropriate.
The key comparison is not simply whether the prescribed dose changes. Clearance can also change with age, and the dosing interval may change along with the dose. A larger dose can therefore produce a smaller-than-expected exposure change, no change, or even an exposure change in the opposite direction. The boundary check makes these effects explicit.
Set Up the Comparison
First write the rule precisely: specify which ages belong to each band, the dose per administration, the dosing interval, and any applicable dose cap or formulation constraint. The equality convention matters. For example, a rule may assign age \(c\) to the older band; the comparison still needs predictions immediately below and above \(c\), not an ambiguous calculation at the cutoff.
Next specify the exposure metric and the model used to predict it. Under linear pharmacokinetics at steady state, with constant bioavailability \(F\), clearance \(CL\), dose per administration \(D\), and dosing interval \(\tau\), the area under the concentration-time curve over one dosing interval is \(F D/CL\). If the interval divides 24 hours, the 24-hour AUC is the interval AUC multiplied by \(24/\tau\). Equivalently, the average steady-state concentration is the 24-hour AUC divided by 24 hours.
Here \(D\) is in mg per administration, \(CL\) is in L/h, and \(\tau\) is in hours; \(AUC_{24}\) is in mg·h/L. This relation assumes linear, time-invariant pharmacokinetics over the interval and a stable regimen. If clearance depends on concentration, bioavailability changes with dose, or a regimen is not at steady state, use an appropriate model prediction instead of treating this formula as exact.
For a fixed interval and the same \(F\), the exposure ratio across the boundary is
The subscripts \(-\) and \(+\) denote predictions just below and just above the cutoff. If the interval also changes, its effect must be included through \(24/\tau\). This ratio is useful because it separates the contribution of the dose rule from the change in modeled clearance.
Worked Boundary Checks
Worked Example: A Dose Increase at a Two-Year Cutoff
Hypothetical rule and model. A once-daily regimen recommends 10 mg/kg below age 2 years and 15 mg/kg from age 2 years onward. To isolate the rule’s boundary effect, suppose weight is 12.0 kg on both sides, \(F=0.80\), and model-predicted clearance is 12.0 L/h immediately below the cutoff and 12.5 L/h immediately above it. Assume the linear steady-state exposure relation applies.
Below the cutoff, \(D_-=10\times12.0=120\) mg per day. Above it, \(D_+=15\times12.0=180\) mg per day. Because \(\tau=24\) h, the 24-hour AUC is \(F D/CL\):
The modeled exposure rises by \(11.52-8.00=3.52\) mg·h/L, or \(3.52/8.00\times100=44.0\%\). The dose rises by 50%, while clearance rises by about 4.17%; both changes explain why the exposure increase is slightly smaller than the dose increase. If the hypothetical target were 9.00 mg·h/L, the lower-band prediction would be below target and the upper-band prediction above it. That result would warrant review of the rule; it would not, by itself, establish that either regimen is clinically unacceptable.
Worked Example: A Small Growth Change Does Not Remove a Rule-Induced Jump
Hypothetical rule and model. At age 6 years, a child moves from 8 mg/kg once daily to 10 mg/kg once daily. Consider a child weighing 20.00 kg just below the cutoff and 20.02 kg just above it. Let \(F=0.90\), with predicted clearance increasing from 4.00 to 4.02 L/h. Assume linear steady-state pharmacokinetics.
The respective doses are \(D_-=8\times20.00=160.0\) mg and \(D_+=10\times20.02=200.2\) mg. With a 24-hour interval:
The increase is \(8.82\) mg·h/L, or \(8.82/36.00\times100=24.5\%\). The weight and clearance changes across the comparison are small, so the dose-band change accounts for most of the exposure jump. If a prespecified review criterion were a boundary change greater than 20%, this hypothetical result would trigger review. That criterion would be a policy for screening, not a universal pharmacological threshold.
Worked Example: A Dose-per-Administration Jump Can Preserve Daily AUC
Hypothetical rule and model. At a cutoff, a regimen changes from 100 mg every 12 hours to 200 mg every 24 hours. Let \(F=0.75\), with clearance changing from 5.00 L/h just below the cutoff to 5.05 L/h just above it. Assume linear steady-state pharmacokinetics.
Below the cutoff there are two administrations per 24 hours, for 200 mg total per day. Above it there is one 200 mg administration per day. Thus:
The 24-hour AUC falls by \(0.30\) mg·h/L, or 1.00%. The dose per administration doubles, but the dosing interval doubles too, leaving total daily dose unchanged; the slight exposure decrease follows from the small modeled clearance increase. This comparison supports near-continuity of daily AUC under the stated assumptions. It does not show that peak and trough concentrations are continuous: those depend on the concentration-time profile, which an AUC calculation alone does not determine.
Why the Exposure Comparison Works
Under the stated linear model, exposure over one interval is proportional to the amount of drug entering the systemic circulation and inversely proportional to clearance. With interval length \(\tau\), the steady-state average concentration is the interval AUC divided by \(\tau\):
Multiplying by 24 hours gives the 24-hour AUC when the regimen is stable. This also explains why comparing dose per administration alone can mislead when \(\tau\) changes. The ratio of daily AUC predictions is
If bioavailability is the same on each side, its ratio is one. If the interval is unchanged, the interval ratio is also one. The expression then reduces to the simpler dose-clearance ratio used in the first example. Stating these cancellations is important: they are assumptions, not automatic properties of every pediatric regimen.
A continuous age-based model can also help interpret the check. If the underlying age, weight, and maturation functions change smoothly with age, the modeled clearance should usually have a smooth boundary prediction. A large predicted jump in exposure may then come from the discrete dosing rule rather than from the modeled physiology. Conversely, a developmental transition represented in the model can alter clearance appreciably near a boundary. The relevant question is whether the dose rule compensates for the predicted clearance change well enough to maintain the intended exposure, not whether the dose itself is continuous.
A Practical Review Sequence
Record the cutoff, the age convention, and which band includes the cutoff itself. Evaluate predictions at ages close to the cutoff on both sides.
State which covariates are held fixed and which follow a plausible age trajectory. Use the same model, exposure metric, and assumptions on both sides.
Include dose per administration, interval, bioavailability, and model-predicted clearance, or obtain exposure from a suitable pharmacokinetic simulation.
Report the absolute and relative change in dose and exposure, with units. Compare the predictions with the intended exposure target or range if one has been specified.
Use a clinically justified, prespecified review criterion and consider the consequence of the change. A flagged discontinuity prompts examination of the rule; it does not by itself dictate a replacement dose.
Alternatives, Diagnostics, and Limitations
An age-band rule is operationally simple and can be practical when prescribing needs clear categories. A continuous model-based dose can instead calculate a dose from predicted clearance and an exposure target. Under the linear model, the dose needed for a target \(AUC_{24}^{*}\) is
This expression yields dose per administration when \(AUC_{24}^{*}\) is the 24-hour target and \(\tau\) is the interval in hours. A hybrid approach may use age bands for operational reasons but check each band’s predicted exposure across its age and size range. These approaches serve different purposes: a continuous calculation is not automatically preferable if its inputs or predictions are not reliable, and a simple band is not automatically adequate merely because it is easy to apply.
Several diagnostics help determine whether a boundary result is informative. Recalculate using ages increasingly close to the cutoff; a genuine step in the rule should persist, whereas a difference caused only by choosing widely separated ages may shrink. Check that units, interval conversion, band assignment, and dose cap have been applied consistently. Inspect whether clearance and other model covariates change smoothly or whether an actual model transition is represented. If the predicted jump changes markedly under small, plausible changes in covariates or model assumptions, report that dependence rather than presenting one calculation as definitive.
The principal limitations are equally important. A boundary comparison assesses predictions under the selected model; it is not evidence that the model is correct for every child. AUC continuity does not guarantee similar peak concentration, trough concentration, duration above or below a threshold, safety, or efficacy. These may require a concentration-time model and endpoints beyond AUC. The simple formula also does not cover nonlinear elimination, dose-dependent absorption, changing bioavailability, missed doses, or non-steady-state conditions without further modeling. Finally, a smooth mathematical dose may not be usable as written if available formulations, dose caps, or administration constraints intervene.
Common Mistake
A boundary discontinuity warrants a revised-rule discussion when its predicted exposure effect is clinically meaningful under a justified criterion, conflicts with the intended target, or cannot be explained by the modeled developmental change. The review may lead to a different cutoff, a different band dose, a continuous or hybrid rule, or no change if the difference is justified. The calculation identifies where to investigate; it does not choose among those decisions by itself.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions below.
- A once-daily hypothetical rule changes from 6 mg/kg to 8 mg/kg at a cutoff. Weight is 15 kg on both sides, \(F=0.80\), and clearance rises from 6.0 to 6.4 L/h. Calculate both 24-hour AUC predictions and the relative change.
- Why is it insufficient to compare dose per administration when a regimen also changes from every 12 hours to every 24 hours?
- At a cutoff, predicted 24-hour AUC changes by 2%, but the regimen changes substantially in dose interval. What important conclusion cannot be drawn from AUC alone?
- Name two assumptions behind using \(AUC_{24}=FD/CL\) for a once-daily steady-state regimen.
Solutions. (1) The doses are \(6\times15=90\) mg and \(8\times15=120\) mg. Thus \(AUC_{24,-}=0.80\times90/6.0=12.0\) mg·h/L and \(AUC_{24,+}=0.80\times120/6.4=15.0\) mg·h/L. The relative increase is \((15.0-12.0)/12.0\times100=25.0\%\). (2) Daily exposure depends on the number of administrations per day as well as dose per administration; the interval factor \(24/\tau\) must be included. (3) Similar daily AUC does not establish similar peak or trough concentrations, or similar concentration-time profiles. (4) Examples include linear pharmacokinetics, stable clearance and bioavailability, and steady-state dosing. Any two, stated precisely, are sufficient.