Why Compare Age-Band and Model-Based Doses?
Pediatric regimens often assign one dose to a range of ages. This is practical, but children do not all have the same clearance within an age band, and clearance may change as a child grows. A dose that gives the desired exposure near the middle of a band can therefore produce different predicted exposures at its lower and upper ends. A second issue arises at the boundary: when the assigned dose changes abruptly but clearance changes smoothly, predicted exposure can jump even for children whose ages differ by only a small amount.
A continuous model-based regimen instead uses a model-predicted clearance that changes with age, and potentially with other supported covariates, to calculate a dose. The comparison is about predicted exposure under stated assumptions; it does not by itself establish clinical efficacy, safety, or the best practical regimen.
Linking Dose to Predicted Exposure
For an oral drug with linear pharmacokinetics, the AUC over a dosing interval at steady state is \(F D/CL\), where \(D\) is dose per interval, \(F\) is bioavailability, and \(CL\) is clearance. It is convenient to express the relationship using apparent clearance, \(CL/F\). The same exposure equation applies to a single dose's total AUC under linear pharmacokinetics. In this tutorial, \(D\) is the dose per interval and AUC refers to exposure over that interval at steady state.
Here \(a\) is age, \(CL/F(a)\) is model-predicted apparent clearance in L/h, and \(D(a)\) is the dose in mg per interval. The resulting AUC is in mg·h/L. If the route, formulation, and bioavailability are consistent across ages, changes in apparent clearance can represent the age-related exposure differences of interest.
The equation requires dose-proportional exposure over the dose range being considered, a suitable clearance model for the population and ages evaluated, and comparable dosing conditions. If bioavailability changes with age or formulation, it must be accounted for in the model or the apparent-clearance prediction; otherwise, the predicted AUC comparison may be misleading. A typical population prediction describes a typical child with the specified covariates, not every child in that age group.
Suppose a target interval AUC, \(AUC_{\text{target}}\), is chosen. Solving the exposure equation for dose gives the continuous model-based dose:
This calculation makes predicted AUC equal to the target at each age for which the model predicts apparent clearance. It does not guarantee that each child will attain the target: individual clearance may differ from the model prediction, and the model itself is uncertain.
Worked Examples
Worked Example: Exposure Variation Within One Age Band
Assume, hypothetically, that children in one age band receive 57.70 mg per dosing interval. The model predicts apparent clearances of 0.5333 L/h at the younger end and 0.6286 L/h at the older end. Assume linear pharmacokinetics and the same formulation across the band. Calculate predicted AUC at each end.
At the younger end:
At the older end:
If the illustrative target is 100 mg·h/L, the predicted AUC is about 8.19% above target at the younger end and 8.21% below target at the older end. These values are calculations from the displayed dose and clearances; small differences may result if unrounded model predictions are used. The fixed dose does not keep predicted exposure constant as apparent clearance changes.
Worked Example: Exposure Jump at an Age Boundary
Consider a hypothetical regimen that gives 60 mg per interval below age 6 years and 75 mg per interval from age 6 years onward. Suppose the model predicts \(CL/F=0.600\) L/h immediately before the boundary and \(CL/F=0.610\) L/h immediately after it. Use the same linear-PK assumptions as above.
Immediately before age 6:
Immediately after age 6:
Thus, predicted exposure increases by approximately 22.95% across the boundary: \((122.9508-100.0)/100.0\times100\%\). The dose rises by 25%, while predicted apparent clearance rises by only about 1.67%. If the target is 100 mg·h/L, a continuous model-based calculation would give 60 mg at 0.600 L/h and 61 mg at 0.610 L/h. Those doses predict 100 mg·h/L on both sides of the boundary under the stated assumptions.
Worked Example: Typical-Age Dosing and Individual Prediction
Suppose a hypothetical child’s age corresponds to a typical model-predicted \(CL/F\) of 0.600 L/h, and the target AUC is 100 mg·h/L. The age-based continuous dose is:
If two children of the same age have actual apparent clearances of 0.480 L/h and 0.720 L/h, the 60.0 mg dose would predict:
If valid individual covariate information instead supported clearance predictions of 0.480 L/h and 0.720 L/h, the corresponding model-based doses for the same target would be 48.0 mg and 72.0 mg. Each would predict 100 mg·h/L when divided by its corresponding apparent clearance. This is an illustration of the calculation, not evidence that individual clearance can always be predicted accurately. Residual variability and parameter uncertainty remain, even when a continuous model is used.
Why Continuous Dosing Can Remove an Age-Driven Pattern
The key property follows directly from the dose equation. If the model-based dose is calculated as \(AUC_{\text{target}}\times CL/F(a)\), substitute that dose into the exposure equation:
The cancellation shows why this approach removes the systematic exposure change caused solely by the modeled age trend in clearance. It depends on using the same clearance prediction in both the dose calculation and the exposure prediction, and on the model and linear-PK assumptions being adequate. It does not show that observed exposures will be identical across ages.
For a fixed age-band dose \(D_j\), where \(j\) identifies the band, exposure instead varies inversely with predicted apparent clearance. If clearance rises with age within a band, predicted AUC falls. At a boundary, the ratio of exposure just after and just before is:
This expression separates two effects: the dose change and the clearance change. A dose increase may be larger than, smaller than, or equal to the increase in clearance. Only when their ratios match does predicted AUC remain unchanged across the boundary.
Comparison, Diagnostics, and Limitations
| Feature | Fixed age-band regimen | Continuous model-based regimen |
|---|---|---|
| Dose rule | One dose is assigned to everyone in an age interval. | Dose is calculated from predicted apparent clearance and a chosen exposure target. |
| Variation within a band | Predicted AUC changes as clearance changes while dose stays fixed. | The modeled age-related change can be offset by a dose that changes with predicted clearance. |
| At an age boundary | A dose step can produce an exposure jump if it does not match the clearance change. | A smooth clearance model gives a smooth calculated dose before any practical adjustments. |
| Main practical consideration | Simple to prescribe, but the chosen bands and doses may poorly match developmental change. | Depends on a suitable model and on being able to prescribe the calculated dose. |
A useful diagnostic is to plot predicted AUC against age for the proposed regimen, using a fine age grid that includes points just below and above every band boundary. Plotting dose and predicted apparent clearance on the same age grid can help identify whether an exposure jump comes from the dose rule or the model prediction. Also inspect within-band minima and maxima rather than relying only on the band’s midpoint: a midpoint can conceal substantial exposure variation at the edges.
Check that the clearance model is supported across the ages being evaluated, and compare predictions with available data across developmental stages. A sharp change in predicted clearance at an age where biology is expected to change smoothly may indicate a model or data problem; conversely, a smooth average model does not rule out meaningful variation between children. Predictions outside the model’s data support should be treated cautiously.
Age bands remain useful when operational simplicity, available dose increments, or the evidence base favors categories. More, narrower bands can approximate a continuous dose curve, but also create more thresholds at which dose changes must be checked. A continuous model reduces the age-related trend in typical predicted exposure under its assumptions, but can be sensitive to model misspecification and individual variability. Neither approach removes the need to assess uncertainty and clinical suitability. The calculations here are not a method for choosing an exposure target.
Common Pitfall
Wrong reasoning: “The dose is constant within an age band, so predicted exposure is constant there.” This confuses dose with exposure. Under the stated model, \(AUC_{\tau}=D/(CL/F)\); if apparent clearance changes while dose stays fixed, predicted AUC changes too. A second mistake is to assume that a continuous dose rule guarantees the target AUC for every child. It guarantees the target only for the clearance value used in that calculation, assuming the model and pharmacokinetic relationship are correct.
Correction: Calculate exposure at relevant ages using the corresponding clearance predictions, and inspect both within-band variation and the values on either side of each boundary. Interpret model-based doses as predictions conditional on the model, not as certainty about an individual child’s exposure.
Check Your Understanding
Use \(AUC_{\tau}=D/(CL/F)\) and show units in your calculations.
- A fixed dose of 45 mg is given across an age band. Predicted apparent clearance is 0.450 L/h at one end and 0.600 L/h at the other. What are the two predicted AUC values?
- A regimen changes from 40 mg to 50 mg at an age boundary. Predicted apparent clearance changes from 0.400 L/h to 0.420 L/h. Calculate predicted AUC on each side and the percentage change across the boundary.
- For a target AUC of 90 mg·h/L and predicted apparent clearance of 0.550 L/h, what continuous model-based dose is calculated?
- Why can a continuous age-based dose still produce different exposures in two children of the same age?
Solutions: 1. \(45/0.450=100.0\) mg·h/L and \(45/0.600=75.0\) mg·h/L. 2. Before the boundary, \(40/0.400=100.0\) mg·h/L; after it, \(50/0.420=119.05\) mg·h/L. The increase is \((119.05-100.0)/100.0\times100\%=19.05\%\), approximately. 3. \(90\ \mathrm{mg\cdot h/L}\times0.550\ \mathrm{L/h}=49.5\) mg. 4. Age-based dosing uses a model prediction, while individual children may have different actual clearances or bioavailability; model and residual variability mean their actual exposures need not equal the typical prediction.