Pediatric Dose Extrapolation from Adult PK Models

Why Adult Models Need More Than a Weight Conversion

An adult pharmacokinetic (PK) model can provide a useful starting point for pediatric dosing, but its parameters cannot generally be carried over unchanged. Children differ from adults in body size, and the processes responsible for drug elimination may also be developmentally immature. A defensible extrapolation therefore adjusts relevant model parameters for both size and maturation, then checks whether the resulting predictions are plausible for the child and the intended dosing question.

The key distinction is that size and maturation are not interchangeable adjustments. A size function describes how a parameter changes with body size; a maturation function describes a developmental change in a physiological process, such as the capacity of an elimination pathway. The appropriate function and its parameters depend on the drug and pathway. There is no single age adjustment that is correct for every medicine.

Definition: Pediatric dose extrapolation from an adult PK model uses an adult model as a reference and adjusts its parameters for relevant differences between the adult and child, commonly body size and maturation. The extrapolated model predicts pediatric exposure under stated assumptions; it does not, by itself, establish that the prediction is accurate or that the resulting regimen is clinically appropriate.

Scaling Clearance and Volume

A commonly used starting point for body-size scaling is allometry. Let \(WT\) be body weight in kilograms, and let \(CL_{70}\) and \(V_{70}\) be the adult model's typical clearance and volume of distribution, respectively, referenced to a 70-kg person. A conventional model uses an exponent of 0.75 for clearance and 1 for volume:

CL_{\mathrm{size}}=CL_{70}\left(\frac{WT}{70}\right)^{0.75}, \qquad V_{\mathrm{size}}=V_{70}\left(\frac{WT}{70}\right)^1.

These exponents are modeling conventions supported as useful starting points in many settings, not biological laws that must fit every drug. Pediatric data or strong mechanistic knowledge may support different exponents or a different size descriptor. The 70-kg reference is a scaling anchor, not an assertion that every adult weighs 70 kg.

For clearance influenced by development, a maturation function can be applied in addition to size scaling. One possible form is a Hill function:

F_{\mathrm{mat}}(a)=\frac{a^h}{a^h+TM_{50}^{\,h}}, \qquad CL_{\mathrm{child}}=CL_{70}\left(\frac{WT}{70}\right)^{0.75}F_{\mathrm{mat}}(a).

Here \(a\) is the selected age metric, \(TM_{50}\) is the value of that metric at which the function reaches one-half of its mature value, and \(h\) controls the steepness of the transition. The age metric must match the model and evidence used to estimate the function. The function is normalized to a mature value of one; it should not be treated as a universal maturation curve. In particular, an effect on clearance should not automatically be applied to volume. Volume may require its own model and evidence.

For a one-compartment model, concentration is related to amount in the body by \(C=A/V\), and clearance links concentration to the rate of elimination. Scaling both clearance and volume can therefore affect the concentration-time profile. However, under linear PK, it is clearance—not volume—that determines AUC for a specified systemic dose. Volume and other model parameters matter when predicting the shape and timing of concentrations, including peak concentrations.

From Scaled Clearance to Exposure

The AUC relationship follows from mass balance, provided its conditions are stated. Consider a repeated-dose regimen at steady state over a dosing interval of length \(\tau\). Assume clearance is constant over the interval, PK is linear, and the systemic input during the interval is \(F D\), where \(D\) is the administered dose and \(F\) is the fraction reaching systemic circulation. At steady state, the amount in the body at the end of the interval equals the amount at its beginning. Thus, net change in body amount is zero: systemic input equals elimination over the interval. Since elimination over the interval is \(CL\) times the concentration integrated over time, the result is:

F D=CL\int_0^\tau C(t)\,dt=CL\cdot AUC_{\tau,ss}, \qquad AUC_{\tau,ss}=\frac{F D}{CL}.

For intravenous dosing, \(F=1\). For an oral regimen, the formula uses bioavailability \(F\), if it is known and adequately represented by the model. It is not valid to infer this steady-state interval balance for an arbitrary interval in which the amount in the body changes. For a single dose under linear PK, the corresponding AUC relationship applies when integrating over the complete elimination period.

To target a specified steady-state AUC, rearrange the equation to obtain \(D=AUC_{\mathrm{target}}\,CL/F\). The predicted dose is proportional to clearance when the target AUC and bioavailability are held fixed. A calculation based on this relationship targets AUC; it does not guarantee a particular peak, trough, or safety outcome.

Worked Example: Scaling Clearance and Setting an Intravenous Dose

Hypothetical data. An adult model estimates \(CL_{70}=5.00\) L/h. For a 12-kg child, suppose the relevant maturation function has been estimated as 0.80. Assume the child follows the model, clearance is constant, and the drug has linear PK. What IV dose at steady state targets an AUC over the dosing interval of 25.0 mg·h/L?

First scale clearance for body size and maturation:

CL_{\mathrm{child}} =5.00\left(\frac{12}{70}\right)^{0.75}(0.80) =5.00(0.266417289\ldots)(0.80) =1.065669156\ldots\ \mathrm{L/h}.

For IV administration \(F=1\). The steady-state dose for the target AUC is:

D=\frac{AUC_{\mathrm{target}}\,CL_{\mathrm{child}}}{F} =(25.0\ \mathrm{mg\cdot h/L})(1.065669156\ldots\ \mathrm{L/h}) =26.6417289\ldots\ \mathrm{mg} \approx 26.64\ \mathrm{mg}.

The units reduce to milligrams. Under the stated assumptions, this is the dose per interval that predicts the target AUC. It is a model-based calculation, not a clinical recommendation; the regimen, dose rounding, formulation, and safety must also be evaluated.

Worked Example: Oral Exposure and a Target AUC

For oral dosing, both clearance and bioavailability contribute to the predicted AUC. The following hypothetical calculation illustrates why omitting \(F\) changes the dose.

Worked Example: Calculating an Oral Dose for a Child

Hypothetical data. An adult model has \(CL_{70}=6.00\) L/h and \(V_{70}=40.0\) L. A 20-kg child has a maturation factor of 0.95 for clearance. Assume oral bioavailability \(F=0.80\), linear PK, and a target steady-state AUC per interval of 30.0 mg·h/L. Calculate scaled clearance, scaled volume, and the oral dose for that target.

The size factor for clearance is \((20/70)^{0.75}=0.390794965\ldots\). Thus:

CL_{\mathrm{child}} =6.00(0.390794965\ldots)(0.95) =2.2275313\ldots\ \mathrm{L/h}.

Using the stated size exponent for volume, with no additional maturation adjustment assumed for volume:

V_{\mathrm{child}} =40.0\left(\frac{20}{70}\right) =11.4286\ \mathrm{L}.

The oral dose is obtained from the steady-state AUC equation:

D=\frac{AUC_{\mathrm{target}}\,CL_{\mathrm{child}}}{F} =\frac{(30.0\ \mathrm{mg\cdot h/L})(2.2275313\ldots\ \mathrm{L/h})}{0.80} =83.5324\ldots\ \mathrm{mg} \approx 83.53\ \mathrm{mg}.

The volume estimate may help describe concentration-time behavior, but it does not enter this AUC-based dose calculation. If instead an 80.0-mg oral dose were given per interval, predicted AUC would be:

AUC_{\tau,ss} =\frac{(0.80)(80.0\ \mathrm{mg})}{2.2275313\ldots\ \mathrm{L/h}} =28.73\ \mathrm{mg\cdot h/L}.

This is below the 30.0 mg·h/L target under the model. The calculation assumes the stated bioavailability applies to this child and regimen; if \(F\) is uncertain or changes with dose, the prediction changes.

Assumptions, Diagnostics, and Failure Modes

Extrapolation is most credible when the adult model captures the drug's relevant disposition and the pediatric changes are represented adequately by the selected covariates. In practice, this requires more than inserting a child's weight and age into equations. Consider whether the adult data support the base parameter estimates, whether pediatric physiology changes the same elimination pathway, whether the maturation function is based on a suitable age metric, and whether the child lies within the range over which the model can reasonably be used. The model also needs appropriate assumptions about dose proportionality, route, bioavailability, and the stability of relevant covariates over the prediction interval.

When pediatric concentration or exposure data are available, compare predictions with observations. Useful checks include prediction errors, observed-versus-predicted plots, and residual patterns against weight and age. Systematic underprediction in younger children, for example, can indicate that the maturation function or its age metric is inadequate; a pattern across body size can point to a poor size relationship. Check uncertainty as well as the central prediction. Uncertainty in adult clearance, maturation parameters, and bioavailability can compound, so a precise-looking dose calculation may still have a wide prediction range.

A simple sensitivity check can show the consequence of a modeling assumption. In the first worked example, omitting maturation would give clearance \(5.00(12/70)^{0.75}=1.332086445\ldots\) L/h and a target dose of 33.30 mg. If the true clearance followed the stated maturation factor of 0.80, that dose would predict an AUC of \(33.3022/1.065669\approx31.25\) mg·h/L, rather than 25.0 mg·h/L. The discrepancy is not a small rounding effect: it follows directly from leaving out a relevant clearance adjustment.

Alternative approaches answer related but different questions. A simple mg/kg dose assumes a direct proportionality between dose and body weight; it does not by itself represent nonlinear scaling of clearance or developmental maturation. An allometric model adds a size relationship, while a size-and-maturation model adds a developmental effect on specified parameters. A model fitted to pediatric data can estimate child-specific relationships more directly when the data are adequate, but may be unstable if the pediatric evidence is sparse. Physiologically based pharmacokinetic models can represent organ and pathway mechanisms in greater detail, but require suitable physiological and drug-specific inputs. None of these approaches removes the need to evaluate evidence and uncertainty.

Worked Example: Checking the Consequence of Omitting Maturation

Return to the hypothetical 12-kg child, with \(CL_{70}=5.00\) L/h and maturation factor 0.80. Suppose someone uses size scaling but omits maturation, then administers the resulting IV dose for a target AUC of 25.0 mg·h/L. What AUC would the maturation-adjusted model predict?

The dose based on size scaling alone would be:

D_{\mathrm{size\ only}} =(25.0\ \mathrm{mg\cdot h/L}) \left[5.00\left(\frac{12}{70}\right)^{0.75}\ \mathrm{L/h}\right] =33.3022\ldots\ \mathrm{mg}.

Under the model that includes maturation, clearance is \(1.065669156\ldots\) L/h. Therefore, the AUC predicted for that dose is:

AUC_{\tau,ss} =\frac{33.3022\ldots\ \mathrm{mg}}{1.065669156\ldots\ \mathrm{L/h}} =31.25\ \mathrm{mg\cdot h/L}.

The predicted AUC is 25% above the target because the size-only dose was based on clearance 25% larger than the maturation-adjusted value. This comparison assumes the maturation factor is valid; it illustrates the implication of that assumption, not proof that the factor is correct.

Common Pitfalls and Practical Limits

Using a universal maturation curve. Different elimination pathways mature differently. Use a function supported for the relevant drug or pathway, and state its age metric and parameter basis.
Applying clearance maturation to every parameter. A clearance adjustment does not automatically justify the same adjustment to volume, absorption, or bioavailability.
Treating allometry as exact. The 0.75 and 1 exponents are common starting choices, not guarantees. Check whether evidence supports them for the drug and population.
Equating a target AUC with a complete regimen recommendation. AUC alone does not determine peak and trough concentrations, formulation suitability, or clinical benefit and risk.
Using steady-state mass balance outside its conditions. Over a steady-state dosing interval, net change in body amount is zero. At non-steady state, input minus elimination equals the change in body amount, so the same interval argument cannot be used without accounting for that change.

Adult-to-child extrapolation is valuable because it makes assumptions explicit and can guide predictions when pediatric data are limited. Its main limitation is equally important: numerical output does not validate the model. The strongest application combines a plausible structural model, justified size and maturation functions, suitable pediatric evidence where available, and clear reporting of the conditions and uncertainty attached to the prediction.

Check Your Understanding

Answer each question, then compare your reasoning with the brief solutions.

A linear IV drug has a steady-state interval dose of 40 mg and clearance of 2 L/h. What is its predicted AUC per interval?
A child has \(WT=35\) kg. Using the conventional volume exponent of 1, what fraction of \(V_{70}\) does the size-scaled volume equal?
Why is it not generally appropriate to apply a clearance maturation factor automatically to volume?
If oral bioavailability decreases while dose and clearance remain unchanged, what happens to predicted steady-state AUC?

Solutions. (1) \(AUC_{\tau,ss}=D/CL=40/2=20\) mg·h/L, using \(F=1\) for IV dosing and assuming the stated steady-state linear model. (2) \(V/V_{70}=35/70=0.500\), so the scaled volume is one-half the reference volume. (3) Maturation is parameter- and pathway-specific; evidence for an effect on clearance does not establish the same developmental effect on volume. (4) AUC decreases in direct proportion to \(F\), because \(AUC_{\tau,ss}=FD/CL\), under the assumptions of linear PK and unchanged clearance.