Bridging Pediatric and Adult Pharmacokinetics

Why a Pediatric-to-Adult Bridge Needs a Shared Model

A clearance estimate for a child should not be extended to adults by changing the age label while leaving the size relationship implicit. Clearance depends on body size, and in children it can also depend on physiological maturation. A model that represents both effects can describe the gradual rise from low pediatric clearance toward a mature level while retaining a clear adult reference.

The bridge is useful only if the parts fit together. In particular, the predicted clearance should approach the adult reference as maturation approaches its plateau. A model may be smooth in age but still fail this requirement if its reference values or size scaling are defined inconsistently. Conversely, joining one pediatric rule to a different adult rule at a chosen age can create an artificial jump in predicted clearance.

Definition: A size-and-maturation clearance model expresses clearance as a function of body size and a maturation factor. A common form for the typical individual is $$ CL(W,a)=CL_{\mathrm{ref}}\left(\frac{W}{W_{\mathrm{ref}}}\right)^\theta F_{\mathrm{mat}}(a), $$ where $W$ is body weight, $W_{\mathrm{ref}}$ is a reference weight, $CL_{\mathrm{ref}}$ is clearance at that weight when maturation is at its plateau, $\theta$ is the allometric exponent, $a$ is the selected developmental-age measure, and $F_{\mathrm{mat}}(a)$ is a dimensionless maturation factor. An individual-level NLME form may multiply this typical value by $\exp(\eta_i)$, where $\eta_i$ is an individual random effect.

For illustration, use a sigmoid maturation function:

F_{\mathrm{mat}}(a)=\frac{a^\gamma}{TM_{50}^{\gamma}+a^\gamma}, \qquad a\geq 0,\quad TM_{50}>0,\quad \gamma>0.

Here $TM_{50}$ is the age at which the maturation factor equals 0.5, and $\gamma$ controls the curve's steepness. The factor lies between 0 and 1 for finite nonnegative age, equals 0 at age zero, and tends to 1 as age increases without bound. Thus the mature plateau is not a second arbitrary parameter: it is specified by the size-scaled reference clearance $CL_{\mathrm{ref}}(W/W_{\mathrm{ref}})^\theta$.

Conditions for a Coherent Bridge

The model has an interpretable adult anchor when the following choices are explicit and compatible:

Reference definition: $CL_{\mathrm{ref}}$ refers to clearance at $W_{\mathrm{ref}}$ when the maturation factor is at its plateau. The reference weight and clearance must use stated units.
Consistent size scaling: The same allometric relationship is applied on the pediatric and adult sides of the bridge unless the data support, and the scientific question requires, a justified alternative.
One age scale: The age variable and its units must be defined consistently for all observations. A maturation function based on age in years cannot be evaluated using age in months without converting the function's parameter units as well.
Supported maturation curve: The pediatric and adult data must contain enough information to estimate or constrain the maturation parameters. Adult observations near a plateau anchor mature clearance, but by themselves may say little about the early-age curve.
Appropriate variability and error assumptions: In an NLME analysis, individual random effects are generally specified with mean zero on the chosen scale and a stated distribution; residual errors are specified conditionally on the structural model. These assumptions should be checked rather than inferred merely from a good-looking prediction curve.

The earlier material on combining allometry and maturation for clearance establishes the rationale for multiplying the size and maturation components. The key additional bridge check is to verify exactly what that product predicts at the mature plateau. With the stated function, $F_{\mathrm{mat}}(a)\to 1$, so

\lim_{a\to\infty}CL(W,a) = CL_{\mathrm{ref}}\left(\frac{W}{W_{\mathrm{ref}}}\right)^\theta.

This equality is the basic alignment condition. At $W=W_{\mathrm{ref}}$, the limiting clearance is $CL_{\mathrm{ref}}$. At another adult weight, the plateau is scaled from that same reference using the allometric relationship. The model therefore does not predict identical clearance for every adult; it predicts a mature size-adjusted relationship.

Why the Plateau Is Continuous

A continuous age function approaches the plateau gradually rather than switching to a new clearance rule at a selected birthday. The limiting result follows directly by dividing numerator and denominator of the maturation function by $a^\gamma$:

F_{\mathrm{mat}}(a) = \frac{1}{(TM_{50}/a)^\gamma+1} \longrightarrow \frac{1}{0+1}=1 \quad\text{as }a\to\infty.

Because the size multiplier does not depend on age in this model, multiplying the limit by that multiplier gives the mature clearance above. This is a limit statement: for any finite age, the factor is still slightly below 1 under this function. In practice, it may be close enough to 1 to be negligible relative to estimation uncertainty, but that should be checked numerically and scientifically rather than assumed.

A piecewise model can instead use one pediatric prediction below an age boundary and an adult prediction above it. If the pediatric rule at boundary $a_c$ gives $CL_{\mathrm{peds}}(W,a_c)$, while the adult rule gives $CL_{\mathrm{adult}}(W)$, continuity requires

CL_{\mathrm{peds}}(W,a_c)=CL_{\mathrm{adult}}(W)

for every relevant weight $W$. If the two sides differ, the prediction jumps at $a_c$. Matching at one weight alone may not be enough if the two rules use different allometric exponents. A continuous maturation model avoids a hard age switch, but it still needs the correct reference definition and size scaling to reach the intended adult plateau.

Worked Example: Verifying the Adult Plateau

Suppose a hypothetical model has $CL_{\mathrm{ref}}=5.00$ L/h at $W_{\mathrm{ref}}=70.0$ kg, $\theta=0.75$, $TM_{50}=1.50$ years, and $\gamma=2.50$. Verify the mature prediction at 70.0 kg and describe the plateau at 35.0 kg.

At the reference weight, the size multiplier is $(70.0/70.0)^{0.75}=1$. As age increases, $F_{\mathrm{mat}}(a)$ tends to 1, so

\lim_{a\to\infty}CL(70.0,a) = 5.00\left(\frac{70.0}{70.0}\right)^{0.75}(1) =5.00\ \mathrm{L/h}.

At 35.0 kg, the mature plateau is instead

5.00\left(\frac{35.0}{70.0}\right)^{0.75} = 5.00(0.5^{0.75}) = 5.00(0.594604) \approx 2.973\ \mathrm{L/h}.

The model therefore connects to the same 5.00 L/h adult reference at 70.0 kg while predicting a lower mature clearance at 35.0 kg through size scaling. The plateau check is not a claim that all adults have clearance of 5.00 L/h.

Calculating How Close a Prediction Is to the Plateau

A practical bridge assessment asks how far predictions at observed older ages remain below the mature size-adjusted value. For the specified sigmoid function, the shortfall at fixed weight is especially simple. If $CL_{\mathrm{plateau}}(W)=CL_{\mathrm{ref}}(W/W_{\mathrm{ref}})^\theta$, then

CL_{\mathrm{plateau}}(W)-CL(W,a) = CL_{\mathrm{plateau}}(W)\{1-F_{\mathrm{mat}}(a)\}.

The relative shortfall is therefore $1-F_{\mathrm{mat}}(a)$. This provides a direct way to judge whether treating a particular older-age prediction as effectively mature is reasonable. The judgment depends on the scientific context, data precision, and intended use; there is no universal age at which the function becomes exactly 1.

Worked Example: Pediatric and Near-Plateau Predictions

Use the hypothetical parameters from the previous example. Calculate clearance for a 17.5 kg child aged 0.500 years, and for a 35.0 kg individual aged 12.0 years. Compare the latter prediction with its mature plateau.

For the 17.5 kg child, the size multiplier is $(17.5/70.0)^{0.75}=0.25^{0.75}=0.353553$. The size-scaled plateau is therefore $5.00(0.353553)=1.76777$ L/h. At age 0.500 years,

F_{\mathrm{mat}}(0.500) = \frac{0.500^{2.50}}{1.50^{2.50}+0.500^{2.50}} = \frac{0.176777}{2.755676+0.176777} \approx 0.0602829.

Thus the predicted clearance is

CL(17.5,0.500) = 1.76777(0.0602829) \approx 0.10657\ \mathrm{L/h}.

For the 35.0 kg individual, the size-scaled plateau is $5.00(0.5^{0.75})=2.97302$ L/h. At age 12.0 years, $12^{2.50}\approx498.831$, giving

F_{\mathrm{mat}}(12) = \frac{498.831}{2.755676+498.831} \approx 0.994506.

The predicted clearance is $2.97302(0.994506)\approx2.95668$ L/h. Its absolute shortfall from the mature plateau is $2.97302-2.95668=0.01634$ L/h, or approximately $0.550\%$ of the plateau. These calculations show both a low-age prediction and an older-age prediction close to, but not exactly at, the plateau.

Assessing the Bridge Against Adult Observations

An adult observation is informative only in relation to its weight, age, and uncertainty. A useful first check is to compare an observed or estimated clearance with the model's typical prediction at that individual's weight and age. If the age is sufficiently mature, the comparison can also be made with the size-adjusted plateau. A single value cannot establish that the bridge is valid; assessment should examine the pattern across subjects and the uncertainty in both the parameters and predictions.

Worked Example: Checking an Adult-Side Prediction

Suppose an 18.0-year-old hypothetical individual weighs 70.0 kg. Under the model above, the typical prediction is $5.00F_{\mathrm{mat}}(18)$ L/h. A hypothetical individual clearance estimate is 5.20 L/h. Calculate the model prediction and the observed value's relative difference from that prediction.

Since $18^{2.50}\approx1374.62$,

F_{\mathrm{mat}}(18) = \frac{1374.62}{2.755676+1374.62} \approx0.997999.

The model prediction is $5.00(0.997999)\approx4.9900$ L/h. The difference is $5.20-4.9900=0.2100$ L/h. Relative to the prediction, this is

\frac{0.2100}{4.9900}\times100\% \approx4.21\%.

The hypothetical observation is about 4.21% above the typical prediction. This does not by itself show a model failure: the individual may differ from the typical subject, and the clearance estimate and model prediction have uncertainty. Across a dataset, however, consistent positive differences in older subjects could indicate a biased plateau, an inadequate size relationship, or a systematic data-source difference.

Diagnostics, Alternatives, and Limitations

Bridge diagnostics should distinguish a wrong plateau from an inadequate maturation curve. Examine individual predictions and residuals against age, weight, and predicted clearance, with pediatric and adult observations identified. Check whether adult residuals are centered around zero and whether pediatric residuals show age-related trends. Also inspect observed age and weight coverage: a model can appear smooth while relying on unsupported extrapolation between small children and adults.

The main alternatives answer different questions. A pediatric-only model may describe the observed pediatric range but leave the adult plateau weakly anchored. An adult-only allometric model can establish mature size scaling but cannot describe developmental change. Separate pediatric and adult models may be convenient when populations or mechanisms differ, but need an explicit boundary and matching condition if predictions are to join continuously. A continuous size-and-maturation model uses one coherent relationship across the span, at the cost of assumptions about the shape of maturation and the applicability of the same allometry.

The approach has important limitations. The sigmoid's $TM_{50}$ and $\gamma$ can be difficult to estimate when few observations cover the ages where maturation changes most. Strong correlation between age and weight can make their effects difficult to separate. Adult data may anchor the plateau without identifying the pediatric curve, while sparse pediatric data may make the adult reference an extrapolation. A good continuity check is necessary, but it is not sufficient evidence that the model is physiologically correct or predictive.

The chosen age measure also matters. If developmental age definitions change across the sample, or if age is recorded with inconsistent units, the maturation parameter loses a coherent interpretation. Physiological evidence and residual diagnostics should be considered alongside numerical fit. A model with a plausible plateau but systematic residual trends is not rescued by its limiting equation.

Common pitfall: It is tempting to declare the bridge continuous because a pediatric curve approaches an adult value in a plot, or because both groups use the same clearance units. Continuity requires equality of the predicted values at the joining point; matching units or visual proximity is not enough. For an asymptotic maturation model, state the mature limit and quantify the finite-age shortfall. For a piecewise model, calculate both predictions at the boundary and verify they are equal across the relevant weights.

Define the adult anchor.
State the reference weight, reference clearance, units, and the meaning of the mature plateau.

Specify the shared prediction rule.
Give the size relationship, age measure, maturation function, and assumptions for individual and residual variability.

Verify the limiting value.
Take the age limit of the maturation function and confirm that the resulting clearance equals the intended size-adjusted adult reference.

Check observed coverage and fit.
Compare predictions with pediatric and adult observations, assess residual patterns, and report uncertainty and extrapolation limits.

Check Your Understanding

Answer each question, then compare with the solutions.

If $F_{\mathrm{mat}}(a)\to1$, what is the mature clearance predicted at weight $W$?
For $CL_{\mathrm{ref}}=4.00$ L/h at $W_{\mathrm{ref}}=50.0$ kg and $\theta=0.75$, calculate the mature plateau at 25.0 kg.
For a piecewise pediatric and adult model, what equality is required for continuity at age $a_c$?
Why does an adult dataset alone generally fail to identify the early-age maturation curve?

Solutions. (1) The mature clearance is $CL_{\mathrm{ref}}(W/W_{\mathrm{ref}})^\theta$, because the maturation factor tends to 1. (2) $4.00(25.0/50.0)^{0.75}=4.00(0.594604)\approx2.378$ L/h. (3) The predictions must satisfy $CL_{\mathrm{peds}}(W,a_c)=CL_{\mathrm{adult}}(W)$ for each relevant weight $W$. (4) Adult observations may characterize mature clearance and size scaling, but if they occur where maturation is already near its plateau, they contain little information about the shape or timing of the early-age rise.