Modeling Maturation of Clearance

From Maturation to a Clearance Prediction

A maturation curve describes how a physiological function changes with age. To predict clearance, the curve must be linked to a clearance value: the model treats clearance at a given age as a fraction of the value expected after maturation is complete. This gives a direct way to calculate pediatric clearance predictions at ages between birth and maturity.

The construction here is a maturation component on its own. It does not adjust for differences in body size. Accordingly, the mature clearance value must be understood on a consistent reference basis, and predictions should not be interpreted as accounting for size-related clearance differences. The age variable must also be specified consistently; in the examples, it is postmenstrual age (PMA), measured in weeks, or an explicitly stated age in months.

Definition: Let $A$ be age, $CL_{\mathrm{mat}}$ the mature clearance value, $TM_{50}$ the age at which half of maturation is reached, and $h$ the Hill coefficient. The age-dependent mature fraction and predicted clearance are $$ F_{\mathrm{mat}}(A)=\frac{A^h}{A^h+TM_{50}^h}, \qquad CL(A)=CL_{\mathrm{mat}}F_{\mathrm{mat}}(A). $$ Here $A$ and $TM_{50}$ must use the same age units, $h>0$, $TM_{50}>0$, and $CL_{\mathrm{mat}}>0$. The fraction is dimensionless, so $CL(A)$ has the same units as $CL_{\mathrm{mat}}$.

This is the sigmoid Emax maturation form with a zero baseline fraction and a maximum fraction of one. Its parameters have distinct roles: $CL_{\mathrm{mat}}$ sets the mature clearance level; $TM_{50}$ sets the age location of the curve; and $h$ controls how gradually or sharply the fraction rises around that age. As in the earlier material on Hill and half-maturation parameters, $h$ is dimensionless and is not itself an age.

Why the Product Gives Clearance

The model defines the maturation fraction as the proportion of mature clearance attained at age $A$. If the mature value is $CL_{\mathrm{mat}}$, then a fraction $F_{\mathrm{mat}}(A)$ of that value is $CL_{\mathrm{mat}}F_{\mathrm{mat}}(A)$. This multiplication is a modeling assumption: it says that the age-dependent change acts proportionally on the mature clearance value.

The key properties follow directly from the formula. At $A=TM_{50}$, the numerator and the two terms in the denominator are equal, so the fraction is one half. For positive $A$, increasing $A$ increases $A^h$, and the fraction rises toward one. At age zero, the fraction is zero for $h>0$; as age becomes very large relative to $TM_{50}$, the fraction approaches one.

F_{\mathrm{mat}}(TM_{50}) = \frac{TM_{50}^h}{TM_{50}^h+TM_{50}^h} = \frac{1}{2}, \qquad \lim_{A\to\infty}F_{\mathrm{mat}}(A)=1.

Thus, $CL(TM_{50})=CL_{\mathrm{mat}}/2$, and $CL(A)$ approaches $CL_{\mathrm{mat}}$ as age increases. The mature clearance is an asymptote, not necessarily the prediction at the oldest age observed in a pediatric data set. The curve also imposes a zero fraction at age zero; whether that behavior is plausible depends on the age definition and the physiological setting.

Specify age and units.
Choose the age metric appropriate to the setting and express $A$ and $TM_{50}$ in the same units.

Set the mature value and curve parameters.
State $CL_{\mathrm{mat}}$, $TM_{50}$, and $h$, with units for the clearance and age parameters.

Calculate the mature fraction.
Substitute age and curve parameters into $F_{\mathrm{mat}}(A)$, retaining adequate precision during calculation.

Calculate clearance and check its scale.
Multiply the unrounded fraction by $CL_{\mathrm{mat}}$. For the stated assumptions, the prediction lies between zero and the mature value.

Worked Predictions Across Pediatric Ages

Worked Example: Clearance at Three Postmenstrual Ages

Suppose a hypothetical model has $CL_{\mathrm{mat}}=8.00\ \mathrm{L/h}$, $TM_{50}=40$ weeks PMA, and $h=3$. Calculate predicted clearance at 20, 40, and 80 weeks PMA. The assumptions are that the selected age measure is appropriate, all ages and $TM_{50}$ are in weeks, and the same mature clearance reference applies at each age.

At 20 weeks PMA, $$ F_{\mathrm{mat}}(20)=\frac{20^3}{20^3+40^3} =\frac{8{,}000}{72{,}000} =\frac{1}{9}\approx 0.1111. $$ Using the unrounded fraction for the clearance calculation, $$ CL(20)=8.00\left(\frac{20^3}{20^3+40^3}\right) \approx 0.8889\ \mathrm{L/h}. $$

At 40 weeks PMA, age equals $TM_{50}$, so the fraction is $0.5000$, and $$ CL(40)=8.00\left(\frac{40^3}{40^3+40^3}\right) =8.00(0.5000)=4.00\ \mathrm{L/h}. $$

At 80 weeks PMA, $$ F_{\mathrm{mat}}(80)=\frac{80^3}{80^3+40^3} =\frac{512{,}000}{576{,}000} =\frac{8}{9}\approx 0.8889, $$ and therefore $$ CL(80)=8.00\left(\frac{80^3}{80^3+40^3}\right) \approx 7.111\ \mathrm{L/h}. $$

The predictions increase with age and remain below $8.00\ \mathrm{L/h}$, as expected for this model. The values $0.1111$ and $0.8889$ are rounded displays of the fractions; the clearance calculations use their exact fractional forms so that rounding does not change the reported results.

Worked Example: Using a Different Age Unit Consistently

Consider a second hypothetical model with $CL_{\mathrm{mat}}=1.50\ \mathrm{L/h}$, $TM_{50}=6$ months, and $h=2$. Calculate clearance at ages 3, 6, and 12 months. The inputs use months consistently, so no conversion is needed.

At 3 months, $$ F_{\mathrm{mat}}(3)=\frac{3^2}{3^2+6^2} =\frac{9}{45}=0.2000, \qquad CL(3)=1.50(0.2000)=0.300\ \mathrm{L/h}. $$

At 6 months, the age is $TM_{50}$, so $$ F_{\mathrm{mat}}(6)=\frac{36}{36+36}=0.5000, \qquad CL(6)=1.50(0.5000)=0.750\ \mathrm{L/h}. $$

At 12 months, $$ F_{\mathrm{mat}}(12)=\frac{144}{144+36}=0.8000, \qquad CL(12)=1.50(0.8000)=1.20\ \mathrm{L/h}. $$

These calculations illustrate why matching age units matters. If age were entered in weeks, $TM_{50}$ would also have to be converted to weeks; changing only one would alter the ratio in the model and produce an incorrect prediction.

Finding the Age for a Target Mature Fraction

The model can also be rearranged to find the age at which a chosen fraction of mature clearance is reached. Let the target fraction be $q$, with $0<q<1$. Starting from $q=A^h/(A^h+TM_{50}^h)$, multiply both sides by the denominator, collect the terms involving $A^h$, and take the positive $h$th root:

q(A^h+TM_{50}^h)=A^h \quad\Longrightarrow\quad qTM_{50}^h=(1-q)A^h \quad\Longrightarrow\quad A=TM_{50}\left(\frac{q}{1-q}\right)^{1/h}.

Worked Example: Age at 80% of Mature Clearance

For the hypothetical model with $CL_{\mathrm{mat}}=8.00\ \mathrm{L/h}$, $TM_{50}=40$ weeks PMA, and $h=3$, find the age at which predicted clearance is 80% of the mature value. Set $q=0.8000$ in the rearranged equation: $$ A=40\left(\frac{0.8000}{1-0.8000}\right)^{1/3} =40(4)^{1/3} \approx 40(1.5874) \approx 63.50\ \text{weeks PMA}. $$

The corresponding clearance is $$ CL(A)=8.00(0.8000)=6.40\ \mathrm{L/h}. $$ This age is greater than $TM_{50}$, which is consistent with reaching a fraction greater than one half. The result is a model-based age prediction, not a guarantee that an individual child reaches that exact fraction at that age.

Assumptions, Diagnostics, and Alternatives

The model is most defensible when the clearance process represented by $CL_{\mathrm{mat}}$ changes smoothly and monotonically with the selected age measure, and when its mature level is meaningful for the reference context. It also assumes that the age effect can be represented as a proportion of that mature level. A maturation curve alone does not account for body-size differences, changes in other covariates, or individual variation unless those features are handled elsewhere in the model.

When assessing whether this structure is useful, compare predictions with the age pattern in the data and consider whether there are enough observations across the range where the curve rises. Warning signs include systematic overprediction or underprediction in an age range, an apparent plateau that conflicts with the model's implied asymptote, or parameter estimates that change substantially when a small number of observations are removed. Sparse data near birth or near the mature range may leave those parts of the curve weakly informed. A numerical prediction can still be calculated there, but precision in the arithmetic does not establish reliability of the extrapolation.

A constant mature clearance applied to all ages is simpler, but cannot represent an age-related increase in clearance. Treating age groups separately can describe group-specific differences without imposing a smooth curve, but does not provide the same continuous prediction between groups and can require many group-specific estimates. The sigmoid maturation model makes a more structured assumption: a smooth rise toward a mature value. That structure is useful when biologically plausible, but should not be mistaken for a pattern proven solely by fitting the equation.

A practical check is to calculate the implied fraction at $A=TM_{50}$, verify that $CL(A)\leq CL_{\mathrm{mat}}$ for positive ages, and inspect predictions at the youngest and oldest ages relevant to use. If the data support a nonzero fraction at the youngest modeled age, or a different plateau, the stated zero-baseline, unit-maximum form may not describe the process adequately. Such a mismatch calls for reconsidering the model assumptions rather than forcing the parameters to give an implausible curve.

Common Mistakes and Model Limits

Mixing age units. Using age in days with $TM_{50}$ in weeks changes the fraction. Convert one quantity so both use the same units before calculating.
Treating $TM_{50}$ as a mature age. It is the age at which the model predicts half of the mature fraction, not the age at which maturation is complete.
Reporting a rounded intermediate as exact. Keep the fraction unrounded when multiplying by $CL_{\mathrm{mat}}$, then round the final clearance to an appropriate number of significant figures.
Assuming the curve accounts for body size. It does not. The predictions describe the specified maturation component on the chosen mature-clearance reference basis.
Interpreting the mature value as an observed maximum. $CL_{\mathrm{mat}}$ is the model's asymptotic value; the data may not include ages sufficiently high to observe that level directly.

Key takeaway: Calculate pediatric clearance by multiplying the mature clearance value by the age-specific mature fraction. Check age units, parameter assumptions, and the predicted curve before interpreting the resulting values.

Check Your Understanding

Answer each question using the model definition above. Brief solutions follow the questions.

A model has $CL_{\mathrm{mat}}=5.00\ \mathrm{L/h}$, $TM_{50}=10$ months, and $h=2$. What clearance does it predict at 10 months?
For the same model, calculate the mature fraction and clearance at 5 months.
If age is recorded in weeks, what must be done to $TM_{50}$ before using the model?
For $h=2$, $TM_{50}=10$ months, and target fraction $q=0.8000$, find the age at which the model reaches that fraction.
State one reason a maturation-only prediction might not adequately describe clearance differences among children of the same age.

Solutions. (1) At $A=TM_{50}$, the fraction is $0.5000$, so clearance is $5.00(0.5000)=2.50\ \mathrm{L/h}$. (2) At 5 months, $F_{\mathrm{mat}}(5)=5^2/(5^2+10^2)=25/125=0.2000$; clearance is $5.00(0.2000)=1.00\ \mathrm{L/h}$. (3) Convert $TM_{50}$ from months to weeks so age and $TM_{50}$ use matching units. (4) $A=10(0.8000/0.2000)^{1/2}=10(2)=20$ months. (5) Children of the same age may differ in body size, which this maturation-only model does not represent.