Predicting Tissue Partitioning Across Pediatric Ages

Why Tissue Partitioning Can Change With Age

A tissue-to-plasma partition coefficient describes how much drug is present in a tissue relative to plasma at a specified condition. In pediatric PBPK models, it is tempting to assume that this ratio is constant from infancy through childhood. That assumption can fail even when the drug’s intrinsic properties do not change: the proportions of water, lipid, phospholipid-rich membranes, and protein-containing material in a tissue can differ with age. Age-related plasma binding can also change the amount of unbound drug available to partition.

The direction and size of the age effect depend on the drug. A highly lipophilic compound may respond strongly to a change in neutral lipid content. A compound that associates with phospholipids or tissue proteins may respond differently. Plasma binding can reinforce or oppose either effect. The calculations below show how to keep these contributions distinct rather than treating “pediatric partitioning” as a single universal adjustment.

Definition: For tissue \(t\) at age \(a\), the total tissue-to-plasma partition coefficient is \(K_{p,t}(a)=C_t(a)/C_p(a)\), where \(C_t\) is total drug concentration per tissue volume and \(C_p\) is total drug concentration per plasma volume. It is dimensionless when both concentrations use compatible units. It applies to the condition specified, commonly equilibrium or a model’s assumed equilibrium distribution.

A partition coefficient is not a clearance parameter and does not, by itself, describe how quickly drug enters or leaves a tissue. Nor is it the same as an unbound partition coefficient. The value depends on what the model counts in each concentration and on its binding and distribution assumptions. The earlier material on developmental plasma binding and blood-to-plasma ratios provides the context for choosing age-specific plasma unbound fractions; here, plasma binding enters explicitly in the tissue prediction.

A Composition-Weighted Partition Model

A transparent way to derive an age-specific estimate is to represent a tissue as a set of non-overlapping phases. For illustration, use aqueous tissue, neutral lipid, phospholipid-rich material, and protein-rich material. Let \(\phi_j(a)\) be the fraction of tissue volume assigned to phase \(j\) at age \(a\), so the fractions sum to one. Let \(R_j\) be the effective concentration capacity of that phase: the total drug concentration in the phase, divided by the unbound plasma concentration, under the assumed conditions.

The \(R_j\) values summarize how the drug interacts with each phase. They may reflect partition into a lipid phase or reversible association with sites in that phase. For linear, nonsaturable binding at low concentrations, they can be treated as constant across the compared ages. They are not universal constants for all drugs or tissues.

\sum_j \phi_j(a)=1,\qquad C_t(a)=C_{p,u}(a)\sum_j \phi_j(a)R_j

Here \(C_{p,u}(a)\) is unbound plasma concentration. The second expression follows by adding the drug amount in each phase and dividing by total tissue volume: each phase contributes its volume fraction times its concentration. Define the composition-weighted capacity \(D_t(a)=\sum_j\phi_j(a)R_j\). If the unbound plasma fraction is \(f_{u,p}(a)=C_{p,u}(a)/C_p(a)\), division by total plasma concentration gives the prediction.

K_{p,t}(a)=f_{u,p}(a)D_t(a) =f_{u,p}(a)\sum_j\phi_j(a)R_j

This form is useful because it makes two influences visible: tissue composition changes \(D_t\), while plasma binding changes \(f_{u,p}\). The phase capacities must be defined to include the relevant phase-level partitioning and binding. Do not multiply by a separate tissue unbound fraction as well: doing so would count tissue binding twice under this formulation.

Assumptions: This simplified prediction assumes a defined tissue and age, compatible concentration units, equilibrium or an explicitly intended equilibrium approximation, and a representative set of tissue phases. It assumes that drug is not substantially metabolized or actively transported within the tissue for the purpose of estimating partitioning, and that binding is linear over the concentrations considered. It also assumes that the chosen \(\phi_j\), \(R_j\), and \(f_{u,p}\) values are appropriate for the ages being compared. If local pH gradients, active transport, saturable binding, or slow distribution matter, the effective capacities may need to be age- and concentration-dependent or the model must be expanded.

Worked Predictions Across Ages

The numerical values in the following examples are hypothetical. They illustrate the calculation and its assumptions; they are not measured values for a particular tissue, drug, or pediatric population. Each example compares the same tissue, represented by different phase fractions in an infant and an older child.

Worked Example: A Lipophilic Neutral Drug

Suppose an illustrative tissue has the following phase fractions. The infant fractions are 0.78 aqueous, 0.03 neutral lipid, 0.06 phospholipid-rich, and 0.13 protein-rich. The older-child fractions are 0.70, 0.06, 0.08, and 0.16, respectively. Each set sums to 1. For a neutral, lipophilic drug, suppose the effective phase capacities are \(R=(1,80,20,2)\) in that same order. Let \(f_{u,p}\) be 0.08 in the infant and 0.05 in the older child.

For the infant, the composition-weighted capacity is \(D_t=0.78(1)+0.03(80)+0.06(20)+0.13(2)=0.78+2.40+1.20+0.26=4.64\). Thus \(K_{p,t}=0.08(4.64)=0.3712\), or about 0.371.

For the older child, \(D_t=0.70(1)+0.06(80)+0.08(20)+0.16(2)=0.70+4.80+1.60+0.32=7.42\). Thus \(K_{p,t}=0.05(7.42)=0.371\).

In this hypothetical example, the greater lipid and phospholipid fractions increase tissue capacity with age, while the lower unbound plasma fraction reduces the total tissue-to-plasma ratio. The two effects nearly cancel: the predicted coefficients are approximately 0.371 in both age groups. It would be incorrect to conclude from this numerical coincidence that the underlying tissue composition or unbound concentrations are the same.

Worked Example: A Drug With Strong Phospholipid Association

Use the same tissue fractions, but consider a hypothetical basic drug with effective capacities \(R=(1,3,30,12)\), reflecting much greater association with phospholipid-rich and protein-rich material than with neutral lipid. Suppose \(f_{u,p}=0.20\) in the infant and 0.10 in the older child.

For the infant, \(D_t=0.78(1)+0.03(3)+0.06(30)+0.13(12)=0.78+0.09+1.80+1.56=4.23\). Therefore, \(K_{p,t}=0.20(4.23)=0.846\).

For the older child, \(D_t=0.70(1)+0.06(3)+0.08(30)+0.16(12)=0.70+0.18+2.40+1.92=5.20\). Therefore, \(K_{p,t}=0.10(5.20)=0.520\).

The older-child tissue capacity is higher, but the assumed reduction in unbound plasma fraction is larger in relative terms. The predicted total tissue-to-plasma coefficient is consequently lower in the older child. This is a prediction conditional on the stated binding values; it is not a general rule that infants have higher tissue partitioning for basic drugs.

Worked Example: A Relatively Hydrophilic Drug

For a third hypothetical drug, let \(R=(1,1.2,1.5,4)\). Its phase capacities are comparatively close to the aqueous value, except for some association with protein-rich material. Use the same tissue fractions and suppose \(f_{u,p}=0.30\) in the infant and 0.15 in the older child.

For the infant, \(D_t=0.78(1)+0.03(1.2)+0.06(1.5)+0.13(4)=0.78+0.036+0.090+0.52=1.426\). The coefficient is \(K_{p,t}=0.30(1.426)=0.4278\).

For the older child, \(D_t=0.70(1)+0.06(1.2)+0.08(1.5)+0.16(4)=0.70+0.072+0.120+0.64=1.532\). The coefficient is \(K_{p,t}=0.15(1.532)=0.2298\).

The tissue capacity rises modestly, but the assumed unbound plasma fraction halves. The resulting coefficient is lower for the older child. This example emphasizes that even a small or moderate composition effect does not determine the net age trend by itself.

Interpreting and Checking the Prediction

The calculation gives a conditional partition estimate, not a direct observation. Before comparing ages, check that each age-specific phase-fraction set sums to one, that the phases do not overlap, and that each \(R_j\) uses the same concentration reference. Check that \(f_{u,p}\) is a fraction between zero and one and corresponds to the same plasma definition used for \(C_p\). If a model uses whole-blood concentration rather than plasma concentration, the relevant blood-to-plasma relationship must be handled consistently; substituting one denominator for the other changes the coefficient.

A useful diagnostic is to calculate the composition-weighted capacity and plasma-binding contribution separately. If the predicted age effect is surprising, inspect which terms dominate \(D_t\) and whether the change in \(f_{u,p}\) reverses the direction. For example, a large lipid coefficient makes the result sensitive to \(\phi_{\mathrm{lipid}}\); a phospholipid-binding drug may instead be sensitive to \(\phi_{\mathrm{phospholipid}}\). Small errors in a highly influential phase fraction can then matter more than similar errors in a weakly influential phase.

Specify the comparison.
Name the tissue, age groups, concentration basis, and whether the target is an equilibrium or approximate partition coefficient.

Choose composition and binding inputs.
Use age-appropriate phase fractions, phase capacities for the drug, and plasma unbound fractions. Keep units and definitions consistent.

Calculate and inspect contributions.
Compute \(D_t(a)=\sum_j\phi_j(a)R_j\), then \(K_{p,t}(a)=f_{u,p}(a)D_t(a)\). Check the arithmetic and identify influential assumptions.

This approach is more interpretable than assigning the same adult coefficient to every pediatric age: it displays which compositional and binding assumptions create the prediction. A direct empirical age-specific \(K_p\), when reliable measurements are available for the relevant drug and tissue, can avoid some mechanistic assumptions but may be unavailable or imprecise. More detailed composition-based partition models can distinguish additional lipid and binding pools; their added structure is useful only when the needed inputs are credible. A whole-body PBPK simulation can then use tissue-specific partition estimates alongside other model components, but a partition coefficient alone does not establish the accuracy of the full simulation.

Limitations and Common Pitfalls

Tissue composition is tissue-specific: a change in whole-body fat percentage is not a substitute for age-specific fractions in a particular organ. The illustrative \(R_j\) values also need support, because a drug’s ionization, binding affinity, local environment, and concentration can affect effective capacity. A linear model may be inadequate if binding sites saturate. A single equilibrium coefficient may not describe transient distribution when tissue entry is slow, and passive partitioning assumptions may fail when transport or pH gradients are important.

Pitfall: “The tissue has more lipid, so \(K_p\) must rise.” More lipid raises the lipid contribution only when the drug has a sufficiently large lipid-phase capacity. A simultaneous change in plasma binding or other phases can offset it.
Pitfall: adding a tissue unbound fraction to this phase-capacity calculation. The \(R_j\) values already represent phase-level capacity, including the modeled association. Multiplying by an additional tissue-binding correction would double-count that contribution.
Pitfall: treating one calculated coefficient as a universal pediatric value. The result is specific to the tissue, age, drug properties, plasma binding, and assumptions supplied.

Key takeaway: Under the stated phase-capacity assumptions, pediatric tissue-to-plasma partitioning is the product of age-specific plasma unbound fraction and composition-weighted tissue capacity. The net age trend depends on both contributions, so calculate and interpret them separately.

Check Your Understanding

Use the model and assumptions in this tutorial. The solutions are included so you can check the reasoning as well as the arithmetic.

A tissue has phase fractions 0.75 aqueous, 0.05 lipid, and 0.20 protein-rich material. For \(R=(1,10,2)\) and \(f_{u,p}=0.10\), calculate \(K_p\).
If tissue capacity rises from 3 to 4 while \(f_{u,p}\) falls from 0.20 to 0.10, which age has the larger predicted \(K_p\)?
Why must phase fractions sum to one in the model as written?
A calculated tissue-to-plasma coefficient uses total tissue concentration divided by unbound plasma concentration. Is it directly the \(K_p\) defined here? Explain.

Solutions. (1) \(D_t=0.75(1)+0.05(10)+0.20(2)=0.75+0.50+0.40=1.65\), so \(K_p=0.10(1.65)=0.165\). (2) The first prediction is \(0.20(3)=0.60\); the second is \(0.10(4)=0.40\). The first age has the larger predicted coefficient. (3) The fractions represent non-overlapping portions of the total tissue volume; they must account for that volume exactly once. (4) No. It is \(C_t/C_{p,u}\), a capacity relative to unbound plasma. To obtain the total tissue-to-plasma coefficient, multiply by \(f_{u,p}=C_{p,u}/C_p\).