Worked Plausibility Audit for Tissue Distribution

Why audit tissue-distribution inputs?

A tissue partition coefficient can look numerically plausible while being produced by an inconsistent collection of inputs. For example, a tissue’s composition fractions might sum to more than one, a binding fraction might use the wrong scale, or a lipid-partition input might not match the units expected by the model. A prediction based on such a set is not repaired simply because its final value falls within a familiar range.

A useful audit follows the inputs into the calculation. Check what each value represents, confirm that the inputs are internally compatible, calculate the partition coefficient, and state how a supported correction changes the result. The calculation below is a simplified equilibrium model for organizing that audit. It illustrates the logic of composition, binding, and phase partitioning; it is not a universal replacement for a tissue-specific mechanistic model.

A simplified composition-based model

Suppose a tissue is represented by aqueous space, neutral lipid, phospholipid, and protein or other binding material. Let \(f_w\), \(f_{nl}\), \(f_{pl}\), and \(f_{pr}\) be their volume fractions on a common tissue-volume basis. These fractions must be nonnegative and, if they are intended to describe the complete tissue, must satisfy

Let \(P_{nl}\), \(P_{pl}\), and \(P_{pr}\) be effective, dimensionless concentration ratios for the three nonaqueous components relative to unbound aqueous concentration in the tissue. In particular, \(P_{pr}\) summarizes linear binding or partitioning into the protein/other component at the modeled conditions. Under a linear additive representation, total tissue concentration is the unbound aqueous tissue concentration multiplied by

The term \(R_t\) is a tissue capacity factor: it accounts for the amount of drug in all represented tissue components per unit unbound aqueous tissue concentration. It is not itself the tissue-to-plasma partition coefficient.

Let \(f_{u,p}\) be the fraction of total plasma drug that is unbound. By definition, unbound plasma concentration is \(f_{u,p}\) times total plasma concentration. Also, \(K_{p,uu}\) relates unbound tissue concentration to unbound plasma concentration. Together, these definitions give

This equation assumes that the effective component ratios are dimensionless and refer to the same modeled conditions; that the component contributions can be added linearly; and that the reported fractions and partition inputs use compatible volume bases. It also assumes that \(f_{u,p}\) and \(K_{p,uu}\) are defined consistently with the total and unbound concentrations in the equation. In passive equilibrium without processes that separate unbound concentrations across tissue and plasma, \(K_{p,uu}=1\) may be a suitable assumption. It should not be imposed automatically when transport, ion trapping, or other mechanisms could produce a different ratio.

Deriving the calculation

The expression follows from writing the tissue and plasma totals in terms of unbound concentrations. Let \(C_{u,t}\) and \(C_{u,p}\) denote unbound tissue and plasma concentrations, and \(C_{t,\mathrm{tot}}\) and \(C_{p,\mathrm{tot}}\) the corresponding total concentrations. Under the additive tissue representation, each component contributes its fraction times its concentration ratio relative to \(C_{u,t}\). Thus the aqueous contribution is \(f_w C_{u,t}\), and, for example, the neutral-lipid contribution is \(f_{nl}P_{nl}C_{u,t}\). Adding all four terms gives

By the definition of \(K_{p,uu}\), \(C_{u,t}=K_{p,uu}C_{u,p}\). By the definition of plasma unbound fraction, \(C_{u,p}=f_{u,p}C_{p,\mathrm{tot}}\). Substitute both relations into the tissue total, then divide by plasma total:

The ratio on the left is \(K_p\), establishing the stated expression. This derivation also shows why a large tissue capacity factor does not necessarily imply a large \(K_p\): the result is scaled by plasma unbound fraction and by \(K_{p,uu}\).

Worked audits

What to inspect, and how approaches differ

An audit is more than an arithmetic check. Confirm that each fraction is a fraction rather than a percentage; that fractions have the same denominator and do not overlap; and that each partition or binding term is dimensionless if it is used as a concentration ratio in this equation. Check that an unbound fraction lies between zero and one, and determine whether the record represents plasma or tissue binding. A value outside its defined range or on a different scale is a reason to investigate, not permission to make an arbitrary replacement.

The simplified composition calculation is useful because it makes the contribution of each input visible. An alternative is to use an empirically measured \(K_p\) directly for the relevant tissue and conditions. That avoids some composition assumptions, but the measurement may not transfer to another species, population, concentration, or physiological state. A more detailed mechanistic model can represent specific binding, ionization, transport, or additional tissue phases. It may be preferable when those mechanisms are important, but it needs suitable parameters and does not eliminate the need to check their definitions and units.

These approaches answer related but not identical questions. A measured \(K_p\) describes a total-concentration ratio in its measurement setting; a composition model attributes a predicted ratio to specified components and assumptions; and \(K_{p,uu}\) concerns unbound concentrations. Do not treat them as interchangeable inputs. If a measured \(K_p\) is used to calibrate a component term, using both the measurement and that calibrated prediction as independent evidence can also count the same information twice.

Diagnostics, failure modes, and limitations

A useful diagnostic is to calculate each term in \(R_t\) separately. This exposes an unexpected dominant contribution, such as a large protein term caused by a fraction entered as \(30\) rather than \(0.30\). Check the computed \(K_p\) by recomputing it from \(K_{p,uu}f_{u,p}R_t\), and verify that \(K_p\), \(K_{p,uu}\), and the effective phase ratios have not been confused. If changing an input that enters additively with a positive fraction does not change \(K_p\) in the expected direction while other inputs are fixed, inspect the equation implementation and the parameter mapping.

The additive expression has important limitations. Real tissues may have more components, nonlinear or saturable binding, pH gradients, active uptake or efflux, and concentration-dependent behavior. Effective coefficients may vary with conditions, and the assumption \(K_{p,uu}=1\) may fail. Fractions that do not sum to one may reflect omitted components or incompatible measurement bases rather than a single mistaken number. In those cases, forcing the inputs into this formula can conceal the actual problem. The audit should state which assumptions are being used and whether the available information supports them.

A common mistake is to see an implausible \(K_p\), adjust the tissue fraction or binding term until the output looks reasonable, and then describe the result as a correction. That reverses the logic: the output does not identify which input is wrong. First establish what each input represents and whether the set is internally consistent. Make a change only when evidence or a clearly stated modeling assumption supports it, then show the resulting direction and magnitude of the prediction change.