Poulin-Theil Kp Prediction for Neutral Compounds

Purpose and scope

A tissue-to-plasma partition coefficient, $K_p$, connects a tissue concentration to a plasma concentration at equilibrium. For a neutral compound, the Poulin–Theil method predicts this ratio by representing tissue as aqueous, neutral-lipid, and phospholipid phases. The compound’s octanol–water partitioning informs its affinity for lipid phases, while its plasma unbound fraction links unbound concentration to total plasma concentration.

The calculation is mechanistic in form, but its result is still a model prediction. It depends on how well the selected tissue composition represents the tissue, and on whether the compound behaves as a passively partitioning neutral solute. The earlier material on neutral-drug partitioning across tissue phases establishes the phase-partitioning idea; here that idea is assembled into a practical $K_p$ prediction.

Definition: For this tutorial, $K_p$ is the equilibrium ratio of total concentration in a specified tissue to total concentration in plasma, with both concentrations expressed on a compatible volume basis. The neutral-compound Poulin–Theil calculation is $$ K_p = f_{u,p}\left(f_w + f_{nl}K_{nl,w} + f_{ph}K_{ph,w}\right), $$ where $f_{u,p}$ is the plasma unbound fraction; $f_w$, $f_{nl}$, and $f_{ph}$ are the tissue volume fractions assigned to water, neutral lipid, and phospholipid; and $K_{nl,w}$ and $K_{ph,w}$ are the phase-to-water partition coefficients for neutral lipid and phospholipid. In the standard neutral-compound approximation used here, $K_{nl,w}=K_{ow}$ and $K_{ph,w}=0.3K_{ow}+0.7$, where $K_{ow}$ is the octanol–water partition coefficient.

The input $K_{ow}$ is dimensionless and is obtained from the neutral compound’s $\log P$:

K_{ow}=10^{\log P}.

The tissue fractions are proportions of tissue volume, not masses or percentages. Use fractions such as $0.70$, rather than $70$. They need not sum to one if the composition data include a residual fraction; in the stated model, that residual is treated as having no additional partitioning capacity beyond the modelled aqueous contribution. Do not count the same volume in two phases.

Why plasma unbound fraction enters the calculation

The tissue-phase terms describe storage relative to the concentration of unbound compound in plasma. For a neutral compound at passive equilibrium, the unbound concentration in tissue water is taken to equal the unbound plasma concentration. Tissue lipid phases can hold additional compound according to their partition coefficients. But $K_p$ compares total tissue concentration with total plasma concentration, and total plasma concentration exceeds unbound plasma concentration when plasma binding is present.

Since $f_{u,p}=C_{u,p}/C_{p}$, the total plasma concentration is $C_p=C_{u,p}/f_{u,p}$. If the tissue-phase calculation gives $C_t/C_{u,p}=A_t$, then

K_p=\frac{C_t}{C_p} =\frac{A_tC_{u,p}}{C_{u,p}/f_{u,p}} =f_{u,p}A_t.

This explains the direction of the plasma-binding adjustment: with tissue composition and partitioning held fixed, a smaller $f_{u,p}$ gives a smaller predicted ratio of total tissue concentration to total plasma concentration. This derivation assumes that the same unbound plasma concentration is the relevant reference for exchange with tissue water and that binding is not saturable over the concentration range of interest.

Worked calculations

Worked Example: Prediction for a Moderately Lipophilic Neutral Compound

Consider a hypothetical neutral compound with $\log P=1.00$ and $f_{u,p}=0.250$. For a hypothetical tissue, let $f_w=0.700$, $f_{nl}=0.040$, and $f_{ph}=0.060$. The remaining $0.200$ of tissue volume is residual material not assigned additional storage in this calculation. Calculate $K_p$.

First convert $\log P$ to $K_{ow}$:

K_{ow}=10^{1.00}=10.0.

Then calculate the phospholipid-to-water partition coefficient:

K_{ph,w}=0.3(10.0)+0.7=3.70.

Substitute the tissue fractions and plasma unbound fraction:

K_p=0.250\left[0.700+0.040(10.0)+0.060(3.70)\right] =0.250(1.322)=0.3305.

The predicted $K_p$ is $0.331$ (dimensionless, on the stated concentration basis). Under the model assumptions, the equilibrium total concentration in this tissue is about $0.331$ times the total plasma concentration.

Worked Example: A Highly Lipophilic Compound in a Lipid-Poor Tissue

For a second hypothetical neutral compound, let $\log P=2.00$ and $f_{u,p}=0.100$. A hypothetical tissue has $f_w=0.750$, $f_{nl}=0.020$, and $f_{ph}=0.040$. Calculate $K_p$, showing how the lipid terms contribute.

Here $K_{ow}=10^{2.00}=100$, so $K_{ph,w}=0.3(100)+0.7=30.7$. The phase-weighted tissue capacity is

$$ A_t=0.750+0.020(100)+0.040(30.7) =0.750+2.000+1.228=3.978. $$

Applying the plasma unbound fraction gives

$$ K_p=0.100(3.978)=0.3978. $$

Thus the prediction is $K_p=0.398$. Although the compound is highly lipophilic, this tissue has relatively small lipid fractions and the plasma unbound fraction is $0.100$. A large $K_{ow}$ alone does not determine a large total tissue-to-plasma ratio.

Worked Example: Separating Lipophilicity and Plasma-Binding Effects

For a third hypothetical neutral compound, let $\log P=0.500$, $f_{u,p}=0.800$, and tissue fractions $f_w=0.650$, $f_{nl}=0.050$, and $f_{ph}=0.080$. Calculate the prediction and compare it with the same tissue and compound properties if $f_{u,p}$ were $0.400$.

The octanol–water partition coefficient is $K_{ow}=10^{0.500}=3.162$. The phospholipid coefficient is $K_{ph,w}=0.3(3.162)+0.7=1.649$. Therefore,

$$ A_t=0.650+0.050(3.162)+0.080(1.649) =0.650+0.1581+0.1319=0.9400. $$

With $f_{u,p}=0.800$, $K_p=0.800(0.9400)=0.752$. If only $f_{u,p}$ changes to $0.400$, then $K_p=0.400(0.9400)=0.376$. Halving $f_{u,p}$ halves the predicted $K_p$ because it is a multiplicative factor in this equation. This is a model-based comparison, not evidence that changing plasma binding alone would cause a real tissue concentration change in every setting.

Assumptions and checks

Confirm neutrality.
The calculation treats the compound as neutral under the relevant conditions. If appreciable ionization occurs, a neutral-only partition calculation does not represent the full distribution process.

Use compatible inputs.
Use a neutral-compound $\log P$, a plasma unbound fraction on the relevant basis, and tissue fractions referring to the tissue of interest. Check that fractions are entered as proportions and that phase volumes are not duplicated.

Evaluate the partitioning assumptions.
The model assumes passive equilibrium, linear and nonsaturable partitioning, and that octanol–water partitioning is a useful proxy for neutral-lipid affinity. Its phospholipid approximation is also an empirical simplification.

Inspect the result.
Check each term in the bracket and verify that $K_p$ is nonnegative. A prediction dominated by a small lipid fraction multiplied by a very large $K_{ow}$ can be particularly sensitive to errors in $\log P$ or composition.

Useful boundary checks catch arithmetic or input errors. At $K_{ow}=1$, the model gives $K_{ph,w}=1$, so the tissue capacity reduces to the sum of the modelled water and lipid fractions. If $f_{u,p}$ is reduced while all other inputs remain unchanged, the predicted $K_p$ must also decrease proportionally. If either check fails, revisit the conversion from $\log P$, the bracketed sum, and the placement of $f_{u,p}$.

Interpretation, alternatives, and limitations

The Poulin–Theil prediction is useful when a neutral compound’s tissue composition, $\log P$, and plasma unbound fraction are available but a measured tissue partition coefficient is not. Its phase terms make the prediction interpretable: water contributes directly, while neutral lipid and phospholipid contributions are weighted by their tissue fractions and their affinity for the compound. This transparency also exposes which inputs are driving the result.

A direct measurement of tissue and plasma concentrations at equilibrium can provide an empirical $K_p$ and avoid relying on the model’s phase proxies, although measurement quality and experimental conditions still matter. Composition-based predictions are useful when measurements are unavailable or for comparing hypotheses, but other mechanistic methods make different assumptions about tissue binding and partitioning. They are not interchangeable without checking the compound class and required inputs. In particular, this neutral-only formulation should not be treated as a complete prediction method for ionizable compounds.

The equation simplifies biological tissue. It does not explicitly represent active uptake or efflux, saturable binding, transporter-mediated distribution, tissue-specific pH effects, or all forms of macromolecular binding. It also relies on representative composition fractions and on a correlation between octanol and tissue phases. Errors in $\log P$ matter increasingly at high lipophilicity because $K_{ow}=10^{\log P}$: a change of one log unit multiplies $K_{ow}$ by ten. Consequently, a plausible numeric answer is not, by itself, evidence that the assumptions are satisfied.

Common mistake

Common mistake: “The plasma unbound fraction belongs in the denominator, so stronger plasma binding must increase $K_p$.” That reasoning confuses the tissue capacity relative to unbound plasma concentration with the ratio to total plasma concentration. Since $C_p=C_{u,p}/f_{u,p}$, the conversion is $K_p=f_{u,p}A_t$, not $A_t/f_{u,p}$. For fixed tissue composition and partition coefficients, stronger plasma binding (smaller $f_{u,p}$) lowers the predicted total-tissue-to-total-plasma ratio in this model.

Check Your Understanding

Answer each question using the neutral-compound equation and its stated assumptions. Full solutions follow.

A compound has $\log P=0$. What are $K_{ow}$ and $K_{ph,w}$?
For fixed tissue fractions and $K_{ow}$, what happens to predicted $K_p$ if $f_{u,p}$ is reduced by one half?
Why is a neutral-only Poulin–Theil calculation inappropriate when a substantial fraction of a drug is ionized?
For a hypothetical compound, $K_{ow}=10$, $f_{u,p}=0.200$, $f_w=0.700$, $f_{nl}=0.030$, and $f_{ph}=0.050$. Calculate $K_p$.

Solutions. (1) $K_{ow}=10^0=1$, and $K_{ph,w}=0.3(1)+0.7=1$. (2) $K_p$ is directly proportional to $f_{u,p}$, so it is also halved. (3) The equation represents neutral-compound partitioning and does not account for the distribution of ionized species; its assumptions no longer describe the relevant partitioning. (4) $K_{ph,w}=0.3(10)+0.7=3.70$. Thus $K_p=0.200[0.700+0.030(10)+0.050(3.70)]=0.200(1.185)=0.237$.