When Tissue Subdivision Changes Predictions

Why Subdivide a Tissue?

A single tissue compartment assigns one concentration to a region of the body. That can be adequate when the region behaves as a coherent unit for the prediction of interest. But two tissues with different partition coefficients or exchange rates can respond differently to the same plasma concentration. Their volume-weighted average may then have a time course that no single tissue compartment reproduces. If tissue exchange is coupled back to plasma, those differences can also alter the predicted plasma profile.

The question is not whether more compartments are always better. It is whether the chosen single-compartment representation preserves the quantities being predicted. The earlier material on exact lumping establishes that aggregation can preserve predictions under specific conditions. Here the focus is on why those conditions may fail when tissue properties differ.

Definition: In the linear exchange model used here, tissue $i$ has volume $V_i$, concentration $C_i(t)$, and exchange conductance $G_i$. The plasma concentration is $C_p(t)$. Its partition coefficient $K_i$ is the equilibrium ratio $C_i/C_p$ when exchange is at equilibrium. The tissue amount changes according to $$ V_i\frac{dC_i}{dt}=G_i\left(C_p-\frac{C_i}{K_i}\right). $$ Here $V_i$ has units L, concentrations have units mg/L, and $G_i$ has units L/h. Thus $dC_i/dt$ has units mg/L/h. $K_i$ is dimensionless.

This model assumes well-mixed plasma and well-mixed tissue compartments; constant volumes, conductances, and partition coefficients; and linear, reversible exchange. There is no tissue metabolism or tissue-specific input in these examples. The equation says that net transfer into a tissue is proportional to the difference between plasma concentration and the tissue concentration scaled by its partition coefficient. If plasma concentration is held constant at $C_p$, exchange tends toward $C_i=K_iC_p$.

A whole-body model can make plasma concentration change as tissue exchange proceeds. For the examples below, plasma volume is $V_p$, elimination is linear with clearance $CL$, and there is no further input after the initial condition. The plasma balance is

V_p\frac{dC_p}{dt}=-CL\,C_p-\sum_i G_i\left(C_p-\frac{C_i}{K_i}\right).

The exchange terms have the opposite sign in the plasma and tissue balances, so exchange alone transfers amount rather than creating or destroying it. Elimination removes amount from plasma at rate $CL\,C_p$.

What the Pooled Concentration Represents

If two tissues are combined, their volume-weighted concentration is

\bar{C}_t(t)=\frac{V_1C_1(t)+V_2C_2(t)}{V_1+V_2}.

This quantity preserves the combined tissue amount: $V_1C_1+V_2C_2=(V_1+V_2)\bar{C}_t$. But preserving amount at each instant does not by itself show that a one-compartment model predicts the same time course. To see the issue, add the two tissue equations:

(V_1+V_2)\frac{d\bar{C}_t}{dt} =(G_1+G_2)C_p-\frac{G_1C_1}{K_1}-\frac{G_2C_2}{K_2}.

When $K_1$ and $K_2$ differ, the final two terms depend on the separate concentrations, not just on their volume-weighted average. A single equation for $\bar{C}_t$ therefore cannot generally be obtained from this sum alone. In a coupled model, the plasma balance also contains those separate tissue concentrations, so differences between the tissues can feed back into $C_p(t)$.

A useful sufficient case for exact pooling is when the tissues have the same $K$, begin at the same concentration, and have the same exchange rate per unit volume, $G_1/V_1=G_2/V_2$. Their equations then imply the same concentration at all times under a shared plasma concentration. The pooled concentration follows the same tissue equation, with total volume $V_1+V_2$ and total conductance $G_1+G_2$. Other special cases may also permit exact aggregation, but one should check the equations rather than assume it from a shared label or an averaged parameter.

Worked Example: Two Tissues Under a Fixed Plasma Concentration

Suppose plasma concentration is held at $1.00$ mg/L, and both tissues initially have concentration $0$ mg/L. Let $V_1=V_2=2.00$ L and $G_1=G_2=1.00$ L/h, but let $K_1=1.00$ and $K_2=4.00$. Find each tissue concentration after $1.00$ h.

Model and conditions. The plasma concentration is an imposed constant, not a prediction from a coupled plasma balance. Volumes, conductances, and partition coefficients are constant, and each tissue follows the linear exchange equation. With constant $C_p$, the solution for a tissue starting at zero is

C_i(t)=K_iC_p\left(1-e^{-G_it/(V_iK_i)}\right).

Tissue 1. Substituting the values gives

C_1(1)=1.00(1.00)\left(1-e^{-1.00(1.00)/(2.00(1.00))}\right) =1-e^{-0.5} \approx 0.3935\ \text{mg/L}.

Tissue 2.

C_2(1)=4.00(1.00)\left(1-e^{-1.00(1.00)/(2.00(4.00))}\right) =4(1-e^{-0.125}) \approx 0.4700\ \text{mg/L}.

The volume-weighted concentration is $(2.00(0.3935)+2.00(0.4700))/4.00\approx0.4318$ mg/L. A single compartment with $V=4.00$ L, $G=2.00$ L/h, and the volume-weighted equilibrium partition coefficient $K=(2.00(1.00)+2.00(4.00))/4.00=2.50$ would instead predict $2.50(1-e^{-2.00/(4.00(2.50))})\approx0.4532$ mg/L. That candidate pooled model matches the weighted equilibrium partitioning but not this transient prediction.

When Pooling Does Preserve the Prediction

A comparison with a case that does pool exactly helps isolate what matters. If both tissues have the same partition coefficient and the same exchange rate per unit volume, they have identical dynamics when they start at the same concentration. In that case their volume-weighted average is not hiding distinct trajectories.

Worked Example: Identical Tissue Dynamics Pool Exactly

Let $V_1=1.00$ L, $V_2=3.00$ L, $K_1=K_2=2.00$, $G_1=0.500$ L/h, and $G_2=1.50$ L/h. Both tissues start at zero concentration, and plasma is maintained at $1.00$ mg/L. Find the pooled concentration at $2.00$ h and verify it using one compartment.

Check the required matching condition. The exchange conductance per unit volume is $G_1/V_1=0.500$ h$^{-1}$ and $G_2/V_2=0.500$ h$^{-1}$. With common $K$ and the same initial concentration, both tissues follow the same equation and therefore have the same concentration at every time.

For either tissue, the exponential rate is $G_i/(V_iK)=0.500/2.00=0.250$ h$^{-1}$. Thus at $2.00$ h,

C_1(2)=C_2(2)=2.00(1.00)\left(1-e^{-0.250(2.00)}\right) =2(1-e^{-0.5})\approx0.7869\ \text{mg/L}.

The pooled compartment has $V=1.00+3.00=4.00$ L and $G=0.500+1.50=2.00$ L/h. Its rate $G/(VK)=2.00/(4.00(2.00))=0.250$ h$^{-1}$, so it gives $2.00(1-e^{-0.250(2.00)})\approx0.7869$ mg/L as well. Here pooling preserves the tissue concentration because subdivision has not introduced a distinct tissue trajectory.

How Subdivision Can Change Plasma Predictions

A difference in tissue concentration matters to the plasma prediction only if the tissue states are coupled back to plasma. The following example compares two models with the same total tissue volume and total conductance, and with the same initial plasma concentration. The tissues differ in partitioning in the subdivided model. Explicit Euler calculations are used for a short numerical comparison; they approximate, rather than exactly solve, the continuous differential equations.

Worked Example: Different Partitioning Changes the Plasma Curve

Take $V_p=3.00$ L, $CL=1.00$ L/h, and initial plasma concentration $C_p(0)=1.00$ mg/L. All tissue concentrations initially equal zero. In the subdivided model, $V_1=V_2=2.00$ L and $G_1=G_2=1.00$ L/h, with $K_1=1.00$ and $K_2=4.00$. Compare this with a pooled model having $V_t=4.00$ L, $G=2.00$ L/h, and $K=2.50$. The pooled $K$ is the volume-weighted equilibrium value.

Check the setup. Both models use positive volumes and conductances, linear elimination, and the same initial plasma and total tissue concentrations. The exchange and elimination terms in the plasma equation have consistent units of amount per time. A fixed Euler step of $0.100$ h is used, and updates use the concentrations at the start of each step. For the subdivided model the updates are

C_p^{\mathrm{new}}=0.900C_p+0.03333C_1+0.008333C_2,\quad C_1^{\mathrm{new}}=0.950C_1+0.05000C_p,\quad C_2^{\mathrm{new}}=0.9875C_2+0.05000C_p.

For example, the first step gives $C_p=0.9000$, $C_1=0.05000$, and $C_2=0.05000$ mg/L. Repeating the updates for five steps gives $C_p(0.500)\approx0.6067$, $C_1(0.500)\approx0.1842$, and $C_2(0.500)\approx0.2001$ mg/L.

For the pooled model, write $C_t$ for its tissue concentration. The equations give the updates $C_p^{\mathrm{new}}=0.900C_p+0.02667C_t$ and $C_t^{\mathrm{new}}=0.05000C_p+0.980C_t$. Starting from $C_p=1.00$ and $C_t=0$, five steps give $C_p(0.500)\approx0.6011$ mg/L and $C_t(0.500)\approx0.1965$ mg/L. The subdivided model's approximate plasma concentration is therefore $0.6067-0.6011=0.0056$ mg/L higher at $0.500$ h. This is a difference between two Euler approximations, not an exact difference between continuous-time solutions.

The initial plasma slopes are the same: each model has zero initial tissue concentrations and total conductance $2.00$ L/h, so $dC_p/dt=-(1.00+2.00)(1.00)/3.00=-1.00$ mg/L/h at time zero. The subsequent curves can diverge because the two tissues return material to plasma at different rates. The example demonstrates that a matched initial slope and an equilibrium-based pooled $K$ do not guarantee a matched later plasma profile.

Deriving the Pooling Condition

The closure issue can be stated directly. Let $V_t=V_1+V_2$, $G_t=G_1+G_2$, and $\bar{C}_t=(V_1C_1+V_2C_2)/V_t$. A single pooled tissue with parameters $K$ and $G_t$ would satisfy

V_t\frac{d\bar{C}_t}{dt}=G_t\left(C_p-\frac{\bar{C}_t}{K}\right).

But adding the true tissue equations gives the expression with separate terms $G_1C_1/K_1+G_2C_2/K_2$. For the pooled equation to hold for all relevant states, those separate terms must be representable using $\bar{C}_t$ with a fixed $K$. That is generally impossible when the two concentrations can vary independently and the partition coefficients differ. Even if the partition coefficients match, the volume-weighted average alone need not determine the conductance-weighted return term unless further conditions link the tissue concentrations. If $C_1=C_2=\bar{C}_t$ and $K_1=K_2=K$, however, the return term is $G_t\bar{C}_t/K$, giving the pooled equation exactly.

This is a sufficient condition, not a claim that it is the only possible special case. For example, particular initial conditions or a restricted input may constrain the states in ways that allow aggregation for a particular prediction. Such a case should be demonstrated from the equations and not inferred just because a weighted average parameter seems plausible.

Diagnostics, Alternatives, and Limitations

A direct warning sign is persistent divergence between tissue concentrations predicted by the subdivided model. Large differences in $K_i$ change equilibrium ratios; differences in $G_i/V_i$ change response speeds. Either can make one concentration a poor stand-in for the other. A second diagnostic is a failure of the aggregated balance to be expressible in terms of $C_p$ and $\bar{C}_t$ alone. If the summed tissue equation still requires $C_1$ and $C_2$ separately, the pooled state is not closed for those assumptions.

The main alternatives are to retain one pooled tissue, to subdivide it into the relevant regions, or to use a simpler effective model for a narrowly defined target. Pooling reduces the number of states and can be appropriate when the tissues have sufficiently similar dynamics for the required prediction. Subdivision represents distinct responses but requires defensible tissue-specific parameters and may increase data requirements. A model can also be adequate for a total tissue amount while inadequate for concentrations in an individual tissue; the prediction target must be specified.

These examples are intentionally limited to linear, well-mixed exchange with constant parameters. Real tissues may have heterogeneous perfusion, binding, metabolism, or time-varying physiology. The equations do not establish that any chosen partition coefficient or conductance is biologically correct. Nor does a difference between model predictions alone show which model is closer to reality; that requires evidence relevant to the intended prediction.

Common mistake: “The tissues have the same total volume and their volume-weighted partition coefficient is known, so one pooled compartment must reproduce their time course.” The weighted coefficient can preserve an equilibrium relationship, but transient predictions also depend on exchange rates and on how concentrations are distributed between tissues. Check whether the summed dynamic equations close using only the proposed pooled state.

Check Your Understanding

Answer each question, then compare your reasoning with the solutions.

Two tissues have different partition coefficients but the same initial concentration. Why does that not ensure that one pooled compartment will reproduce both tissue concentration trajectories?
For a single tissue with $V=2.00$ L, $G=0.800$ L/h, $K=2.00$, constant $C_p=1.00$ mg/L, and initial concentration zero, calculate $C(1.00\text{ h})$.
In the coupled plasma model, what happens to the plasma balance if $C_i=K_iC_p$ for every tissue at an instant?
Two tissues have equal $K$, equal $G_i/V_i$, and equal initial concentrations. State why their volume-weighted concentration follows a pooled tissue equation.
A pooled model and a subdivided model have identical initial plasma slopes. Does that prove their later plasma predictions are identical? Explain.

Solutions. (1) Different $K_i$ values produce different equilibrium ratios $C_i/C_p$; as plasma changes, the separate concentrations generally evolve differently. The summed exchange term then depends on both tissue states, not only on their volume-weighted average. (2) The rate is $G/(VK)=0.800/(2.00(2.00))=0.200$ h$^{-1}$. Therefore $C(1)=2.00(1.00)(1-e^{-0.200(1.00)})\approx0.3625$ mg/L. (3) Each exchange term is zero because $C_i/K_i=C_p$; at that instant plasma changes only through elimination, so $V_p\,dC_p/dt=-CL\,C_p$. (4) Under the shared plasma concentration, the tissues obey identical concentration equations and start at the same value, so their concentrations remain equal. Adding their amount balances then gives a pooled equation with total volume and total conductance. (5) No. Equal initial slopes only establish agreement at that instant. Later tissue concentrations can differ and return material to plasma at different rates, changing subsequent plasma slopes.