When Equal Inputs Give Different Clearance Predictions

Why Equal Inputs Can Produce Unequal Predictions

A hepatic clearance prediction is not determined by flow and drug properties alone: it also depends on how the model represents blood movement and mixing within the liver. The well-stirred model assumes rapid mixing and a common hepatic concentration; the parallel-tube model represents blood as moving through separate paths with concentration falling along each path; the dispersion model allows axial mixing along the flow direction. Thus, when hepatic blood flow, unbound fraction, and intrinsic clearance are identical, different assumptions about the concentration profile can still give different clearances.

The comparison is useful only if the inputs have the same definitions and units. Here, hepatic blood flow is \(Q\), unbound fraction in blood is \(f_u\), and intrinsic clearance is \(CL_{int}\), defined relative to unbound drug concentration. The product \(f_uCL_{int}\) is the effective intrinsic clearance in the liver model. All clearance quantities in the examples are expressed in L/h.

Definition: Define the dimensionless ratio \(R=f_uCL_{int}/Q\). For each model, the hepatic extraction ratio is \(E_H=CL_H/Q\), where \(CL_H\) is hepatic clearance. The well-stirred model gives \(E_H=R/(1+R)\); the parallel-tube model gives \(E_H=1-e^{-R}\). In the axial-dispersion model, \(E_H=1-C_{out}/C_{in}\), where \(C_{in}\) and \(C_{out}\) are inlet and outlet concentrations, and the concentration ratio depends on both \(R\) and the dispersion number \(\mathrm{DN}\).

Comparing models at fixed \(Q\), \(f_u\), and \(CL_{int}\) fixes \(R\) as well. Consequently, differences in their predicted extraction ratios are attributable to model structure, not to different input values. The dispersion model has one additional structural parameter, \(\mathrm{DN}\), which describes axial dispersion relative to flow. It is not an alternative name for intrinsic clearance.

Equations and Assumptions

The well-stirred model treats the liver as a single, rapidly mixed compartment. Its central concentration is represented by one value, and the drug is eliminated from that well-mixed pool. Under this assumption, the clearance is

CL_{H,\mathrm{WS}}=\frac{Q f_uCL_{int}}{Q+f_uCL_{int}} =Q\frac{R}{1+R}.

The parallel-tube model instead assumes parallel flow paths with no axial mixing between paths. Within a path, drug concentration falls as blood travels through the liver. With the model’s uniform path properties and total flow \(Q\), the fraction remaining after passage is \(e^{-R}\), so extraction is \(1-e^{-R}\):

CL_{H,\mathrm{PT}}=Q(1-e^{-R}).

The dispersion model describes flow with axial dispersion: concentration can spread along the direction of flow, rather than being either perfectly mixed throughout the liver or confined to unmixed parallel paths. For the standard axial-dispersion formulation, let

a=\sqrt{1+4\,\mathrm{DN}\,R}, \qquad \frac{C_{out}}{C_{in}} = \frac{4a\exp\!\left(\frac{1-a}{2\,\mathrm{DN}}\right)} {(1+a)^2-(1-a)^2\exp\!\left(-\frac{a}{\mathrm{DN}}\right)}.

Then \(CL_{H,\mathrm{disp}}=Q(1-C_{out}/C_{in})\). This expression assumes positive \(Q\), \(f_u\), and \(CL_{int}\), and a specified finite positive \(\mathrm{DN}\). Its limiting structures are also informative: as dispersion tends to zero, it approaches the parallel-tube result; as dispersion becomes very large, it approaches the well-stirred result. These limits express how the model’s mixing assumption changes, not changes to the drug or the liver inputs.

Why the Models Diverge

The well-stirred and parallel-tube expressions can be compared directly. For \(R>0\), \(e^R>1+R\), so \(e^{-R}<1/(1+R)\). Subtracting each remaining fraction from one gives

1-e^{-R}>\frac{R}{1+R}.

Thus, at the same positive \(R\), the parallel-tube model predicts greater extraction and clearance than the well-stirred model. The concentration decline along parallel flow paths produces more extraction than the well-mixed representation in this comparison. For the standard finite-dispersion model, predictions lie between these limiting structures: increasing axial mixing moves the prediction toward the well-stirred limit, while reducing it moves toward the parallel-tube limit.

At low extraction, the models nearly agree because the leading term in each extraction ratio is \(R\). Their difference appears in higher-order terms. Using the Taylor expansion of the exponential around \(R=0\),

E_{H,\mathrm{WS}} =R-R^2+R^3+O(R^4), \qquad E_{H,\mathrm{PT}} =R-\frac{R^2}{2}+\frac{R^3}{6}+O(R^4).

The first-order terms match, while the leading difference is \(R^2/2\). This explains why modest absolute model differences are expected at sufficiently low \(R\), but it does not imply exact equality. At high \(R\), both extraction ratios approach one and both clearances approach \(Q\). The most noticeable disagreement is therefore often at intermediate extraction, where neither limit makes the structural differences negligible.

Worked Example: Three Predictions at Intermediate Extraction

Suppose, hypothetically, that \(Q=90.0\ \mathrm{L/h}\), \(f_u=0.100\), and \(CL_{int}=900\ \mathrm{L/h}\). Assume the values are all on a compatible blood basis and \(CL_{int}\) is referenced to unbound concentration. For the dispersion prediction, take \(\mathrm{DN}=0.200\). The shared dimensionless ratio is

R=\frac{(0.100)(900\ \mathrm{L/h})}{90.0\ \mathrm{L/h}}=1.00.

The well-stirred prediction is

CL_{H,\mathrm{WS}} =90.0\frac{1.00}{1+1.00} =45.0\ \mathrm{L/h}.

For the parallel-tube prediction, \(e^{-1}=0.367879\), giving

CL_{H,\mathrm{PT}} =90.0(1-e^{-1}) =90.0(0.632121) =56.8909\ \mathrm{L/h}.

For the dispersion model, \(a=\sqrt{1+4(0.200)(1.00)}=\sqrt{1.8}=1.34164\). Substitution into the concentration-ratio expression gives \(C_{out}/C_{in}\approx0.41661\), and hence

CL_{H,\mathrm{disp}} =90.0(1-0.41661) \approx52.505\ \mathrm{L/h}.

The three results are approximately 45.0, 52.505, and 56.8909 L/h. The dispersion prediction falls between the two limiting structures for this positive finite dispersion number. The inputs \(Q\), \(f_u\), and \(CL_{int}\) did not change; the differences arise from how each model represents intrahepatic concentration and mixing.

How the Difference Changes Across Extraction

The following comparisons use the well-stirred and parallel-tube equations. They show two limiting patterns without changing the basis of the inputs: \(R\) is varied while \(Q\) is held fixed. The values are model predictions, not measurements.

Worked Example: Low and High Values of the Shared Ratio

Let the hypothetical hepatic flow be \(Q=100\ \mathrm{L/h}\). First take \(R=0.100\), representing low extraction. Then

CL_{H,\mathrm{WS}} =100\frac{0.100}{1.100} =9.09091\ \mathrm{L/h}, \qquad CL_{H,\mathrm{PT}} =100(1-e^{-0.100}) =100(0.0951626) =9.51626\ \mathrm{L/h}.

The parallel-tube prediction is higher by \(9.51626-9.09091=0.42535\ \mathrm{L/h}\). Both are close because their extraction ratios share the same first-order term \(R\).

Now take \(R=10.0\), representing high extraction, still with \(Q=100\ \mathrm{L/h}\). The predictions are

CL_{H,\mathrm{WS}} =100\frac{10.0}{11.0} =90.9091\ \mathrm{L/h}, \qquad CL_{H,\mathrm{PT}} =100(1-e^{-10}) =100(0.9999546) =99.9955\ \mathrm{L/h}.

Both approach the flow ceiling of \(100\ \mathrm{L/h}\), although their difference here is \(99.9955-90.9091=9.0864\ \mathrm{L/h}\). “Approach the same limit” does not mean that predictions are identical at every finite \(R\); it means both tend to \(Q\) as \(R\) grows without bound.

Worked Example: Converting Shared Inputs to Model-Specific Predictions

For a second hypothetical comparison, let \(Q=80.0\ \mathrm{L/h}\), \(f_u=0.200\), and \(CL_{int}=200\ \mathrm{L/h}\). These values give \(R=(0.200)(200)/80.0=0.500\). The well-stirred prediction is

CL_{H,\mathrm{WS}} =80.0\frac{0.500}{1+0.500} =26.6667\ \mathrm{L/h}.

For the parallel-tube prediction, \(e^{-0.500}=0.606531\), so

CL_{H,\mathrm{PT}} =80.0(1-0.606531) =31.4775\ \mathrm{L/h}.

The difference is \(31.4775-26.6667=4.8108\ \mathrm{L/h}\). The matching \(R\) confirms that the comparison uses the same effective intrinsic clearance relative to flow in both formulations. It does not make the clearance equations interchangeable: the predicted values differ because the models impose different mixing structures.

Diagnostics, Alternatives, and Limitations

Before interpreting a model difference as a consequence of mixing structure, check that the comparison is genuinely like-for-like. Flow must use the same units as clearance, \(f_u\) must refer to the relevant blood concentration basis, and \(CL_{int}\) must have the same unbound-reference definition in each calculation. A value of \(f_uCL_{int}/Q\) with units remaining is a warning that the inputs or units have been combined incorrectly. Also verify that each calculated clearance is nonnegative and does not exceed \(Q\) under these model formulations.

A practical comparison should report the shared inputs, \(R\), and the model structure used. For the dispersion model, report \(\mathrm{DN}\) as well. If the dispersion result does not approach the parallel-tube prediction as \(\mathrm{DN}\) tends toward zero, or does not approach the well-stirred prediction as \(\mathrm{DN}\) becomes large, check the parameterization and implementation. At extreme values, direct evaluation of exponentials can also be numerically delicate; a suspicious result should be checked with a numerically stable implementation rather than accepted solely because the formula was transcribed.

The main alternatives are not simply three competing algebraic rearrangements. Each makes a different statement about intrahepatic mixing: a single well-mixed compartment, separate unmixed flow paths, or axial dispersion. None automatically describes every real liver. Heterogeneous blood flow, spatially varying enzyme activity, transport processes, or other departures from the assumed structure can limit the relevance of a simple formulation. A difference between predictions is therefore not, by itself, evidence that one model is correct or that the input parameters are wrong. The calculations quantify the consequences of the stated assumptions.

Common Pitfall

Common mistake: “The same \(Q\), \(f_u\), and \(CL_{int}\) must give the same hepatic clearance, so any different answer is an arithmetic error.”

Correction: Those shared inputs fix \(R\), but the models use different concentration and mixing assumptions. At positive \(R\), the well-stirred and parallel-tube equations generally predict different extraction, and the dispersion model depends additionally on \(\mathrm{DN}\). First verify units and formula inputs; if those are consistent, a difference is an expected model-structure consequence.

Key takeaway: Equal physiological and drug inputs do not force equal predictions across liver models. At low extraction the models tend to agree to first order; at intermediate extraction their mixing assumptions can produce material differences; at high extraction their clearances approach the shared flow limit.

Check Your Understanding

Use the equations and assumptions above. Each problem is followed by its solution.

For \(R=0.500\) and \(Q=80.0\ \mathrm{L/h}\), which predicts greater clearance, the well-stirred or parallel-tube model? Calculate both.
Why do the well-stirred and parallel-tube models agree to first order at low \(R\) but not exactly?
What structural feature does \(\mathrm{DN}\) represent, and which limiting model is approached as it tends to zero?
A calculation gives a hepatic clearance above \(Q\) under one of the formulations in this tutorial. Name two checks to make before interpreting the result.

Solutions. 1. The well-stirred result is \(80.0(0.500/1.500)=26.6667\ \mathrm{L/h}\). Since \(e^{-0.500}=0.606531\), the parallel-tube result is \(80.0(1-0.606531)=31.4775\ \mathrm{L/h}\), which is greater. 2. Their expansions begin \(E_H=R+\) higher-order terms, but the second-order terms differ: \(-R^2\) for well-stirred and \(-R^2/2\) for parallel tube. 3. \(\mathrm{DN}\) describes axial dispersion relative to flow; as it tends to zero, the dispersion formulation approaches the parallel-tube limit. 4. Check the units and blood-versus-plasma basis of \(Q\), \(f_u\), and \(CL_{int}\), and verify the formula and parameter values. Under the stated formulations, clearance should be nonnegative and no greater than \(Q\).