Why Albumin Binding Does Not Equal Clearance
Liver disease can reduce plasma albumin, changing how much of an albumin-bound drug is free in plasma. This affects measured total concentration and can affect distribution and hepatic elimination, but these effects are not interchangeable. A larger free fraction does not, by itself, establish that unbound exposure will rise: the result also depends on how the drug is cleared.
The key distinction is between a change in binding and a change in the processes that remove drug. Binding describes the partition of drug between free and protein-bound forms. Clearance describes removal from the body. In a PBPK model, changing albumin can change the free fraction while other inputs, such as intrinsic metabolic capacity, are held fixed. Whether those other inputs are also altered in a particular patient is a separate question.
A simple binding model assumes reversible equilibrium between drug and albumin, one effective class of binding sites, and no saturation over the concentrations considered. If \(K_a\) is an association constant and \([\text{Alb}]\) is albumin concentration, the bound concentration is \(C_b=K_a[\text{Alb}]C_u\). Since \(C_{\text{tot}}=C_u+C_b\), this model gives:
Within these assumptions, lowering albumin increases \(f_u\). The relationship is not a universal albumin correction: other proteins, competing ligands, binding-site saturation, and disease-related changes in drug distribution can matter. In particular, this equation describes binding in plasma; it does not specify hepatic clearance.
Worked Example: Albumin Concentration and Free Fraction
Suppose a hypothetical drug follows the linear binding model above, with \(K_a=0.50\) L/g. Albumin concentration is 4.0 g/L in a reference condition and 2.0 g/L in a liver-disease scenario. Calculate \(f_u\) in each condition.
At 4.0 g/L, \(K_a[\text{Alb}]=(0.50\ \text{L/g})(4.0\ \text{g/L})=2.0\), so \(f_u=1/(1+2.0)=0.333\).
At 2.0 g/L, \(K_a[\text{Alb}]=(0.50\ \text{L/g})(2.0\ \text{g/L})=1.0\), so \(f_u=1/(1+1.0)=0.500\). The model therefore predicts a higher free fraction at the lower albumin concentration.
For illustration, if total plasma concentration were 10 mg/L in both conditions, \(C_u\) would be \((0.333)(10)=3.33\) mg/L and \((0.500)(10)=5.00\) mg/L, respectively. This comparison holds total concentration fixed; it is not a prediction that total concentrations remain the same after dosing. Clearance and distribution can change the concentrations that result.
Binding Enters the Well-Stirred Liver Model
To connect binding to hepatic elimination, consider the well-stirred liver model. It assumes that drug entering the liver mixes rapidly, that hepatic blood flow \(Q_h\) and intrinsic clearance \(CL_{\text{int}}\) are defined for the modeled condition, and that elimination depends on unbound drug available to the eliminating process. Here \(CL_{\text{int}}\) is treated as clearance with respect to unbound drug. The model is useful for isolating the role of \(f_u\); it does not assert that liver disease leaves blood flow, intrinsic clearance, or shunting unchanged.
The distinction is especially clear for a constant intravenous input rate \(R_{\text{in}}\) at steady state, assuming hepatic clearance is the only elimination route and the model parameters remain constant. Total concentration is input rate divided by clearance, \(C_{\text{tot}}=R_{\text{in}}/CL_h\). Multiplying by \(f_u\) gives unbound concentration. Substituting the well-stirred expression yields:
This derivation makes an important point: the unbound steady-state concentration is not determined by \(f_u\) alone. It also depends on intrinsic clearance and blood flow. The limiting cases help interpret the result.
If \(f_u CL_{\text{int}}\ll Q_h\), then \(CL_h\approx f_u CL_{\text{int}}\). For the specified constant intravenous input, \(C_u\approx R_{\text{in}}/CL_{\text{int}}\). With \(CL_{\text{int}}\) fixed, increasing \(f_u\) can increase clearance and lower total concentration, while unbound concentration changes little.
If \(f_u CL_{\text{int}}\gg Q_h\), then \(CL_h\approx Q_h\). For the same input, \(C_{\text{tot}}\approx R_{\text{in}}/Q_h\) and \(C_u\approx f_u R_{\text{in}}/Q_h\). With flow held fixed, increased \(f_u\) can therefore raise unbound concentration while total concentration changes relatively little.
These are limiting approximations, not categories that make binding irrelevant in one group and decisive in another. Intermediate-extraction drugs require the full expression. The steady-state results also depend on the stated intravenous-input and single-clearance assumptions; an oral dose, additional routes of elimination, altered distribution, or changing hepatic physiology can produce different concentration profiles.
Worked Example: A Low-Extraction Drug at Steady State
Consider a hypothetical drug given by constant intravenous infusion at \(R_{\text{in}}=6.00\) mg/h. Let \(Q_h=90.0\) L/h and \(CL_{\text{int}}=3.00\) L/h. Compare \(f_u=0.0200\) with \(f_u=0.0400\), holding flow and intrinsic clearance fixed. Assume hepatic elimination is the only route and steady state has been reached.
For \(f_u=0.0200\), \(f_u CL_{\text{int}}=(0.0200)(3.00)=0.0600\) L/h, which is much smaller than \(Q_h\). The well-stirred clearance is:
Thus \(C_{\text{tot}}=6.00/0.05996=100.1\) mg/L, and \(C_u=(0.0200)(100.1)=2.001\) mg/L.
For \(f_u=0.0400\), \(f_u CL_{\text{int}}=0.120\) L/h, still much smaller than flow. Now:
The resulting total concentration is \(C_{\text{tot}}=6.00/0.1198=50.07\) mg/L, while \(C_u=(0.0400)(50.07)=2.003\) mg/L. In this low-extraction example, doubling \(f_u\) approximately doubles clearance and halves total concentration, but unbound concentration is nearly unchanged. The small difference reflects use of the full model rather than its low-extraction approximation.
Worked Example: A High-Extraction Drug at Steady State
Now consider a separate hypothetical drug with the same \(R_{\text{in}}=6.00\) mg/h and \(Q_h=90.0\) L/h, but \(CL_{\text{int}}=90{,}000\) L/h. Compare \(f_u=0.0200\) and \(f_u=0.0400\), again assuming hepatic elimination alone and fixed model parameters.
At \(f_u=0.0200\), \(f_u CL_{\text{int}}=1{,}800\) L/h, much greater than \(Q_h\). The full model gives \(CL_h=(90.0)(1{,}800)/(90.0+1{,}800)=85.71\) L/h. Therefore \(C_{\text{tot}}=6.00/85.71=0.07000\) mg/L and \(C_u=(0.0200)(0.07000)=0.001400\) mg/L.
At \(f_u=0.0400\), \(f_u CL_{\text{int}}=3{,}600\) L/h, and \(CL_h=(90.0)(3{,}600)/(90.0+3{,}600)=87.80\) L/h. Thus \(C_{\text{tot}}=6.00/87.80=0.06833\) mg/L and \(C_u=(0.0400)(0.06833)=0.002733\) mg/L. Here, doubling \(f_u\) nearly doubles unbound concentration, while total concentration decreases slightly. This is consistent with clearance approaching the fixed blood-flow limit.
Using the Relationship in a PBPK Model
A PBPK model should represent binding and clearance as distinct inputs or processes. If the drug is albumin-bound and binding is plausibly linear, albumin concentration can inform \(f_u\) through a binding relationship or through measured condition-specific \(f_u\). The resulting \(f_u\) can then enter the relevant tissue-distribution and hepatic-clearance calculations. It should not be used as a substitute for disease-related changes in hepatic flow, functional liver mass, or other determinants of elimination.
The earlier material on scaling functional liver mass and on hepatic blood flow establishes that liver disease can affect more than one physiological determinant. When evaluating a binding effect, hold those other determinants fixed as an explicit scenario assumption. If they are not fixed, change them separately and examine their combined consequences; otherwise a predicted exposure change cannot be attributed to albumin binding alone. The disease-dependent enzyme-capacity question is likewise distinct from the binding calculation.
A simple albumin adjustment to total concentration is useful only for a narrow purpose: under a specified binding model, it can describe how total and unbound concentrations relate at a given condition. It does not generally predict the concentration after a dose, because that prediction also requires clearance, input, distribution, and relevant disease physiology. A PBPK model with explicit binding and clearance is more informative when the aim is to predict time courses or compare total and unbound exposure.
| Approach | What it can answer | Important limitation |
|---|---|---|
| Binding calculation alone | How a specified albumin change alters \(f_u\) under stated binding assumptions | Does not predict clearance or post-dose concentrations |
| Well-stirred clearance calculation | How \(f_u\), \(Q_h\), and \(CL_{\text{int}}\) jointly determine hepatic clearance under model assumptions | Does not by itself capture every disease effect or full body distribution |
| PBPK model with binding and relevant physiology | How changed binding and specified physiological inputs affect predicted concentration profiles | Predictions depend on input quality, structural assumptions, and validation |
Diagnostics and Common Pitfalls
Compare predicted or measured total and unbound concentrations together when possible. A change in total concentration without a corresponding change in unbound concentration can be consistent with altered binding, especially for a low-extraction drug under the assumptions illustrated above. A rise in unbound concentration may be expected under some high-extraction conditions, but it may also reflect altered clearance, flow, distribution, or other processes. These patterns are diagnostic clues, not proof of a single mechanism.
Check whether the assumed binding relationship is appropriate. If measured \(f_u\) does not change as predicted from albumin, possible explanations include binding to other plasma proteins, competition from endogenous substances, nonlinearity, or measurement issues. Also check that the extraction regime used for an approximation is supported by the parameters: compare \(f_u CL_{\text{int}}\) with \(Q_h\), rather than classifying a drug by its name or by albumin binding alone.
- Mistake: “Albumin fell, so unbound concentration must rise by the same factor.” That conclusion treats total concentration as fixed. After dosing, clearance and distribution can change total concentration; under low extraction, the rise in \(f_u\) can be offset by a rise in hepatic clearance.
- Correction: State what is held fixed. For a specified total concentration, \(C_u=f_u C_{\text{tot}}\). For a concentration prediction after dosing, use a clearance and input model as well.
- Mistake: “A larger free fraction means hepatic clearance is unchanged.” In the well-stirred model, \(f_u\) appears directly in \(CL_h\).
- Correction: Evaluate \(CL_h\) using \(Q_h\), \(f_u\), and \(CL_{\text{int}}\), and distinguish the binding change from any concurrent change in those other parameters.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions.
- Under the linear albumin-binding model, what happens to \(f_u\) when albumin concentration falls, assuming \(K_a\) is unchanged?
- A low-extraction drug has fixed \(CL_{\text{int}}\) and is given by constant intravenous infusion. In the low-extraction limit, what happens approximately to total and unbound steady-state concentrations when \(f_u\) doubles?
- Why is it incorrect to infer a twofold rise in unbound concentration from a twofold rise in \(f_u\) without specifying what is held fixed?
- For a well-stirred liver, which comparison determines whether the low- or high-extraction approximation is appropriate?
Solutions. (1) Since \(f_u=1/(1+K_a[\text{Alb}])\), lowering albumin lowers the denominator and increases \(f_u\). (2) In the low-extraction limit, \(CL_h\approx f_u CL_{\text{int}}\); doubling \(f_u\) approximately doubles clearance and halves total steady-state concentration, while \(C_u\approx R_{\text{in}}/CL_{\text{int}}\) remains approximately unchanged. (3) The relation \(C_u=f_u C_{\text{tot}}\) gives a twofold increase only if total concentration is held fixed. After dosing, total concentration can change as clearance and distribution change. (4) Compare \(f_u CL_{\text{int}}\) with \(Q_h\): much smaller supports the low-extraction approximation, and much larger supports the high-extraction approximation.