Why Represent Clearance as Changing Over Time?
A constant-clearance model is useful when elimination capacity can reasonably be treated as stable over the time window being modeled. But when clearance changes during that window, a single value may not describe the resulting concentration decline adequately. The central modeling question is then not just what clearance is at one time, but how clearance accumulates across the interval. That distinction matters because drug concentration responds to the cumulative elimination experienced since the previous time point.
This tutorial considers a one-compartment model with a fixed apparent volume of distribution and clearance that changes according to a specified function of time or a specified state variable. The function may be chosen to represent a smooth trend or a sequence of intervals; a state variable may represent a changing system state without requiring that its biology be fully modeled here. These are structural choices: they describe how the model’s expected concentration changes over time.
Model and Assumptions
Let \(A(t)\) be the amount of drug in the compartment, and suppose concentration is \(C(t)=A(t)/V\), with \(V\) constant and positive. If there is no input after an intravenous bolus at time zero, the model is
The model assumes a well-mixed compartment; linear elimination with respect to concentration at each instant; a positive, specified clearance function; and no further drug input after the bolus. Clearance can change with time, but it is not itself a function of the concentration in this model. The fixed-volume assumption is important: if volume changes, the relation between amount and concentration also changes and requires a different model.
Because \(A(t)=VC(t)\) and \(V\) is constant, the amount equation becomes
This is a linear differential equation with a time-dependent coefficient. Solving it gives the concentration at time \(t\) after a bolus with concentration \(C(0)=C_0\):
The integration variable \(u\) distinguishes the time being accumulated from the final time \(t\). The integral \(\int_0^t CL(u)\,du\) has units of volume; division by \(V\) makes the exponent dimensionless. With input \(R_{\mathrm{in}}(t)\), the concentration equation instead includes an input term, \(dC/dt=R_{\mathrm{in}}(t)/V-[CL(t)/V]C(t)\). The same time-varying elimination coefficient applies, but the no-input bolus formula alone is not the full solution.
Why Cumulative Clearance Determines the Profile
The solution can be justified by separating the concentration change from concentration itself. For positive \(C(t)\), divide the differential equation by \(C(t)\), then integrate from zero to \(t\):
The left side is \(\log(C(t))-\log(C_0)\). Exponentiating both sides yields the stated concentration solution. This argument requires that the specified clearance be integrable over the interval. If clearance is positive, the integral is nondecreasing, so concentration cannot rise in this no-input model.
For \(t>0\), define the interval-average clearance as \(\overline{CL}_{[0,t]}=(1/t)\int_0^t CL(u)\,du\). Then the exponent is \(-t\overline{CL}_{[0,t]}/V\). Thus a changing clearance has the same concentration at time \(t\) as a constant clearance equal to its interval average over \([0,t]\). This does not mean that either the initial or endpoint clearance generally gives that answer.
Worked Example: Clearance Increasing Linearly
Suppose a hypothetical one-compartment IV bolus has \(C_0=10\) mg/L and \(V=20\) L. Let clearance be \(CL(t)=2(1+0.2t)\) L/h, where \(t\) is in hours. Calculate the concentration at 4 h and compare it with a constant-clearance prediction using the initial clearance.
The model has positive clearance throughout the interval, fixed volume, and no post-bolus input. Integrating the specified function gives
The cumulative-clearance exponent is \(11.2/20=0.5600\). Therefore,
If clearance were instead held at its initial value of 2 L/h, the exponent would be \((2\text{ L/h})(4\text{ h})/(20\text{ L})=0.4000\), giving \(10e^{-0.4000}=6.703\) mg/L. The constant-initial-clearance model predicts a higher concentration here because clearance increases above its initial value during the interval.
Other Useful Clearance Functions
A smooth function can be selected when clearance is expected to change gradually. For instance, \(CL(t)=CL_0e^{kt}\), with \(CL_0>0\) and \(k>0\), describes an increasing clearance function; \(k\) has units of inverse time. This formula is a mathematical specification, not by itself an explanation of why clearance changes. Its integral is \((CL_0/k)(e^{kt}-1)\). A decreasing exponential trend can be represented with \(k<0\), provided the resulting clearance remains positive.
Worked Example: Exponentially Increasing Clearance
For a hypothetical bolus, let \(C_0=12\) mg/L, \(V=30\) L, \(CL(t)=3e^{0.25t}\) L/h, and evaluate at \(t=3\) h. Assume no further input and a fixed volume, as required by the model.
The integrated clearance is
Dividing by volume gives an exponent of \(13.404/30=0.4468\). The predicted concentration is
A constant clearance equal to the initial value, 3 L/h, would give exponent \((3)(3)/30=0.3000\), so \(12e^{-0.3000}=8.890\) mg/L. A constant clearance equal to the endpoint value, \(3e^{0.75}=6.351\) L/h, would give exponent \((6.351)(3)/30=0.6351\), or \(12e^{-0.6351}=6.359\) mg/L. The changing-clearance result lies between these two predictions because clearance increases smoothly from 3 to 6.351 L/h. The endpoint calculation is a comparison, not a substitute for integrating clearance over the interval.
Clearance may also be specified through a state variable \(x(t)\). The state variable follows its own equation, and clearance is defined from its value. To use the concentration solution, the state trajectory must be known or solved well enough to calculate the integral of \(CL(t)\). This representation can be convenient when clearance is linked to a changing system state, while leaving the state dynamics explicit rather than hiding them inside an unspecified time trend.
Worked Example: Clearance Driven by a Decaying State Variable
Consider a hypothetical model with \(C_0=8\) mg/L, \(V=25\) L, and a state variable \(x(t)=e^{-0.5t}\). Specify clearance as \(CL(t)=2.5[1+0.6x(t)]\) L/h. Find the concentration at 2 h. The state is prescribed, clearance remains positive, volume is fixed, and there is no post-bolus input.
Substitute the state trajectory into the clearance function and integrate:
The exponent is \(6.896/25=0.2759\), so
For comparison, if clearance remained at its baseline value of 2.5 L/h, the concentration would be \(8e^{-(2.5)(2)/25}=6.550\) mg/L. The state-dependent model predicts a lower concentration because its clearance is above baseline during the interval. This example depends on the specified state trajectory; a different state equation would produce a different clearance history and concentration profile.
Checking the Model and Choosing a Representation
The first diagnostic is structural: plot or calculate the implied \(CL(t)\) across the modeled interval and check that it is positive, finite, and consistent with the intended trend. Then check that the integrated clearance gives a nonincreasing concentration after a bolus when there is no input. A rising predicted concentration in this setting signals a sign error, an invalid clearance function, or an additional process that has not been represented in the model.
Observed data can also reveal mismatch, though they do not identify its cause on their own. Systematic residual patterns over time may indicate that constant clearance is inadequate, but similar patterns can result from an incorrect absorption or input model, an incorrect volume assumption, or other structural misspecification. Sparse sampling can make the shape of clearance change difficult to distinguish from these alternatives. A flexible time function may fit observed concentrations well without establishing a biological explanation for the trend.
The main alternatives differ in what they assume about clearance across the interval:
| Representation | How clearance is specified | Main consideration |
|---|---|---|
| Constant clearance | One clearance value applies throughout. | Simple and interpretable when a stable value is adequate. |
| Piecewise constant clearance | A separate value applies over each defined interval. | Useful for abrupt changes, but interval boundaries and values need support from the design or knowledge of the process. |
| Prescribed smooth function | A formula directly specifies \(CL(t)\). | Represents a smooth trend compactly, but its shape is an assumption to assess. |
| State-variable model | A state trajectory determines \(CL(t)\). | Makes the changing state explicit, but requires a defensible state equation and sufficient information to estimate its parameters. |
A constant-clearance approximation can be reported using the interval-average clearance if only the concentration at the end of an interval matters. Using the initial or endpoint value instead generally gives a different result. More precisely, the endpoint-based exponent \(tCL(t)/V\) equals the correct exponent when the endpoint clearance equals the interval-average clearance. This equality holds for constant clearance, but it can also hold for a nonconstant function; it is the equality of the averages, not constancy itself, that determines whether the exponents match.
Limitations and a Common Pitfall
The formula \(C(t)=C_0\exp[-\int_0^t CL(u)/V\,du]\) assumes linear elimination at each time, even though its clearance coefficient varies. It is not a general solution for saturable elimination, concentration-dependent clearance, or changing volume. Those situations require a model whose differential equation represents the relevant dependence. Nor does a good fit alone prove that the chosen function describes the underlying mechanism. With limited measurements, several clearance histories can have the same cumulative clearance at the observation times.
- Pitfall: using endpoint clearance for the whole interval. The reasoning “clearance at time \(t\) is \(CL(t)\), so cumulative clearance is \(tCL(t)\)” ignores earlier values. The correction is to integrate the full function, or use its interval average.
- Pitfall: treating a trend as a mechanism. A fitted increasing or decreasing function describes a pattern in the model. It does not, by itself, establish its biological cause.
- Pitfall: forgetting the volume assumption. The concentration solution uses fixed \(V\). If volume changes, substituting a time-varying clearance into this formula is not sufficient.
The practical value of time-varying clearance is that it separates the clearance history from the resulting concentration profile in a transparent way: a specified history can be integrated and checked against concentration data. The choice should nevertheless be restrained. A simple constant model is preferable when it adequately describes the data and scientific question; additional time structure needs justification and information capable of supporting it.
Check Your Understanding
Work through each question, then compare with the solutions below.
- For a bolus model with constant \(V\) and no input, what quantity determines the concentration exponent when clearance varies?
- A hypothetical model has \(V=20\) L, \(C_0=15\) mg/L, clearance 2 L/h from 0 to 2 h, then 4 L/h from 2 to 5 h. Calculate \(C(5)\).
- When does the endpoint-clearance approximation produce the same exponent as the integrated model? Must clearance be constant for this equality?
- Why is it not appropriate to use the bolus formula unchanged when volume varies over time?
Solutions. (1) The exponent is \(-\int_0^t CL(u)/V\,du\), or \(-1/V\) times the cumulative clearance when \(V\) is constant. (2) The integrated clearance is \((2)(2)+(4)(3)=16\) L; the exponent magnitude is \(16/20=0.8000\), so \(C(5)=15e^{-0.8}=6.740\) mg/L. This uses the stated piecewise clearance, positive throughout, with fixed volume and no input. (3) The endpoint approximation matches when \(CL(t)=(1/t)\int_0^t CL(u)\,du\), so endpoint clearance equals interval-average clearance. Constancy is sufficient but not necessary. (4) The derivation uses \(A(t)=VC(t)\) with the same fixed \(V\) throughout; if \(V\) changes, differentiating amount no longer gives the same concentration equation.