Choosing Among Elimination Structures

Why Compare Elimination Structures?

An elimination structure specifies how drug leaves the body as concentration changes and, in some models, as time passes. Choosing among structures matters because the same observed decline can imply different predictions at concentrations or doses not well represented in the data. A constant fractional decline, a constant amount eliminated per unit time, capacity-limited elimination, and clearance that changes with time are not interchangeable descriptions.

The earlier material on first-order, zero-order, Michaelis–Menten, and time-varying clearance models establishes the component structures. Here the focus is on comparing their predictions. To make the contrasts explicit, consider a one-compartment IV bolus with constant volume \(V\), concentration \(C=A/V\), amount \(A\), and no further input after the bolus. This setup isolates elimination; for other routes or dosing schedules, the input process must also be represented before interpreting the observed concentration profile.

Definition: An elimination structure describes the rate at which drug amount is removed as a function of concentration, time, or both. First-order elimination removes a constant fraction of the amount per unit time; zero-order elimination removes a constant amount per unit time over its stated operating range; Michaelis–Menten elimination has a saturable rate that rises with concentration toward a maximum; and time-varying clearance allows clearance to change with time even when concentration is held fixed.

Predictions From the Candidate Structures

Under first-order elimination, the amount removed per unit time is \(CL\,C\), where \(CL\) is constant clearance. Since \(A=VC\), the concentration satisfies the following equation and solution after the bolus:

\frac{dC}{dt}=-\frac{CL}{V}C=-kC,\qquad C(t)=C_0e^{-kt},\qquad k=\frac{CL}{V}.

The elimination rate in amount per time increases in proportion to concentration, but the fractional rate of decline, \(-\frac{1}{C}\frac{dC}{dt}=k\), is constant. A plot of log concentration against time is therefore a straight line under the model. For an IV bolus, \(D=VC_0\), and the complete AUC is \(C_0/k=D/CL\). With unchanged parameters, doubling dose doubles concentration at every time and doubles AUC.

For zero-order elimination, suppose the body removes a constant amount \(R_0\) per unit time while drug remains available. The concentration declines linearly until it reaches zero; the model is not continued into negative concentrations. The duration of elimination and AUC after a bolus follow directly from the constant slope:

\frac{dC}{dt}=-\frac{R_0}{V},\qquad C(t)=C_0-\frac{R_0}{V}t,\qquad t_{\mathrm{stop}}=\frac{VC_0}{R_0}=\frac{D}{R_0}, \qquad AUC=\frac{C_0t_{\mathrm{stop}}}{2} =\frac{D^2}{2VR_0}.

Thus a constant absolute elimination rate produces a linear concentration decline and, under these restrictive assumptions, AUC increases with the square of dose. In practice, zero-order elimination is often an approximation over a limited concentration range rather than a claim that the same rate applies all the way to zero.

Michaelis–Menten elimination describes a rate that saturates as concentration rises. Let \(V_{\max}\) be the maximum elimination rate in amount per time, and \(K_m\) the concentration at which the rate is half of \(V_{\max}\). With constant \(V\), the concentration equation is:

\frac{dC}{dt} =-\frac{V_{\max}}{V}\frac{C}{K_m+C}.

At \(C\ll K_m\), the rate is approximately \((V_{\max}/K_m)C\), so the behavior approaches first order. At \(C\gg K_m\), the rate in amount per time approaches \(V_{\max}\), so the concentration slope approaches \(-V_{\max}/V\), resembling zero order. Unlike a single constant zero-order model, Michaelis–Menten elimination can move between these regimes as concentration falls.

For time-varying clearance that is specified as a function of time, \(CL(t)\), the model is linear in concentration at each time:

\frac{dC}{dt}=-\frac{CL(t)}{V}C,\qquad C(t)=C_0\exp\left[-\frac{1}{V}\int_0^t CL(u)\,du\right].

If the same \(CL(t)\) applies across doses, concentration remains proportional to \(C_0\), and hence to dose. The log-concentration slope can change with time, but the change is not caused by concentration itself. If clearance changes through a process coupled to drug exposure, such as an exposure-dependent enzyme response, proportionality need not hold; the time course of that process must then be represented rather than treated as a fixed \(CL(t)\).

Worked Comparisons

Worked Example: Constant Fraction Versus Constant Amount

Suppose \(V=20\) L. Compare first-order elimination with \(CL=4\) L/h to zero-order elimination with \(R_0=10\) mg/h at concentrations of 2 and 10 mg/L. Assume the zero-order rate applies at both concentrations.

For first-order elimination, the amount rate is \(CL\,C\). At 2 mg/L it is \(4\times2=8\) mg/h; at 10 mg/L it is \(4\times10=40\) mg/h. The fractional rate is \(k=CL/V=4/20=0.200\) h\(^{-1}\) at both concentrations. The concentration slope is \(-kC\): \(-0.400\) mg/L/h at 2 mg/L and \(-2.00\) mg/L/h at 10 mg/L.

For zero-order elimination, the amount rate remains 10 mg/h at either concentration. The concentration slope is \(-R_0/V=-10/20=-0.500\) mg/L/h at both concentrations. The fractional rate is not constant: it is \(0.500/2=0.250\) h\(^{-1}\) at 2 mg/L and \(0.500/10=0.0500\) h\(^{-1}\) at 10 mg/L.

The distinguishing prediction is therefore not simply that one model has a “faster” decline. First order has a constant fractional decline, while zero order has a constant absolute decline over the assumed range. Observations spanning different concentrations can test which pattern is more consistent with the data.

Worked Example: How Michaelis–Menten Rate Changes With Concentration

Let \(V=10\) L, \(V_{\max}=20\) mg/h, and \(K_m=5\) mg/L. Calculate the elimination rate in amount per time and the apparent clearance (rate divided by concentration) at \(C=1\) and \(C=20\) mg/L.

At 1 mg/L, the rate is \(20(1)/(5+1)=3.333\) mg/h. The apparent clearance is \(3.333/1=3.333\) L/h. At 20 mg/L, the rate is \(20(20)/(5+20)=16.00\) mg/h, and apparent clearance is \(16/20=0.8000\) L/h.

Concentration increased twentyfold, but the amount rate increased only \(16/3.333=4.800\)-fold because the elimination process is approaching its capacity. Apparent clearance falls as concentration rises. At low concentrations relative to \(K_m\), the rate approaches first-order behavior; at concentrations much greater than \(K_m\), it approaches the maximum rate.

Worked Example: Dose and AUC Under Saturable Elimination

Consider an IV bolus with \(V=10\) L, \(V_{\max}=20\) mg/h, and \(K_m=5\) mg/L. Compare initial concentrations \(C_0=2\) and 10 mg/L. Assume Michaelis–Menten elimination continues until concentration reaches zero, with no other elimination route or input.

Separate variables in the concentration equation: \(dt=-(V/V_{\max})(K_m+C)dC/C\). Integrating from \(C_0\) to zero gives:

AUC=\int_0^\infty C\,dt =\frac{V}{V_{\max}}\int_0^{C_0}(K_m+C)\,dC =\frac{V}{V_{\max}}\left(K_mC_0+\frac{C_0^2}{2}\right).

Here \(V/V_{\max}=10/20=0.5000\) h·L/mg. At \(C_0=2\) mg/L, the dose is \(VC_0=20\) mg and \(AUC=0.5000[(5)(2)+2^2/2]=0.5000(12)=6.000\) mg·h/L.

At \(C_0=10\) mg/L, the dose is 100 mg and \(AUC=0.5000[(5)(10)+10^2/2]=0.5000(100)=50.00\) mg·h/L. The dose rises fivefold, while AUC rises \(50/6=8.333\)-fold. In this IV-bolus setting, saturable elimination predicts a more-than-proportional increase in AUC: the fraction of dose eliminated per unit concentration-time exposure is lower at higher concentrations as the pathway saturates. This conclusion is specific to the stated elimination model and assumptions; changes in bioavailability or input can alter exposure for other reasons.

Worked Example: A Clearance Change Over Time

Suppose \(V=20\) L and clearance is 2 L/h for the first 2 h after an IV bolus, then 4 L/h for the next 2 h. Find the fraction of concentration remaining at 4 h, and describe what happens if the starting concentration is doubled.

The integrated clearance over 4 h is \((2)(2)+(4)(2)=12\) L. Therefore, \(C(4)/C_0=\exp(-12/20)=\exp(-0.600)=0.5488\). If \(C_0=10\) mg/L, then \(C(4)=5.488\) mg/L. If \(C_0\) is doubled to 20 mg/L under the same clearance schedule, \(C(4)=10.98\) mg/L, exactly twice the first prediction. The log-concentration slope changes when clearance changes at 2 h, but the fractional change is identical across the two starting concentrations.

How to Choose and Diagnose a Structure

Structural choice should be based on the pattern the model must explain, not on a single visual feature or the smallest residual from one concentration range. First ensure that the input has been represented adequately. After accounting for input, inspect both absolute concentration changes and fractional changes across the observed range. The following comparison summarizes the core predictions under the assumptions given above.

Structure	Rate signature	Concentration profile after IV bolus	Dose response when parameters are unchanged
First order	Constant fractional rate; amount rate proportional to \(C\)	Exponential; straight line on log scale	Concentrations and AUC proportional to dose
Zero order	Constant amount rate over its operating range	Linear decline until the operating range ends or drug is depleted	Under the idealized model, AUC proportional to dose squared
Michaelis–Menten	Rate rises with \(C\), approaching \(V_{\max}\)	Transitions from near-linear fractional decline at low \(C\) to near-linear absolute decline at high \(C\)	IV-bolus AUC can rise more than proportionally with dose
Time-varying clearance	Fractional rate changes with time through \(CL(t)\)	Log slope can change at particular times, irrespective of concentration	Proportional if the same clearance schedule applies across doses

Useful diagnostics include plotting observed concentrations against time on both ordinary and log scales, checking whether residual patterns change systematically across concentration ranges, and comparing profiles across doses when the input and other conditions are comparable. Curvature in a log-concentration plot alone does not establish saturation: time-varying clearance, distribution behavior, unmodeled input, or measurement error can also produce curvature. Conversely, a narrow concentration range may make first-order and Michaelis–Menten models hard to distinguish because the latter is approximately first order when \(C\ll K_m\).

Dose comparisons are especially informative when dose is the only changed condition. Proportional profiles support a linear elimination structure with fixed parameters, including a fixed time-varying clearance schedule; they do not by themselves distinguish constant clearance from time-varying clearance. A systematic dose-dependent departure from proportionality is consistent with concentration-dependent elimination, but it is not proof of it. Dose-dependent bioavailability, absorption, or other changes can produce a similar pattern and should be considered before attributing the departure to elimination.

Assumptions, Alternatives, and Limitations

These comparisons assume that the structural model is evaluated with an appropriate representation of drug input and that the measured concentration is informative about the modeled concentration. The simple equations also assume a single well-mixed compartment and constant \(V\). In a multi-compartment setting, distribution can shape the observed decline, so a changing slope need not imply changing elimination. A zero-order model must be restricted to a range where its constant rate is plausible; it cannot remove a positive amount at that rate indefinitely after the drug has been depleted.

The main alternative to choosing one of these structures is not to force a choice from sparse data, but to recognize that the data may not identify the distinction. First-order behavior may be a useful approximation to Michaelis–Menten elimination over a low-concentration range. A changing log slope may be explained by time-varying clearance or by a different structural feature. When observations do not cover the relevant concentration range or time window, estimates of \(K_m\), \(V_{\max}\), or the timing of clearance changes can be weakly determined. More flexible models can fit observed data better yet yield unstable extrapolations.

A practical comparison therefore asks whether each candidate has a plausible mechanism, whether its assumptions match the dosing and observation conditions, whether it reproduces the concentration and dose patterns, and whether important systematic residual patterns remain. A more complex structure is justified when it resolves a meaningful pattern and is supported by the available information, not merely because it has additional parameters.

Common Mistakes

Calling any curved log profile saturable elimination. Curvature can arise from time-varying clearance, distribution, or input. Compare dose levels and concentration dependence, and check the broader structural assumptions.
Confusing constant amount with constant fraction. Zero order means a constant amount rate over a stated range; first order means a constant fractional rate.
Assuming saturation makes IV-bolus AUC less than proportional. Under the stated Michaelis–Menten model, AUC per dose increases with dose, so AUC can rise more than proportionally.
Treating time-varying clearance as automatically dose dependent. A fixed clearance schedule changes the profile over time but preserves concentration proportionality across doses.
Interpreting a descriptive pattern as unique proof of a mechanism. Similar profiles can result from different structural features, particularly when sampling is sparse or covers a narrow range.

Key takeaway: Compare elimination models by asking what stays constant: the fractional rate for first order, the absolute amount rate for zero order, the saturating rate law for Michaelis–Menten, or clearance as a function of time for time-varying models. Check concentration and dose predictions together, while accounting for input and the limits of the data.

Check Your Understanding

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

At a concentration that doubles, what happens to the amount elimination rate under first-order elimination with fixed clearance?
For zero-order elimination with \(V=10\) L and \(R_0=5\) mg/h, what is the concentration slope while the model applies?
In the worked Michaelis–Menten AUC example, by what factor did AUC increase when dose increased fivefold?
Can fixed time-varying clearance produce a changing log-concentration slope while preserving proportionality across doses?

Solutions. 1. The amount rate doubles because it equals \(CL\,C\) and \(CL\) is fixed. 2. The slope is \(-R_0/V=-5/10=-0.500\) mg/L/h. 3. AUC increased by \(50/6=8.333\)-fold, more than proportionally. 4. Yes. The slope is \(-CL(t)/V\), so it can change over time; if the same \(CL(t)\) applies at every dose, the concentration solution is proportional to \(C_0\) throughout.