Parameterizing Vmax and Km in Population Models

Why Parameterization Changes Interpretation

Michaelis–Menten elimination is described by two positive parameters, but there is more than one useful way to express them. The direct parameterization uses \(V_{\max}\), the maximum elimination rate, and \(K_m\), the concentration at which the rate is half its maximum. An alternative uses \(CL_{\mathrm{int}}=V_{\max}/K_m\), called intrinsic clearance, together with \(K_m\). Both forms can describe exactly the same concentration-dependent elimination. Their difference is the parameter scale on which population values and covariate effects are expressed.

The choice matters because the parameters foreground different features of the model. \(V_{\max}\) directly describes capacity at high concentration. \(CL_{\mathrm{int}}\) describes the limiting clearance at low concentration. In a nonlinear mixed-effects (NLME) model, a population parameterization also determines how typical parameter values and covariate relationships are stated. It does not, by itself, change the underlying elimination law.

For dimensional consistency, if \(C\) is measured in mg/L and the elimination rate in mg/h, then \(K_m\) is in mg/L, \(V_{\max}\) is in mg/h, and \(CL_{\mathrm{int}}\) is in L/h. The rate divided by concentration, \(R(C)/C\), also has units of clearance. For this model that concentration-specific clearance is

This is a clearance associated with the current concentration, not a constant parameter across all concentrations. It approaches \(CL_{\mathrm{int}}\) as concentration becomes small and decreases as elimination saturates.

Two Equivalent Forms of the Population Model

At the population level, one may describe typical positive parameters directly as \(V_{\max}\) and \(K_m\), or as \(CL_{\mathrm{int}}\) and \(K_m\). For example, the corresponding rate equations are

They are equivalent when \(V_{\max}=CL_{\mathrm{int}}K_m\). Thus, moving between forms is an exact change of parameter coordinates, provided the parameters are positive and the conversion is applied consistently. If the model uses a one-compartment volume \(V\), with amount \(A\) and \(C=A/V\), the elimination contribution to the amount equation is \(dA/dt=-R(A/V)\). The parameter conversion does not alter this equation’s predicted rate at any concentration.

In an NLME model, the equations above specify the structural relationship between concentration and elimination. Population fixed effects may describe typical parameters or their covariate dependence; the separate model for between-subject variability must be defined on a compatible scale. The examples below focus on the parameterization and its implications, not on a particular variability model.

What Each Parameterization Makes Direct

In the direct form, \(V_{\max}\) is the high-concentration capacity: as \(C\) increases without bound, the rate approaches \(V_{\max}\). \(K_m\) locates the bend in the rate curve: substituting \(C=K_m\) gives \(R(K_m)=V_{\max}/2\). This form is convenient when the scientific question or prior knowledge is naturally about maximum capacity and the concentration scale for half-maximal elimination.

In the intrinsic-clearance form, \(CL_{\mathrm{int}}\) is the low-concentration limit of \(R(C)/C\), and \(K_m\) still governs the transition toward saturation. At concentrations much smaller than \(K_m\), elimination is approximately first-order: the rate is approximately \(CL_{\mathrm{int}}C\). This form can be useful when low-concentration elimination is the quantity of direct scientific or covariate interest.

The relationship between a covariate effect and the model parameters depends on which quantities are modeled directly. For a covariate \(x\), a log-linear specification of the direct parameters could be

The implied intrinsic clearance is then

Thus, the covariate effect on intrinsic clearance is the difference between the direct log-scale effects on \(V_{\max}\) and \(K_m\). Alternatively, one can model \(CL_{\mathrm{int}}\) and \(K_m\) directly, in which case the implied capacity is their product. A coefficient on one parameter is not automatically an effect of the same size or direction on the other derived parameter.

Why the Low-Concentration Interpretation Holds

The approximation follows directly from the rate equation. Divide the Michaelis–Menten rate by concentration, and use \(V_{\max}=CL_{\mathrm{int}}K_m\):

When \(C/K_m\) is small, \(1/(1+C/K_m)\) is close to 1. More precisely, for \(u=C/K_m\) near zero, \(1/(1+u)=1-u+O(u^2)\). Therefore,

This is a local approximation for concentrations small relative to \(K_m\), not an exact equality at all concentrations. At the other extreme, dividing numerator and denominator of \(R(C)\) by \(C\) gives \(R(C)=V_{\max}/(1+K_m/C)\), which approaches \(V_{\max}\) as \(C\) becomes large. These limits explain why intrinsic clearance emphasizes low concentrations and \(V_{\max}\) emphasizes capacity.

Choosing and Checking a Parameterization

Neither parameterization is universally superior. The direct form makes capacity explicit; the intrinsic-clearance form makes the low-concentration slope explicit. A sound choice is one whose parameters and covariate effects answer the scientific question clearly, while preserving the same structural model. In either form, positivity is required for the standard Michaelis–Menten interpretation, and units must be consistent in every term of the rate equation.

Useful checks include evaluating the equation at \(C=K_m\), at a concentration far below \(K_m\), and at one far above it. The rate at \(K_m\) must equal half of \(V_{\max}\). The low-concentration clearance should approach \(CL_{\mathrm{int}}\), and the high-concentration rate should approach \(V_{\max}\). Also verify that \(CL_{\mathrm{int}}K_m\) has units of amount per time and equals the reported \(V_{\max}\).

If two parameterizations are fitted to the same data and structural assumptions, compare predictions after translating parameters using \(V_{\max}=CL_{\mathrm{int}}K_m\). Differences may arise from implementation, constraints, or how covariate and variability structures were specified; they are not evidence that the algebraically equivalent rate equations differ. Parameter meaning alone also does not guarantee reliable estimation: the observed concentration range and design determine which aspects of the curve are informed.