Worked Calculation of Clearance From Covariates and Eta

From a Parameter Equation to an Individual Clearance

A hierarchical parameter equation becomes concrete when the model values are substituted for a particular subject. For clearance, the calculation combines a population typical value, a covariate adjustment for weight, and a subject-specific random effect. These components are multiplicative in the parameter equation, so the final individual clearance depends on all three.

This tutorial uses hypothetical values throughout. The goal is to calculate clearance conditional on specified covariate and eta values. It does not estimate those values from concentration data.

For the examples, \(CL_{\mathrm{typ}}=6.00\ \mathrm{L/h}\), \(WT_{\mathrm{ref}}=70\ \mathrm{kg}\), and \(\theta_{WT}=0.50\). The weight ratio is dimensionless because both weights use kilograms, and eta is dimensionless because it appears in an exponential. The equation therefore gives \(CL_i\) in the same clearance units as \(CL_{\mathrm{typ}}\).

The typical clearance applies at the reference weight when \(\eta_{CL,i}=0\). For another subject, the weight term adjusts clearance according to the specified covariate relationship, while \(\exp(\eta_{CL,i})\) adjusts it according to that subject's eta. A positive eta gives a multiplier greater than 1; a negative eta gives a multiplier between 0 and 1; and eta equal to zero gives a multiplier of 1.

Why the Equation Has This Form

The weight component is log-linear: a change in log weight relative to the reference is multiplied by \(\theta_{WT}\). The random effect is also added on the log scale. In a log-scale parameter equation, the combined contributions can be written as

Exponentiating both sides gives the individual-scale equation. In particular, the exponential of a sum is the product of exponentials, and \(\exp\{\theta_{WT}\log(r)\}=r^{\theta_{WT}}\) for a positive ratio \(r\):

This derivation also explains why eta is not added directly to clearance in this model. It is added on the log scale and therefore acts as a multiplier on the clearance scale. The covariate and eta contributions can be calculated separately, then multiplied by the typical clearance.

Calculating and Comparing Several Subjects

When calculating for multiple subjects, keep the population parameters and reference weight fixed, but substitute each subject's own weight and eta. The table below uses hypothetical inputs and the same model throughout. Calculations use unrounded factors, with final clearances rounded to three decimal places.

A useful check is to calculate each multiplicative contribution before combining them. A value above 1 raises clearance relative to the typical value; a value below 1 lowers it. But opposite-direction effects can cancel partially or completely, so one should not infer the direction of the final change from weight or eta alone.

Assumptions and Calculation Checks

A numerical result is meaningful only under the model and input conventions that define it. Before substituting values, verify the following:

• Use the specified parameter equation. The exponent \(\theta_{WT}\), reference weight, and random-effect placement must match the model being calculated.
• Use compatible units. The subject and reference weights must use the same units, so their ratio is dimensionless. The output clearance has the units of \(CL_{\mathrm{typ}}\).
• Use the eta for the subject and parameter in question. \(\eta_{CL,i}\) is not residual error, and it is not automatically shared with another parameter.
• Interpret the inputs as given. If the calculation specifies an eta value, use it as an input. The equation itself does not determine eta from weight.
• Keep precision through the calculation. Round the reported result at the end rather than rounding each factor early, especially when factors offset one another.

The exponential form guarantees a positive clearance for a positive typical clearance and positive weight ratio, regardless of the eta value. If a calculation gives a nonpositive clearance, check whether eta was incorrectly added to clearance, whether a sign or parenthesis was misplaced, or whether an inappropriate parameterization was used. If a result is far from expectation, first inspect the reference weight and the eta value; the equation can be evaluated correctly even when one of those inputs was entered incorrectly.

Common Mistakes and Alternatives

A frequent mistake is to calculate \(6.00+0.25\) and treat \(0.25\) as a clearance increment. In this model, eta is on the log scale, so its contribution is \(\exp(0.25)\), a dimensionless multiplier. Another mistake is to omit the weight ratio or to use \(WT_{\mathrm{ref}}/WT_i\), which reverses the covariate adjustment when the exponent is positive. Writing the full equation before substituting helps prevent both errors.

A second tempting shortcut is to calculate the weight-adjusted value and stop, as though the covariate-adjusted prediction were the individual clearance. That result corresponds to setting eta to zero. It is the model's value for an eta-zero subject with that weight, not the result for a subject whose specified eta is nonzero. Conversely, applying eta twice—once as a multiplier and again as an additive adjustment—also changes the model and is incorrect.

An additive random effect on the parameter scale is a different model: it uses an additive rather than exponential adjustment and has different implications, including for positivity. For this calculation, follow the specified exponential equation rather than switching parameterizations. Also distinguish an individual clearance calculated at a supplied eta from a population summary. A single conditional calculation does not by itself give an arithmetic mean across subjects.