Why In Vitro Scaling Can Overpredict In Vivo Clearance

Why a Scaled Prediction Can Be Too High

In vitro-to-in vivo extrapolation (IVIVE) converts a measured rate in a laboratory preparation into a predicted clearance for an organ or organism. The calculation can be arithmetically correct and still overpredict in vivo clearance: a laboratory activity may not represent activity in the target tissue, an adjustment may be inappropriate for the quantity measured, or a scaling operation may be applied more than once. The practical task is to trace each factor from the assay measurement to the final prediction and verify what it represents.

This tutorial focuses on four sources of overprediction: assay-specific activity treated as representative of in vivo activity, an incorrect incubation-binding correction, unsuitable biological scaling factors, and double counting. These are distinct problems. A large prediction alone does not identify which one occurred; the calculation and the provenance of each input must be examined.

For microsomal scaling, a clearance expressed per unit microsomal protein is multiplied by microsomal protein per gram of liver (MPPGL) and liver mass. The resulting whole-liver intrinsic clearance has units of volume per time. For a hepatocyte assay, the corresponding quantities are clearance per cell number, cells per gram of liver, and liver mass. Each factor must match the preparation, species, and units of the measured clearance.

Intrinsic clearance is not the same as hepatic clearance. Under the hepatic well-stirred model, hepatic clearance depends on intrinsic clearance, hepatic blood flow, and the unbound fraction in plasma. With \(Q_{\mathrm{h}}\) denoting hepatic blood flow and \(f_{\mathrm{u,p}}\) the unbound fraction in plasma, the model is:

The model assumes a well-mixed liver, linear intrinsic elimination over the concentration range considered, and consistent units for flow and clearance. It is useful for translating a scaled intrinsic clearance into an organ-clearance prediction, but it cannot repair an incorrect input.

How the Main Sources of Overprediction Arise

Assay-specific activity. An observed rate belongs to the preparation and conditions in which it was measured. A particular enzyme system may have greater activity per unit enzyme or preparation than the relevant activity in liver. Scaling that rate by an enzyme abundance or tissue quantity does not, by itself, make the activity representative of the target tissue. Any relative activity adjustment must be appropriate to the assay and target system. An excessive or inapplicable adjustment factor can therefore inflate the prediction.

Incorrect binding correction. Dividing apparent clearance by \(f_{\mathrm{u,inc}}\) increases the estimate when some substrate is bound in the incubation. If the measured clearance was already calculated on an unbound-concentration basis, dividing again inflates it. Even when correction is needed, the fraction must belong to the relevant incubation and conditions. A fraction from another preparation or an incorrectly entered value can produce a large error, especially when the fraction is small.

Unsuitable scaling factors. MPPGL, liver mass, and cell number per gram are biological quantities, not universal constants. A value from a different species, preparation, or population may not be suitable. Because these factors multiply the in vitro rate, using factors that are too large inflates the scaled clearance directly. Check both the source and the units of each factor before multiplying.

Double counting. Each adjustment should occur once, and only when required by the measurement. Examples include applying an incubation-binding correction to a clearance already expressed relative to unbound concentration, or applying tissue scaling to a result that already represents a whole-liver equivalent. Dimensional analysis can reveal some duplicated steps: the product of assay clearance per unit protein, MPPGL, and liver mass already has whole-liver clearance units. Multiplying by those tissue factors again has no valid dimensional interpretation.

A Short Justification: Why Intrinsic Clearance Errors Affect Organ Clearance

Under the well-stirred model, hepatic extraction is the fraction of drug removed from blood passing through the liver. The model gives \(E_{\mathrm{h}}=f_{\mathrm{u,p}}CL_{\mathrm{int}}/(Q_{\mathrm{h}}+f_{\mathrm{u,p}}CL_{\mathrm{int}})\), and hepatic clearance is flow multiplied by extraction: \(CL_{\mathrm{h}}=Q_{\mathrm{h}}E_{\mathrm{h}}\). Substitution yields the clearance equation above. This form also explains why a biased intrinsic clearance does not necessarily produce the same proportional bias in hepatic clearance: as \(f_{\mathrm{u,p}}CL_{\mathrm{int}}\) becomes large relative to \(Q_{\mathrm{h}}\), \(CL_{\mathrm{h}}\) approaches \(Q_{\mathrm{h}}\). The result is bounded by hepatic flow, although a prediction approaching that bound can still be based on a flawed input.

Diagnostics, Comparisons, and Limits

A useful review follows the calculation in the order the quantities enter it. First identify whether the assay result is apparent or already corrected for incubation binding. Next check that its denominator is clear: per milligram of microsomal protein, per number of hepatocytes, or per amount of enzyme. Then confirm that each scaling factor matches that denominator, the species, and the intended organ. Finally, verify whether the resulting intrinsic clearance is already scaled or binding-corrected before applying any further adjustment.

A dimensional audit is especially useful. For example, \(\mathrm{\mu L/min/mg}\) multiplied by \(\mathrm{mg/g}\) and \(\mathrm{g}\) gives \(\mathrm{\mu L/min}\), a whole-liver clearance. If a calculation requires multiplying that result by another \(\mathrm{mg/g}\) factor, the units no longer describe clearance. A factor should not be retained merely because a spreadsheet has a cell for it.

The main alternative to a direct scaled prediction is to report scaled intrinsic clearance without converting it to hepatic clearance, or to use a different in vitro system with its own appropriate scaling basis. These approaches answer different questions; neither makes an inappropriate activity or scaling factor valid. A comparison across preparations can help identify whether a result depends strongly on the chosen assay system, but interpretation still requires preparation-specific inputs.

The well-stirred model also has limitations: it represents the liver as a well-mixed organ and assumes linear intrinsic elimination for the calculation. A flow-limited ceiling can compress differences between large intrinsic-clearance estimates, so similar hepatic predictions do not prove that their underlying IVIVE calculations are sound. Conversely, a high predicted clearance is a reason to inspect inputs, not proof of a particular error. Other sources of IVIVE bias require their own evidence and should not be inferred solely from an overprediction.

Common Pitfall

Wrong reasoning: “The apparent clearance is low, so I should divide it by every unbound fraction or apply every available tissue factor to be conservative.” Each adjustment can seem individually plausible, but their cumulative use may correct the same feature twice or combine factors on incompatible bases.

Correction: State what each input measures, write its units, and apply a correction only if the input has not already incorporated it. The incubation fraction applies to apparent clearance measured against total incubation concentration; plasma binding enters the hepatic model separately. Keep those roles distinct, and use a tissue scaling factor only once for the relevant assay denominator.