Bridging In Vitro Potency to In Vivo Response

Why a Potency Measurement Needs a Bridge

An in vitro potency value is not automatically the concentration that will produce a particular response in vivo. The reported assay concentration may include drug bound to proteins or other assay components, while the relevant in vivo concentration may be the unbound concentration at the site of action. Even after both are expressed as free concentrations, the two settings can differ in target availability, pathway amplification, assay conditions, and other biological features.

A useful bridge makes these differences explicit. It first puts in vitro potency and in vivo exposure on compatible concentration scales, then uses a mechanistic concentration-response model to relate them. A fitted context factor summarizes the potency shift that remains after the concentration corrections. It is a calibration parameter, not proof of a particular biological cause.

Define the Concentration-Response Link

Definition: Let \(q\) be the response normalized between a baseline \(E_0\) and a maximum \(E_{\max}\): \(q=(E-E_0)/(E_{\max}-E_0)\). Let \(C_{u,\mathrm{site}}\) be unbound drug concentration at the in vivo site of action, and let \(K_{50,\mathrm{vivo}}\) be the concentration at which the normalized response is one-half. For a Hill coefficient of 1, the saturable response model is:

q=\frac{C_{u,\mathrm{site}}}{K_{50,\mathrm{vivo}}+C_{u,\mathrm{site}}}

Here, \(K_{50,\mathrm{vivo}}\) is a functional potency parameter for the specified in vivo response and conditions. It is not necessarily a binding affinity such as a dissociation constant. More generally, a Hill model uses \(q=C_{u,\mathrm{site}}^h/(K_{50,\mathrm{vivo}}^h+C_{u,\mathrm{site}}^h)\), where \(h\) is the Hill coefficient. The examples below use \(h=1\), either as a justified model assumption or as a specified value.

Suppose an in vitro functional assay gives a nominal half-maximal concentration \(P_{50,\mathrm{nom}}\), and the unbound fraction in that assay is \(f_{u,\mathrm{assay}}\). If binding is linear over the tested range and equilibrium has been reached, then the unbound in vitro half-maximal concentration is:

K_{50,\mathrm{vitro},u}=f_{u,\mathrm{assay}}P_{50,\mathrm{nom}}

This conversion assumes the assay’s nominal concentration is the total concentration to which the measured unbound fraction applies. If binding is concentration-dependent, the fraction must be evaluated across the range rather than treated as constant. The conversion also does not turn every assay potency into a target-binding constant: functional EC50 or IC50 values can depend on the assay system and its conditions.

The bridge relates the in vitro free potency to the in vivo free potency through a dimensionless context factor \(\gamma\):

K_{50,\mathrm{vivo}}=\gamma K_{50,\mathrm{vitro},u}

A value \(\gamma>1\) indicates that a higher unbound concentration is required in vivo than the in vitro free potency would suggest. A value \(\gamma<1\) indicates a lower concentration is required. Neither direction alone identifies the mechanism behind the shift.

Put In Vivo Exposure on the Relevant Scale

The response model requires concentration at the site of action, not automatically the administered dose or total plasma concentration. If site concentration is not measured directly, one possible relationship is:

C_{u,\mathrm{site}}=K_{p,uu} f_{u,p} C_{p,\mathrm{total}}

Here \(C_{p,\mathrm{total}}\) is total plasma concentration, \(f_{u,p}\) is the unbound fraction in plasma, and \(K_{p,uu}\) is the ratio of unbound site concentration to unbound plasma concentration under the relevant conditions. This relationship is useful only when the quantities are appropriately defined and the assumed distribution relationship applies. If unbound site concentrations are already available, use them directly rather than applying this conversion again.

Concentration units must agree throughout the bridge. For example, a potency in micromoles per litre cannot be compared directly with an exposure in nanomoles per litre without conversion. The earlier material on harmonizing units and dose definitions is relevant when the underlying records use different conventions.

Deriving the Potency Estimate From a Response

For the Hill-1 model, the normalized response can be rearranged to estimate the in vivo half-maximal concentration from one response observation. Starting from \(q=C_{u,\mathrm{site}}/(K_{50,\mathrm{vivo}}+C_{u,\mathrm{site}})\), multiply both sides by the denominator and collect the terms involving \(K_{50,\mathrm{vivo}}\):

q(K_{50,\mathrm{vivo}}+C_{u,\mathrm{site}})=C_{u,\mathrm{site}}

qK_{50,\mathrm{vivo}}=C_{u,\mathrm{site}}(1-q) \qquad\Longrightarrow\qquad K_{50,\mathrm{vivo}}=C_{u,\mathrm{site}}\frac{1-q}{q}

This calculation requires \(0<q<1\), a response normalized to the appropriate baseline and maximum, and an applicable Hill-1 model. It is sensitive to uncertainty in \(q\), especially near 0 or 1. With a known Hill coefficient \(h\), the corresponding expression is \(K_{50,\mathrm{vivo}}=C_{u,\mathrm{site}}((1-q)/q)^{1/h}\). If \(h\) is unknown, a single response observation cannot determine both \(h\) and \(K_{50,\mathrm{vivo}}\).

Worked Example: Correcting an In Vitro Potency for Assay Binding

In a hypothetical functional assay, the nominal half-maximal concentration is \(0.800\ \mu\mathrm{M}\). The unbound fraction in the assay medium is estimated to be \(0.250\), constant over the relevant range. Calculate the in vitro free potency.

Under the stated linear-binding and equilibrium assumptions:

K_{50,\mathrm{vitro},u} =f_{u,\mathrm{assay}}P_{50,\mathrm{nom}} =(0.250)(0.800\ \mu\mathrm{M}) =0.200\ \mu\mathrm{M}

The free potency is therefore \(0.200\ \mu\mathrm{M}\). It is lower than the nominal value because only one quarter of the nominal concentration is unbound under the assumed assay conditions. This correction changes the concentration scale; it does not establish that the in vivo half-maximal concentration is also \(0.200\ \mu\mathrm{M}\).

Worked Example: Estimating the In Vivo Context Factor

Suppose the corrected in vitro free potency is \(0.200\ \mu\mathrm{M}\). In a hypothetical in vivo experiment, the unbound site concentration is \(0.600\ \mu\mathrm{M}\), and the normalized response is \(q=0.667\). Assume the response follows the Hill-1 model, with baseline and maximum already defined for this endpoint. Estimate \(K_{50,\mathrm{vivo}}\) and \(\gamma\).

First rearrange the Hill-1 response equation:

K_{50,\mathrm{vivo}} =C_{u,\mathrm{site}}\frac{1-q}{q} =(0.600\ \mu\mathrm{M})\frac{1-0.667}{0.667}

Using the stated rounded response, \(1-0.667=0.333\), so:

K_{50,\mathrm{vivo}} =\frac{0.600(0.333)}{0.667}\ \mu\mathrm{M} =\frac{0.1998}{0.667}\ \mu\mathrm{M} \approx 0.2996\ \mu\mathrm{M}

The context factor is the ratio of in vivo to in vitro free potency:

\gamma=\frac{K_{50,\mathrm{vivo}}}{K_{50,\mathrm{vitro},u}} \approx\frac{0.29955}{0.200} \approx 1.498

Thus, in this hypothetical example, the in vivo half-maximal concentration is about \(1.498\) times the corrected in vitro free potency. The interpretation is conditional on the model and the concentration correction; it does not show that any particular biological process caused the shift.

Calibrating With Multiple In Vivo Observations

A single response point can be transformed into a potency estimate, but a collection of exposure-response observations provides a more informative check of the bridge. Rather than independently transforming every noisy response and treating those estimates as exact, calibration can compare observed responses with predictions from the shared model. For a fixed in vitro free potency and Hill coefficient, the predicted in vivo response at observation \(i\) is:

q_i^{\mathrm{pred}} =\frac{C_{u,\mathrm{site},i}} {\gamma K_{50,\mathrm{vitro},u}+C_{u,\mathrm{site},i}}

The calibration estimates \(\gamma\) so that predictions agree with the in vivo data, while the in vitro potency is informed by the assay data and its uncertainty. A joint calibration objective can include the assay and in vivo observations with their measurement-error models; it should not count a derived potency estimate as an additional independent observation if it was calculated from data already included. Assay-specific error and dependence among replicates matter when specifying those contributions.

Worked Example: Checking a Bridge Against Three Exposure Levels

Consider a hypothetical assay with nominal half-maximal potency \(0.800\ \mu\mathrm{M}\) and assay unbound fraction \(0.250\). Hypothetical in vivo normalized responses are \(0.334\), \(0.667\), and \(0.750\) at unbound site concentrations \(0.150\), \(0.600\), and \(0.900\ \mu\mathrm{M}\), respectively. Assume the same endpoint normalization and Hill-1 response model apply at all three concentrations.

The in vitro free potency is \(0.250(0.800)=0.200\ \mu\mathrm{M}\). For an initial check, transform each in vivo point using \(K_{50,\mathrm{vivo}}=C_{u,\mathrm{site}}(1-q)/q\). For the first point:

K_{50,\mathrm{vivo},1} =(0.150)\frac{1-0.334}{0.334} =\frac{0.0999}{0.334}\ \mu\mathrm{M} \approx 0.2991\ \mu\mathrm{M}

For the second point, the calculation from the previous example gives approximately \(0.2996\ \mu\mathrm{M}\). For the third:

K_{50,\mathrm{vivo},3} =(0.900)\frac{1-0.750}{0.750} =(0.900)\frac{0.250}{0.750}\ \mu\mathrm{M} =0.3000\ \mu\mathrm{M}

These estimates are close, as expected for responses consistent with a shared Hill-1 potency, allowing for the rounding of the reported responses. A representative value of \(0.2996\ \mu\mathrm{M}\) gives \(\gamma=0.2996/0.200=1.498\). Check the predicted responses with that value, so that \(K_{50,\mathrm{vivo}}=0.2996\ \mu\mathrm{M}\):

q^{\mathrm{pred}}(0.150)=\frac{0.150}{0.2996+0.150}\approx0.3336,\quad q^{\mathrm{pred}}(0.600)=\frac{0.600}{0.2996+0.600}\approx0.6670

q^{\mathrm{pred}}(0.900)=\frac{0.900}{0.2996+0.900}\approx0.7503

The predictions are close to the three hypothetical observations. In a real calibration, uncertainty in the responses, concentrations, and assay potency would be represented in the objective rather than ignored; rounded pointwise estimates are a diagnostic, not a substitute for that fit.

Assumptions, Diagnostics, and Alternatives

The bridge is most defensible when the following conditions are considered explicitly:

The in vitro and in vivo concentration units and definitions are harmonized, and assay and plasma binding corrections correspond to the relevant conditions.
The unbound fraction and any site-to-plasma relationship are sufficiently stable over the concentration range, or their concentration dependence is modeled.
The in vitro and in vivo endpoints represent meaningfully comparable pharmacology, with baseline and maximum defined consistently for the response being normalized.
The response is adequately described by the chosen concentration-response function, and concentrations correspond to the time or exposure metric appropriate to the response.
The observations span enough of the response range to inform potency. Data only far below or far above the midpoint may weakly constrain \(K_{50,\mathrm{vivo}}\).

Useful checks include plotting observed normalized response against unbound site concentration alongside the fitted curve, examining whether residuals show a systematic trend, and checking whether different exposure levels imply compatible potency estimates. If an effect follows concentration with a delay, a static concentration-response curve may show a loop or time-ordered residual pattern; an effect-compartment or dynamic model may be needed. A poor fit can also reflect a different Hill slope, an incorrect maximum, an active metabolite, or an inaccurate free-concentration conversion.

Approach	What it does	Main caution
Compare nominal in vitro potency directly with total plasma exposure	Uses the reported values without free-concentration corrections	Can compare unlike concentration scales and misstate the exposure relevant to the target
Correct to free concentrations and estimate a context factor	Separates an explicit binding correction from the remaining potency shift	Depends on valid free-fraction, site-exposure, endpoint, and response-model assumptions
Fit separate in vitro and in vivo curves without a shared bridge	Allows each setting its own potency and curve shape	Describes both data sources but does not quantitatively connect them

The free-concentration bridge is therefore neither an automatic conversion nor a guarantee of mechanistic transportability. It is a transparent way to state what is measured, what is corrected, and what remaining shift must be calibrated. If the context factor changes substantially with dose, time, or endpoint, a single constant \(\gamma\) is not an adequate bridge.

Common Mistakes

Using total concentrations as though they were free. Correct the assay and in vivo concentration scales when the relevant unbound fractions are available and applicable.
Calling the context factor a causal mechanism. \(\gamma\) summarizes a potency shift; it does not by itself identify whether that shift comes from biology, assay conditions, or an exposure conversion.
Treating an assay EC50 or IC50 as a binding constant. A functional potency depends on the assay system and conditions and need not equal a molecular binding affinity.
Using a single response point to estimate both potency and slope. One point can estimate \(K_{50,\mathrm{vivo}}\) only when the response model and Hill coefficient are specified.
Ignoring response normalization. The value \(q\) must be calculated from the relevant baseline and maximum; an unnormalized endpoint is not automatically a fractional response.

Key takeaway: Correct in vitro potency and in vivo exposure to compatible free-concentration scales, define a response model, and interpret the remaining context factor as a conditional calibration—not as a standalone biological explanation.

Check Your Understanding

Answer each question, then compare with the solutions below.

A hypothetical assay has nominal potency \(0.500\ \mu\mathrm{M}\) and a constant assay unbound fraction of \(0.400\). What is the in vitro free potency?
Under the Hill-1 model, what \(K_{50,\mathrm{vivo}}\) corresponds to \(q=0.600\) at \(C_{u,\mathrm{site}}=0.450\ \mu\mathrm{M}\)?
If the corrected in vitro free potency is \(0.150\ \mu\mathrm{M}\) and the in vivo potency is \(0.300\ \mu\mathrm{M}\), calculate \(\gamma\) and interpret it.
Why is one response observation insufficient to estimate both \(K_{50,\mathrm{vivo}}\) and an unknown Hill coefficient?
Name one reason why a bridge based on unbound plasma concentration might fail to represent concentration at the site of action.

Solutions. (1) \(K_{50,\mathrm{vitro},u}=(0.400)(0.500\ \mu\mathrm{M})=0.200\ \mu\mathrm{M}\). (2) \(K_{50,\mathrm{vivo}}=(0.450)(1-0.600)/0.600=0.300\ \mu\mathrm{M}\). (3) \(\gamma=0.300/0.150=2.00\); under the stated model, in vivo potency is twice the in vitro free potency, meaning a higher free concentration is required in vivo. This ratio does not identify the cause. (4) The general response equation contains both unknowns, so one equation from one observation cannot uniquely determine two parameters. (5) For example, unbound distribution into the site may differ from unbound plasma exposure, or the response may be delayed relative to plasma concentration.