Connecting Predicted Exposure to Mechanism
A human PBPK model can predict how much drug reaches a tissue over time. A mechanistic pharmacodynamic (PD) model describes how that exposure changes a biological response. Integrating the two means using the PBPK concentration-time profile as the input to the PD model—not treating a predicted human concentration as though it were itself a predicted clinical effect.
The connection is only meaningful if the concentration supplied to the PD model represents the exposure that drives the mechanism. Depending on the drug and target, that may be unbound plasma concentration, unbound tissue concentration, or another justified measure. A total plasma concentration should not be substituted automatically for an unbound target-site concentration.
For a stimulatory effect on response production, one useful turnover model is
Here \(R_H(t)\) is the human response, \(C_{u,T,H}(t)\) is the human unbound concentration at the target site predicted by PBPK, \(k_{\mathrm{in},H}\) is the response-production rate, and \(k_{\mathrm{out},H}\) is the response-loss rate. \(E_{\max,H}\) is the maximum fractional stimulation of production, \(EC_{50,H}\) is the target-site concentration producing half of that maximum, and \(\gamma_H\) controls the concentration-response curve’s steepness. The subscript \(H\) denotes a human parameter or quantity. This is a model choice, not a universal description of drug action: inhibition of production or a different causal pathway requires a correspondingly different PD equation.
With no drug, the model’s steady-state response is \(R_{0,H}=k_{\mathrm{in},H}/k_{\mathrm{out},H}\). Thus, baseline response and turnover rates must be compatible. For a response measured in units, \(k_{\mathrm{in},H}\) has units of response per time and \(k_{\mathrm{out},H}\) has units of inverse time. The concentration and \(EC_{50,H}\) must use the same concentration units.
Assigning Animal and Human Inputs
The integrated model does not imply that every parameter can be copied from an animal. It helps to separate the inputs by their biological role and the evidence needed to support them. Animal observations can inform drug disposition and PD behavior under the measured animal conditions. They do not, by themselves, establish human physiology, human baseline response, or conserved target sensitivity.
| Input or parameter | Potential evidence | Translation check |
|---|---|---|
| Human \(C_{u,T,H}(t)\) | Human PBPK model, with animal PK and drug-specific disposition information contributing where appropriate | Check tissue, binding, and concentration units; verify the exposure measure matches the PD driver |
| Drug-specific \(E_{\max}\), \(EC_{50}\), and \(\gamma\) | Animal concentration-response data and relevant in vitro or target data | Transport only when target, assay context, and exposure scale are comparable; adjust for supported species differences |
| Human \(k_{\mathrm{in},H}\), \(k_{\mathrm{out},H}\), and baseline | Human physiology or clinical biomarker information, potentially informed by cross-species evidence | Ensure the baseline is consistent with the turnover rates and the human response scale |
| Human target or disease-state differences | Human-specific biology, clinical data, or validated translational evidence | Do not assume animal baseline or target sensitivity is identical without justification |
The table describes possible sources of information, not automatic transfer rules. For example, animal data may estimate an \(EC_{50}\) in unbound tissue concentration, but applying that value to a human PBPK profile still assumes that the concentration scale at the target is comparable. If binding, target abundance, or affinity differs by species, the PD parameter may require a supported human adjustment. Likewise, a human PBPK model needs human physiological inputs; its output is not made human-specific merely by entering an animal dose.
Why Turnover Dynamics Matter
The PD equation makes the response depend on both concentration and the rate at which the system changes. Under a constant target-site concentration, define \(S=E_{\max,H}C_{u,T,H}^{\gamma_H}/(EC_{50,H}^{\gamma_H}+C_{u,T,H}^{\gamma_H})\). The response equation becomes \(dR_H/dt=k_{\mathrm{in},H}(1+S)-k_{\mathrm{out},H}R_H\). Its steady state and time course follow directly.
To justify the time-course expression, subtract \(R_{\mathrm{ss}}\) from both sides of the differential equation. Because \(k_{\mathrm{in},H}(1+S)=k_{\mathrm{out},H}R_{\mathrm{ss}}\), the difference \(D(t)=R_H(t)-R_{\mathrm{ss}}\) satisfies \(dD/dt=-k_{\mathrm{out},H}D\). Separating variables and integrating gives \(D(t)=D(0)e^{-k_{\mathrm{out},H}t}\), which yields the stated solution. The characteristic response time is \(1/k_{\mathrm{out},H}\); a response therefore need not reach its new steady state during a finite exposure.
Worked Example: From Human PBPK Concentration to Steady-State Response
Suppose a hypothetical human PBPK simulation predicts a constant unbound target-site concentration of \(2.0\) mg/L. A mechanistic PD model uses \(E_{\max,H}=0.80\), \(EC_{50,H}=1.5\) mg/L, and \(\gamma_H=1\). Human baseline response is \(500\) response units, with \(k_{\mathrm{out},H}=0.20\) h\(^{-1}\). Assume the model’s stimulatory-production structure is appropriate and \(k_{\mathrm{in},H}\) is consistent with that baseline.
First calculate the fractional stimulation, using the model’s concentration-response term:
The concentration ratio is dimensionless because both concentrations are in mg/L. The drug therefore increases response production by 45.71% under this model. Baseline consistency requires \(k_{\mathrm{in},H}=k_{\mathrm{out},H}R_{0,H} =0.20\ \mathrm{h}^{-1}\times500\ \text{units}=100\ \text{units/h}\). The predicted steady-state response is
This is a steady-state prediction conditional on the stated concentration and model. It is not the response expected immediately after dosing; the time required to approach it depends on \(k_{\mathrm{out},H}\).
Worked Example: A Finite Exposure Does Not Produce the Steady-State Effect Immediately
Use a hypothetical model with baseline response \(400\) units, \(k_{\mathrm{out},H}=0.25\) h\(^{-1}\), \(E_{\max,H}=0.75\), \(EC_{50,H}=1.0\) mg/L, and \(\gamma_H=1\). Suppose PBPK predicts a constant target-site concentration of \(2.0\) mg/L for four hours. Assume the system begins at baseline and the turnover model is appropriate.
The production stimulation is \(S=0.75(2.0)/(1.0+2.0)=0.50\). The implied production rate is \(k_{\mathrm{in},H}=0.25\times400=100\) units/h, and the exposed steady-state response is \(R_{\mathrm{ss}}=100(1+0.50)/0.25=600\) units. Apply the constant-concentration solution with \(R_H(0)=400\) units:
The response after four hours is \(526.4\) units, not \(600\) units. The latter is the asymptote if the concentration is maintained indefinitely. The assumptions for this calculation are a constant concentration over the four hours, a constant model stimulation during that period, and an initial response equal to the untreated baseline.
Handling Species Differences Without Losing the Mechanism
A transported parameter should be stated as an assumption supported by evidence, not hidden inside a calculation. The following example shows how a human-specific potency adjustment changes the projection while leaving the PBPK concentration profile as the exposure input.
Worked Example: Adjusting Animal-Informed Potency for Human Target Affinity
Suppose hypothetical animal data support \(EC_{50,A}=0.80\) mg/L when concentration is measured as unbound target-site concentration. Assume other evidence supports a twofold higher human \(EC_{50}\), reflecting lower human target potency, and supports transporting \(E_{\max}=0.90\) and \(\gamma=1\). A human PBPK prediction gives \(C_{u,T,H}=1.2\) mg/L at the time of interest. Assume the species adjustment and common concentration scale are valid for this illustration.
The adjusted human potency parameter is \(EC_{50,H}=2(0.80)=1.60\) mg/L. The predicted fractional stimulation is
If the animal \(EC_{50,A}\) were transferred without adjustment, the calculation would instead be \(0.90[1.2/(0.80+1.2)]=0.5400\). The unadjusted calculation predicts greater stimulation because it assumes greater potency in humans than the stipulated human-specific evidence supports. The \(0.3857\) result is still conditional on the human PBPK concentration, the affinity-based adjustment, and transport of \(E_{\max}\) and \(\gamma\); it is not a universal rule that affinity differences alone determine clinical potency.
Diagnostics, Alternatives, and Limitations
Before interpreting a projection, check that each connection in the model is coherent. Useful checks include:
- Exposure driver: Confirm the PD model’s concentration measure matches the PBPK output. Mixing total plasma and unbound target-site concentrations can distort the apparent potency.
- Units and scale: \(C_{u,T,H}\) and \(EC_{50,H}\) must have identical concentration units. Response, baseline, \(k_{\mathrm{in},H}\), and \(k_{\mathrm{out},H}\) must also have compatible units.
- Baseline consistency: Verify \(R_{0,H}=k_{\mathrm{in},H}/k_{\mathrm{out},H}\) when the model assumes an untreated steady state. A mismatch can create a predicted drug effect that is actually a baseline error.
- Time-course plausibility: Compare the predicted response timing with the turnover rate and exposure duration. A steady-state calculation is not appropriate when the exposure is too brief for the response to approach steady state.
- Mechanism and transfer: Check that the assumed production effect fits the biology and that animal-informed potency or efficacy is supported for the human target and concentration scale.
A common mistake is to insert an animal dose into a human PD equation and call the result a human response prediction. Dose is not the PD driver in this integrated model: the human PBPK-predicted target-site concentration is. Another mistake is to copy all animal PD parameters unchanged. The correction is to identify which parameters are supported as transportable and supply human-specific physiology, baseline, and justified species adjustments where needed.
Compared with a direct dose-response or empirical exposure-response curve, mechanistic PBPK-PD integration represents intermediate processes: human disposition produces a concentration profile, and that profile drives a specified biological mechanism. This can make assumptions visible and support projection to exposure profiles not directly observed in humans. Simpler empirical models may be preferable when the mechanistic parameters cannot be identified or the mechanism is poorly understood. The integrated approach is limited by its structural assumptions, the quality of each input, and the degree to which animal biology represents human biology. A detailed model does not guarantee a reliable projection if its concentration driver or species-transfer assumptions are wrong.
Check Your Understanding
Answer each question, then compare with the short solutions below.
- A human PBPK model reports total plasma concentration, while the PD \(EC_{50}\) is defined for unbound target-site concentration. What must be resolved before linking the models?
- If \(k_{\mathrm{in},H}=60\) response units/h and \(k_{\mathrm{out},H}=0.15\) h\(^{-1}\), what is the untreated steady-state response?
- For \(E_{\max}=0.60\), \(EC_{50}=3.0\) mg/L, \(\gamma=1\), and \(C_{u,T,H}=3.0\) mg/L, calculate the fractional stimulation.
- Why can a four-hour response differ from the steady-state response even when concentration is constant?
- Name one reason not to transfer an animal \(EC_{50}\) directly to humans.
Solutions. (1) Obtain or justify the unbound target-site concentration profile, or otherwise establish a valid mapping to the PD concentration scale. (2) \(R_{0,H}=60/0.15=400\) response units. (3) \(S=0.60(3.0)/(3.0+3.0)=0.300\), or 30.0% stimulation. (4) The response approaches its new equilibrium at a rate governed by \(k_{\mathrm{out},H}\), so finite exposure may end before equilibrium is reached. (5) Human and animal target affinity, binding, target abundance, or relevant concentration scales may differ; a direct transfer requires evidence that the difference is negligible.