Why Check the Victim Drug Before a DDI Simulation?
A victim PBPK model is the basis for predicting how a perpetrator changes victim exposure. If the model already misses the victim’s single-drug concentrations, AUC, or dose dependence, an apparently plausible interaction prediction may rest on a faulty baseline. The check here is deliberately specific: compare the victim model with single-drug observations across the doses and routes that are available, using dose-normalized profiles and clearance diagnostics to identify patterns of mismatch.
The central expectation is conditional, not universal. When pharmacokinetics are linear and time-invariant over the tested range, the same dose and route produce profiles that scale in proportion to dose. Dividing concentrations and exposure by dose should then make results comparable. A failure to overlap is informative only after checking whether the assumptions for that comparison are reasonable.
When Dose Normalization Is Justified
For a linear, time-invariant disposition model with unchanged physiology and parameters, concentration responds proportionally to the input. In a one-compartment illustration, the amount \(A\) follows \(dA/dt=-kA\) after an intravenous bolus, with \(A(0)=D\), and concentration is \(C=A/V\). The solution is \(C(t)=D e^{-kt}/V\). Thus \(C(t)/D=e^{-kt}/V\), which does not depend on dose. More generally, a linear PBPK system with the same route-specific input shape has the same scaling property: multiplying the dose by a constant multiplies concentrations by that constant, if model parameters and system conditions remain unchanged.
Under those conditions, AUC, \(C_{\max}\), and concentrations at corresponding times are proportional to dose. Accordingly, AUC/D and \(C_{\max}/D\) should be similar across doses, within the variability and precision of the observations. For a repeated-dose comparison, the dosing interval and regimen must also match; compare steady-state profiles at the same point in the dosing interval and use AUC over that interval when appropriate.
The comparison also requires consistent concentration definitions and units, an appropriate time origin, and compatible formulations and administration conditions. A bolus, an infusion, and an oral dose do not have the same input shape. Dose normalization does not make those profiles interchangeable. Instead, compare each route with its corresponding model prediction, and use cross-route measures only when their assumptions are met.
Dose proportionality may fail when absorption, distribution, metabolism, or elimination is nonlinear. Examples include saturable intestinal absorption, saturable metabolism, concentration-dependent protein binding that affects the measured concentration, or dose-dependent formulation behavior. Time-varying enzyme activity or physiology can also invalidate a comparison across occasions. In those situations, a difference after normalization may be real and should not automatically be “corrected” by tuning a linear model.
Worked Example: Checking Intravenous Profiles at Two Doses
Suppose a hypothetical victim drug is given as an intravenous bolus at 100 mg and 200 mg. At 1, 4, and 8 hours, observed concentrations are respectively 4.0, 2.0, and 1.0 mg/L at 100 mg, and 8.1, 3.9, and 2.1 mg/L at 200 mg. The PBPK model predicts 4.0, 2.0, and 1.0 mg/L at 100 mg and exactly twice those values at 200 mg.
At 100 mg, the dose-normalized observations are \(4.0/100=0.0400\), \(2.0/100=0.0200\), and \(1.0/100=0.0100\) (mg/L)/mg. At 200 mg, they are \(8.1/200=0.0405\), \(3.9/200=0.0195\), and \(2.1/200=0.0105\) (mg/L)/mg. The corresponding model values are 0.0400, 0.0200, and 0.0100 (mg/L)/mg at both doses.
These hypothetical observations are close to dose proportional and to the model, with small deviations that should be interpreted in light of measurement and sampling variability. They do not show a systematic increase or decrease in normalized concentrations across time or dose. If the profiles had systematically separated, the next step would be to examine exposure metrics and the assumptions behind dose proportionality, rather than to infer a particular mechanism from three time points.
Use Exposure and Clearance as Complementary Checks
For complete intravenous input, systemic clearance can be calculated from dose and total exposure:
This expression is appropriate when the administered dose reaches the systemic circulation and AUC is measured or estimated over the full concentration-time curve. For linear pharmacokinetics, clearance is dose-independent, so IV dose/AUC should agree across doses. If pharmacokinetics are nonlinear, dose/AUC at a particular dose is a dose-specific exposure ratio; it should not automatically be treated as a constant intrinsic clearance.
For an oral dose, the corresponding ratio is \(D_{\mathrm{oral}}/AUC_{\mathrm{oral}}\), often called apparent clearance, \(CL/F\), under linear conditions. Oral AUC depends on both systemic clearance and the fraction of dose reaching systemic circulation, \(F\). Therefore, a discrepancy in oral AUC alone cannot distinguish a clearance error from a bioavailability error. If dose-normalized oral and IV AUCs are available under linear, dose-independent conditions, absolute bioavailability can be estimated as:
This cross-route calculation assumes comparable systemic clearance between occasions and complete IV input. It summarizes the fraction reaching systemic circulation; it does not, on its own, identify whether a difference arises from absorption, intestinal first-pass loss, or hepatic first-pass loss.
Worked Example: Interpreting an IV Clearance Mismatch
In a hypothetical study, a 100 mg IV dose gives an observed \(AUC_{0-\infty}\) of 20.0 mg·h/L. The PBPK model predicts 25.0 mg·h/L for the same dose. Assume complete IV input and adequate AUC coverage.
The observed clearance estimate is \(100\text{ mg}/20.0\text{ mg·h/L}=5.00\text{ L/h}\). The model-implied clearance is \(100\text{ mg}/25.0\text{ mg·h/L}=4.00\text{ L/h}\). The model therefore predicts 25% greater AUC than observed and, for this IV comparison, 20% lower clearance than the estimate from observed data: \((5.00-4.00)/5.00=0.200\).
This is evidence of a baseline disposition mismatch for the specified conditions, not proof that one particular clearance pathway is wrong. The next checks include the consistency of the AUC window, concentration units, dose, sampling, and the model’s route and physiology assumptions. If the observation and prediction remain comparable, the model’s systemic disposition requires review before using it as a reliable victim baseline.
Compare Routes Without Confounding the Diagnosis
Across routes, compare observed and predicted profiles within each route first. A model may fit an IV profile but miss an oral profile because the systemic disposition is adequate while the oral input or presystemic availability is not. Conversely, mismatches across both routes may point toward shared systemic disposition, although data quality and parameter compensation must still be considered. Route-specific patterns guide investigation; they do not uniquely identify a mechanism.
Worked Example: Separating an Oral Exposure Mismatch from IV Clearance
Consider hypothetical linear, dose-proportional data for a 100 mg dose. Observed IV AUC is 20.0 mg·h/L and observed oral AUC is 8.00 mg·h/L. The model predicts IV AUC of 20.0 mg·h/L and oral AUC of 12.0 mg·h/L. Assume both AUCs represent complete exposure and IV input is complete.
Observed IV clearance is \(100/20.0=5.00\) L/h, matching the model’s IV dose/AUC ratio. The observed bioavailability estimate is \((8.00/100)/(20.0/100)=0.400\). The model-implied value is \((12.0/100)/(20.0/100)=0.600\). Thus the IV exposure check supports the model’s systemic clearance under these assumptions, while its oral prediction implies greater systemic availability than the observations.
The justified conclusion is an oral-route baseline mismatch, not a definitive diagnosis of absorption or first-pass metabolism. Relevant checks include the oral input representation, formulation, and evidence constraining presystemic availability. Since IV and oral AUCs are hypothetical and the calculation assumes linearity and comparable clearance, the estimated fractions are illustrative rather than claims about a real drug.
A Practical Diagnostic Sequence
A disciplined comparison reduces the risk of compensating for a baseline error by changing a parameter that does not explain the observed pattern. The sequence below keeps dose proportionality, route, and exposure interpretation separate.
Check dose, route, formulation, administration conditions, sampling times, concentration units, and whether the data are single-dose or steady-state.
Within each route and regimen, compare \(C(t)/D\), AUC/D, and \(C_{\max}/D\) across doses. Look for a systematic pattern rather than demanding exact equality from noisy observations.
Inspect profiles and exposure metrics at each dose. Determine whether mismatch is shared across routes or concentrated in one route or dose range.
For IV data, compare dose/AUC across doses and with the model. For oral data, interpret dose/AUC as apparent clearance unless bioavailability is known or separately estimated.
Check data handling, AUC coverage, model inputs, and plausible nonlinearity. Revise a mechanism only when the route- and dose-specific evidence supports that revision.
Diagnostics, Limitations, and Common Pitfalls
Useful diagnostics include normalized profile plots on common axes, AUC/D and \(C_{\max}/D\) by dose, and IV dose/AUC by dose where IV data exist. A consistent departure with dose can flag nonlinearity or dose-dependent input. A persistent model-observation difference at all doses may instead indicate a baseline scale or disposition mismatch. A profile that agrees early but diverges late can signal a time-course or terminal-phase problem; however, a terminal slope alone is not a direct clearance measurement, because distribution and the terminal observation window also affect it.
These checks have limits. Sparse sampling can make peak concentrations or AUC estimates unreliable. An AUC extrapolated far beyond the last sample is sensitive to the terminal-phase estimate. Between-occasion differences can reflect changing physiology, adherence, formulation, or measurement conditions rather than model structure. Dose-normalized values can also appear similar despite compensating errors in model components. The checks identify whether the baseline prediction is credible over the observed conditions; they do not guarantee that every PBPK mechanism is uniquely identified.
Common mistake: “The oral AUC is too high, so the model must have underestimated systemic clearance.” This reasoning treats oral AUC as if it depended only on clearance. The correction is to recognize that oral exposure reflects both clearance and bioavailability. Compare with IV exposure when suitable data exist; otherwise, describe the mismatch as an oral exposure or apparent-clearance discrepancy, not a proven systemic-clearance error.
A second mistake is to divide every concentration by dose and expect identical profiles regardless of route or dose. Normalization is justified as a linearity check only when the relevant input and pharmacokinetic conditions are comparable. If the data reveal dose dependence, report it and investigate it rather than forcing a dose-proportional interpretation.
Check Your Understanding
Answer each question using the stated assumptions. Brief solutions follow the questions.
- Under what conditions should \(C(t)/D\) be similar at two doses given by the same route?
- A hypothetical IV dose of 80 mg gives \(AUC_{0-\infty}=16\) mg·h/L. What is the estimated systemic clearance?
- Why does an oral dose/AUC ratio not, by itself, establish systemic clearance?
- Observed dose-normalized oral AUC is lower at a higher dose. Name two broad explanations that should be considered before changing the PBPK model.
Solutions. 1. The pharmacokinetics must be linear and time-invariant over the dose range, with comparable route-specific input, regimen, physiology, and measurement conditions. 2. \(CL=80/16=5.00\) L/h, assuming complete IV input and adequate AUC coverage. 3. Oral AUC depends on both systemic clearance and the fraction reaching systemic circulation, so the ratio is apparent clearance \(CL/F\) under linear conditions. 4. Possible explanations include dose-dependent pharmacokinetics, such as saturable absorption or metabolism, and differences in formulation, administration, or study conditions; the pattern should be checked before attributing it to a model parameter error.