Anatomy of a PBPK Model

1. What Is a PBPK Model?

Physiologically based pharmacokinetic (PBPK) modeling is a mechanistic approach to pharmacokinetics in which the body is represented using physiologically meaningful compartments such as organs and tissues. Instead of describing drug disposition only with abstract kinetic compartments, a PBPK model attempts to connect drug behavior to anatomy, physiology, biochemical processes, and drug-specific properties.

A PBPK model typically represents blood as a circulating transport system and individual organs or tissues as compartments with specified volumes, blood flows, partitioning behavior, and drug elimination or metabolism processes where appropriate.

2. How Is PBPK Different From a Conventional Compartment Model?

In a conventional compartmental PK model, compartments are primarily mathematical constructs chosen to describe the observed concentration-time profile. A one-compartment model, for example, may represent the entire body with a single kinetically homogeneous compartment.

PBPK models instead assign compartments to physiological structures or groups of structures. A liver compartment can have a liver volume, liver blood flow, tissue composition, and metabolic processes. A kidney compartment can represent renal elimination. Muscle, adipose tissue, brain, skin, and other tissues can be represented according to the purpose and resolution of the model.

The distinction is not absolute. Both approaches are mathematical models, and both require assumptions. The defining feature of PBPK is the attempt to make the model structure and parameters correspond meaningfully to physiology and drug-specific mechanisms.

3. The Physiological Compartments

The first major component of a PBPK model is the representation of the body. Depending on the application, a model may contain a relatively small number of tissues or a highly resolved set of organs and tissue subcompartments.

Common physiological compartments include the liver, kidney, lung, heart, brain, muscle, adipose tissue, skin, gut, and reproductive or other specialized tissues. Blood is generally represented explicitly because it provides the transport pathway connecting many organs.

4. Blood Flow: The Circulatory Backbone

Blood flow is one of the defining physiological inputs of a PBPK model. Organs receive arterial blood, exchange drug with tissue, and return venous blood to the circulation.

For a perfusion-limited tissue model, the rate of drug delivery to a tissue is related to blood flow and the difference between incoming and outgoing concentrations. A simplified mass-balance expression can be written as:

where \(A_T\) is the amount of drug in tissue \(T\), \(Q_T\) is tissue blood flow, \(C_A\) is arterial blood concentration, \(C_{V,T}\) is the venous concentration leaving the tissue, and \(R_T\) represents net drug loss from the tissue through processes such as metabolism or other elimination mechanisms.

The exact equations used in a PBPK platform depend on the tissue model and assumptions. Nevertheless, the principle is fundamental: physiological blood flow determines how quickly drug is delivered to and removed from a tissue through the circulation.

5. Tissue Volumes

Each physiological compartment generally requires a volume. Tissue volume affects the relationship between the amount of drug in a compartment and its concentration.

For a tissue compartment:

where \(C_T\) is tissue concentration, \(A_T\) is the amount of drug in the tissue, and \(V_T\) is the tissue volume used by the model.

Physiological tissue volumes can be derived from anatomical and physiological data and may vary with body size, age, sex, species, disease state, or other characteristics depending on the application.

This is one reason PBPK models can support simulations across populations or species: physiological parameters can be modified systematically rather than treating the entire PK profile as a single empirical curve.

6. Tissue-to-Plasma Partitioning

A drug does not generally have the same concentration in every tissue. PBPK models therefore require a description of how the drug distributes between blood or plasma and tissues.

A common conceptual quantity is the tissue-to-plasma partition coefficient:

where \(C_T\) is tissue concentration and \(C_P\) is plasma concentration under the conditions used to define the partition coefficient.

Partition coefficients can be estimated or predicted using information about drug physicochemical properties and tissue composition. Depending on the model, factors such as lipophilicity, ionization, protein binding, phospholipid content, and tissue composition can influence predicted partitioning.

7. Drug-Specific Properties

The second major part of PBPK model anatomy is the drug itself. PBPK models can use experimentally measured or predicted physicochemical and biochemical properties to determine how the drug behaves within the physiological system.

One of the defining features of PBPK modeling is that these drug properties can be connected to physiology rather than being used solely as empirical descriptors of one observed concentration-time curve.

8. Absorption and Administration

A PBPK model also needs to describe how drug enters the modeled physiological system. The appropriate input depends strongly on the route of administration.

Intravenous administration

For an IV bolus, drug is introduced directly into the systemic circulation. An IV infusion instead introduces drug over a specified period according to an infusion rate.

Oral administration

For an oral dose, the model may represent dissolution, intestinal transit, absorption across the gut wall, intestinal metabolism, and subsequent delivery to the liver through the portal circulation.

Other routes

Intramuscular, subcutaneous, inhaled, transdermal, and other routes may require specialized absorption models. The more explicitly the input process is represented, the more information the model can potentially use to predict changes in exposure resulting from formulation or route changes.

9. Mass-Balance Equations

At the mathematical core of a PBPK model are mass-balance equations. For each compartment, the model accounts for drug entering, leaving, being transformed, and sometimes being stored or bound.

A generic mass balance can be expressed as:

The terms in this equation become specific to each organ. For a non-eliminating tissue, input and output may primarily reflect blood flow. For the liver, additional terms may represent metabolism. For the kidney, additional terms may represent renal elimination.

10. Clearance and Organ Elimination

Clearance in a PBPK model is often constructed from more mechanistic components than a single fitted whole-body clearance parameter. The liver and kidney are particularly important elimination organs, although other pathways may be included when scientifically relevant.

For hepatic elimination, a PBPK model may distinguish intrinsic metabolic capacity from blood-flow limitations. A conceptual relationship is:

where \(Q_H\) is hepatic blood flow, \(f_u\) is the unbound fraction, and \(CL_{\mathrm{int}}\) represents intrinsic hepatic clearance. The exact form depends on the chosen hepatic model.

For renal elimination, the model may include glomerular filtration, active secretion, reabsorption, or combinations of these processes. Again, the appropriate level of mechanistic detail depends on the drug and scientific question.

11. Protein Binding and the Unbound Fraction

Many PBPK models distinguish between total and unbound drug. The unbound fraction, often denoted \(f_u\), is the fraction of drug in plasma or blood that is not bound to proteins under the relevant conditions.

A simplified relationship is:

Unbound drug is often particularly important because many distribution, transport, and metabolic processes depend on the concentration available to interact with membranes, enzymes, or transporters.

However, \(f_u\) should not be treated as a universal multiplier applied identically to every PBPK process. Its role depends on the specific mechanistic model and the process being represented.

12. Permeability and Membrane Transport

Some tissues cannot be adequately represented by blood flow and equilibrium partitioning alone. Permeability-limited models can represent situations in which transfer across a tissue membrane or barrier is sufficiently slow to influence observed drug concentrations.

A conceptual permeability-limited exchange term can be represented as:

where \(PS\) represents an effective permeability-surface-area term and \(K_p\) represents an appropriate partitioning relationship.

The choice between perfusion-limited and permeability-limited representations is therefore a modeling decision. Highly perfused tissues with rapid equilibration may be adequately described by a simpler representation, whereas barriers or slowly equilibrating tissues may require greater detail.

13. The Gut–Liver System After Oral Dosing

Oral dosing illustrates why PBPK models can contain considerably more structure than a simple absorption compartment.

After an oral dose, drug may dissolve in the gastrointestinal tract, move through different intestinal segments, cross the intestinal wall, undergo intestinal metabolism or transport, enter the portal circulation, and then pass through the liver before reaching systemic circulation.

14. Why Physiology Makes Scaling Possible

One of the major motivations for PBPK modeling is that physiological quantities can be represented separately from drug-specific properties. This creates a framework for changing the physiological system while retaining appropriate drug characteristics.

For example, a model may use species-specific organ volumes and blood flows together with drug-specific properties such as lipophilicity, binding, and intrinsic clearance. This can support structured animal-to-human extrapolation when the model assumptions and inputs are appropriate.

The separation is useful, but it does not guarantee accurate extrapolation. Species differences in enzymes, transporters, tissue composition, disease biology, and other mechanisms can require explicit adjustments or additional data.

15. Where Does Interindividual Variability Enter?

A PBPK model can represent an individual rather than only a single typical subject. Physiological parameters can vary between individuals, and drug-specific parameters can also vary.

For example, body size may affect organ volumes and blood flows, while age or disease may alter physiology. Enzyme abundance or renal function may change drug elimination.

Conceptually, an individual PBPK model can be written as:

Population simulations can then be performed by sampling physiological characteristics or other parameters from appropriate distributions. This creates a virtual population rather than simply fitting one empirical curve to the average concentration.

16. Putting the Pieces Together

A simplified whole-body PBPK system can be viewed as a set of coupled differential equations:

Here, \(\mathbf{A}(t)\) represents the amounts of drug across compartments, while \(\mathbf{Q}\) contains blood flows, \(\mathbf{V}\) contains physiological volumes, \(\mathbf{K_p}\) contains partitioning parameters, \(\mathbf{CL}_{\mathrm{int}}\) contains intrinsic clearance quantities, \(\mathbf{P}\) represents other drug-specific physiological or biochemical parameters, and \(\mathbf{D}(t)\) describes the dosing input.

The exact mathematical form can become highly detailed. The important conceptual point is that the model is a coupled dynamical system: changing one component can affect the concentration-time behavior throughout the network.

17. What Goes Into a PBPK Model?

A useful way to understand PBPK modeling is to separate inputs into several broad categories.

The model developer must distinguish between quantities that are measured directly, quantities estimated experimentally, quantities predicted from other properties, and assumptions introduced because direct information is unavailable.

18. Which Parameters Are Measured, Predicted, or Fitted?

Not every parameter in a PBPK model comes from the same source. Some parameters are physiological constants or experimentally measured quantities, while others may be predicted or estimated from clinical data.

This distinction is important for model credibility. A PBPK model may look highly mechanistic while still containing parameters that are uncertain or empirically calibrated.

19. Worked Example: A Simplified PBPK Tissue

Consider a hypothetical tissue represented by a perfusion-limited compartment. Suppose the tissue has a volume of 10 L, receives a blood flow of 1 L/min, and has an equilibrium tissue-to-plasma partition coefficient of 4.

Step 1: Interpret the partition coefficient

At equilibrium, the model therefore represents tissue concentration as approximately four times the reference plasma concentration under the assumptions defining \(K_p\).

Step 2: Tissue concentration at a specified plasma concentration

If plasma concentration is \(C_P=2\text{ mg/L}\), the corresponding equilibrium tissue concentration would be:

Step 3: Amount of drug in the tissue

Step 4: Why blood flow matters

The tissue blood flow of \(1\text{ L/min}\) determines how rapidly drug can be delivered to and removed from the tissue through the circulation. It therefore affects the time course of equilibration, while \(K_p\) affects the equilibrium concentration relationship.

20. How Is a PBPK Model Built?

1. Define the scientific question. Determine what the model needs to explain or predict: systemic exposure, tissue concentrations, drug-drug interactions, pediatric dosing, first-in-human exposure, formulation effects, or another question.
2. Select the physiological structure. Choose the organs, tissues, circulation pathways, and level of anatomical resolution required.
3. Assemble physiological parameters. Specify organ volumes, blood flows, tissue composition, and other physiological quantities.
4. Characterize the drug. Compile physicochemical, binding, permeability, metabolic, and transporter information.
5. Specify absorption and elimination mechanisms. Represent the relevant input, metabolism, excretion, and transport processes.
6. Construct the mass-balance equations. Translate the physiological structure into a system of differential equations.
7. Parameterize and calibrate where appropriate. Use experimental or clinical observations to inform uncertain parameters without obscuring which parts of the model are mechanistic and which are empirically informed.
8. Evaluate the model. Compare predictions with appropriate observed data and investigate systematic discrepancies.
9. Use the model for simulation. Once adequately evaluated, simulate doses, populations, physiological conditions, or scenarios relevant to the scientific question.

21. Verification, Qualification, and Validation

A PBPK model should not be considered credible merely because its predictions visually resemble observed concentration-time curves. Model evaluation requires consideration of the model structure, inputs, assumptions, parameter values, and intended use.

Useful questions include:

• Are the physiological parameters internally consistent?
• Are mass balances satisfied?
• Are predicted concentrations physically and physiologically plausible?
• Does the model reproduce relevant observed PK behavior?
• Does it perform adequately across doses, studies, or populations not used to construct it?
• Are the parameters sufficiently identifiable for the intended purpose?
• Are discrepancies systematic rather than random?
• Which assumptions contribute most to predictive uncertainty?

22. Sensitivity Analysis: Which Inputs Matter?

PBPK models can contain many parameters. Not all of them have equal influence on the predicted concentration or exposure.

Sensitivity analysis asks how much a model output changes when one or more model inputs change. Conceptually, a local sensitivity can be expressed as:

where \(Y\) is a model output and \(p\) is a model parameter. A large magnitude of \(S_p\) indicates that the output is relatively sensitive to changes in that parameter near the evaluated conditions.

Sensitivity analysis can help identify which inputs deserve the greatest experimental attention and which assumptions are unlikely to materially affect the prediction in the scenario being studied.

23. Parameter Uncertainty Is Not the Same as Biological Variability

PBPK simulations can contain several different sources of uncertainty. These should not be treated as interchangeable.

Separating these sources helps prevent a common interpretation error: a wide simulated distribution does not necessarily mean that the model parameters are uncertain. The distribution may instead represent expected biological variability.

24. What Can a PBPK Model Be Used to Predict?

Once sufficiently developed and evaluated, PBPK models can be used for simulations in situations where collecting every relevant observation experimentally may be difficult, expensive, or unethical. Applications can include:

• Predicted plasma concentration-time profiles.
• Predicted tissue concentrations.
• Exposure across different doses and dosing regimens.
• Effects of changes in organ function.
• Pediatric or other special-population simulations.
• Drug-drug interaction scenarios.
• First-in-human or clinical dose-selection analyses.
• Food or formulation scenarios when appropriately modeled.
• Changes in exposure associated with altered metabolic or transporter activity.
• Mechanistic interpretation of observed PK differences.

The key advantage is not simply that PBPK models can produce more predictions. It is that the predictions can be linked to explicit physiological and mechanistic assumptions.

25. Anatomy of a PBPK Drug–Drug Interaction Model

Drug-drug interaction modeling illustrates the modular nature of PBPK. The model can represent a victim drug, a perpetrator drug, and the biological mechanism through which one drug alters the disposition of another.

For example, an inhibitor may reduce the activity of a metabolic enzyme. An inducer may increase enzyme abundance or activity. A transporter inhibitor may alter uptake or efflux.

Because the mechanism is represented explicitly, PBPK DDI models can simulate combinations of perpetrator and victim drugs under conditions that were not all directly studied experimentally. The credibility of such predictions depends on the quality of the underlying enzyme, transporter, binding, and physiological information.

26. Why PBPK Models Are Called Mechanistic

The term mechanistic does not mean that every biological mechanism has been modeled. Rather, it means that important relationships are represented using causal or physiological assumptions rather than treating the entire concentration-time curve as an empirical object.

For example, a model may connect:

Similarly:

These connections make it possible to ask mechanistic “what if?” questions. But every mechanistic relationship is still an assumption that must be supported by appropriate evidence.

27. PBPK and Population PK Are Not the Same Thing

PBPK and population PK are sometimes discussed together because both can represent variability between individuals, but they solve different modeling problems.

The approaches are not mutually exclusive. Mechanistic PBPK models and statistical population models can be combined or used sequentially depending on the scientific objective.

28. Common PBPK Modeling Mistakes

• Assuming more compartments automatically means a better model. Additional biological detail is useful only when it is supported by data and relevant to the question.
• Treating every parameter as known with certainty. Many drug-specific and physiological inputs have uncertainty.
• Confusing model calibration with validation. A model that reproduces data used during development has not necessarily demonstrated predictive performance in an independent context.
• Ignoring mass balance. A PBPK implementation should conserve mass appropriately unless the model explicitly includes formation or elimination processes.
• Using inappropriate tissue partition coefficients. Partitioning assumptions can have a major effect on predicted tissue concentrations.
• Overlooking binding and blood/plasma relationships. Incorrect handling of unbound fraction or blood-to-plasma relationships can propagate through the entire model.
• Assuming a mechanistic label guarantees mechanistic validity. A model is only as mechanistic and predictive as its assumptions, inputs, and implementation.

29. A Practical PBPK Modeling Workflow

1. Define the decision or scientific question. Specify what the model must predict or explain.
2. Define the population and scenario. Specify species, age range, disease state, body size, organ function, and other relevant characteristics.
3. Assemble physiological data. Collect organ volumes, blood flows, tissue composition, and relevant physiological relationships.
4. Characterize the compound. Compile physicochemical, binding, permeability, metabolism, and transporter information.
5. Define the model structure. Choose physiological compartments and the mechanistic processes required for the question.
6. Implement the mass balances. Translate the conceptual model into equations and computational components.
7. Check the model. Verify units, mass conservation, limiting behavior, and numerical implementation.
8. Evaluate against observations. Compare predicted and observed PK data across appropriate studies and conditions.
9. Perform sensitivity and uncertainty analyses. Identify parameters and assumptions that materially influence predictions.
10. Simulate the intended scenarios. Use the model for prediction only within a clearly defined domain of applicability.

Where to Go Next

A natural next step is to examine PBPK model construction in greater detail, beginning with physiological parameters and tissue composition and then moving into tissue partitioning, hepatic clearance, renal clearance, permeability-limited distribution, and oral absorption.

From there, PBPK concepts can be extended to population PBPK, drug-drug interaction modeling, pediatric extrapolation, first-in-human dose selection, virtual clinical trials, and model-based drug development.