Tutorials › Pharmacometrics › Turnover Models in Pharmacodynamics
Pharmacodynamics · PK/PD Modeling

Turnover Models in Pharmacodynamics

Learn how turnover models describe pharmacodynamic responses that are continuously produced and removed, and how drug effects on production or loss create delayed, sustained, or indirect concentration-response relationships.

Intermediate PK/PD Modeling Pharmacodynamics Indirect Response
01 · The big picture

1. What Is a Turnover Model?

A turnover model describes a biological quantity that is continuously produced and removed. The quantity might be a biomarker, physiological mediator, cell population, or other measurable response variable.

The defining feature is that the response is not simply created by an instantaneous drug concentration-response relationship. Instead, the response follows its own biological dynamics.

Production Response biomarker physiological quantity Loss The response changes according to the balance between production and loss.

A turnover model represents the dynamic balance between production of a response and its removal or loss.

Core idea: the drug can change the rate at which a biological response is produced or lost. The observed response then changes over time according to the turnover process.
02 · Why turnover models?

2. Why Use a Turnover Model?

Many pharmacodynamic biomarkers do not respond instantaneously to changes in drug concentration. A drug may alter a biological process, but the measured response changes only as the underlying system turns over.

Turnover models are useful when:

  • The response has a meaningful baseline in the absence of drug.
  • The response is continuously generated and removed.
  • Drug concentration changes more rapidly than the measured pharmacodynamic response.
  • There is a delay between drug exposure and the observed response.
  • The response persists after plasma concentration has begun to decline.
Observed featurePossible turnover interpretation
Stable baselineProduction and loss are approximately balanced at baseline
Delayed responseThe response has its own turnover dynamics
Prolonged effectThe response turnover rate is slower than the drug concentration dynamics
Drug-dependent biomarker suppressionDrug inhibits the production or stimulates the loss of the response
Drug-dependent biomarker stimulationDrug stimulates production or inhibits loss
03 · Baseline turnover

3. The Basic Turnover Equation

Let \(R(t)\) represent the pharmacodynamic response. A simple turnover model assumes that the response is generated at a zero-order rate \(k_{in}\) and removed at a first-order rate proportional to the current response.

\[ \frac{dR(t)}{dt}=k_{in}-k_{out}R(t) \]

Here, \(k_{in}\) is the production rate and \(k_{out}\) is the fractional loss rate.

At baseline, assume there is no drug effect and the system is at steady state. Then:

\[ \frac{dR}{dt}=0 \]

Therefore:

\[ R_0=\frac{k_{in}}{k_{out}} \]

This relationship is fundamental to turnover modeling. The baseline response is determined by the balance between its production and loss rates.

Important: \(k_{in}\) and \(k_{out}\) are not usually independent of the baseline response. If \(R_0\) is known, then \(k_{in}=k_{out}R_0\).
04 · Turnover time scale

4. Turnover Half-Life

The loss rate constant \(k_{out}\) determines the characteristic time scale of the response turnover process.

The corresponding turnover half-life is:

\[ t_{1/2,\mathrm{turnover}}=\frac{\ln(2)}{k_{out}} \]

A small \(k_{out}\) corresponds to slow turnover and a long response half-life. A large \(k_{out}\) corresponds to rapid turnover.

\(k_{out}\)Turnover half-lifeInterpretation
0.50 h\(^{-1}\)1.39 hRapid turnover
0.20 h\(^{-1}\)3.47 hModerate turnover
0.10 h\(^{-1}\)6.93 hSlower turnover
0.05 h\(^{-1}\)13.86 hSlow turnover

The turnover half-life is not necessarily the half-life of the drug. It describes the time scale of the response system.

05 · Drug effects

5. How Does the Drug Enter the Turnover Model?

The drug effect can be incorporated into either the production or loss component of the turnover equation.

A general representation is:

\[ \frac{dR}{dt}=k_{in}\,F_{in}(C)-k_{out}R\,F_{out}(C) \]

where \(F_{in}(C)\) represents a concentration-dependent effect on production and \(F_{out}(C)\) represents a concentration-dependent effect on loss.

This gives several commonly used model structures.

ModelDrug actionBiological interpretation
Inhibition of productionDrug reduces \(k_{in}\)Drug suppresses generation of the response
Stimulation of productionDrug increases \(k_{in}\)Drug promotes generation of the response
Stimulation of lossDrug increases \(k_{out}\)Drug accelerates removal of the response
Inhibition of lossDrug decreases \(k_{out}\)Drug prolongs persistence of the response
06 · Four basic structures

6. The Four Classic Turnover Models

Four basic turnover structures are particularly useful for understanding indirect pharmacodynamic effects.

6.1 Inhibition of production

If drug concentration inhibits response production:

\[ \frac{dR}{dt}=k_{in}\left(1-I(C)\right)-k_{out}R \]

A common inhibitory \(E_{\max}\)-type function is:

\[ I(C)=\frac{I_{\max}C}{IC_{50}+C} \]

Increasing concentration therefore reduces the input into the response system.

6.2 Stimulation of production

If drug stimulates production:

\[ \frac{dR}{dt}=k_{in}\left(1+S(C)\right)-k_{out}R \]

A common stimulation function is:

\[ S(C)=\frac{S_{\max}C}{SC_{50}+C} \]

6.3 Stimulation of loss

If drug accelerates removal of the response:

\[ \frac{dR}{dt}=k_{in}-k_{out}\left(1+S(C)\right)R \]

The response is removed more rapidly while drug concentration is present.

6.4 Inhibition of loss

If drug reduces the loss rate:

\[ \frac{dR}{dt}=k_{in}-k_{out}\left(1-I(C)\right)R \]

The response then persists longer because the drug slows its removal.

Key distinction: two drugs can produce similar concentration-response observations while acting on different turnover components. The mechanistic interpretation depends on the model and the information contained in the data.
07 · Why delay occurs

7. Why Can Turnover Models Produce Delayed Responses?

Suppose drug concentration changes rapidly after dosing. The response does not necessarily follow immediately because the response itself must change through the production-loss process.

For example, if a drug suddenly inhibits production, the loss process may continue removing existing response. The measured response therefore declines progressively rather than jumping immediately to a new value.

Drug concentration Response The response changes gradually because the biological system turns over.

A rapid drug effect on production or loss can generate a slower pharmacodynamic response because the response variable has its own turnover time scale.

This is different from a direct-effect model in which effect is specified immediately as a function of concentration.

08 · Steady state

8. Steady-State Behavior

At a constant drug concentration \(C\), a turnover model may reach a new steady state.

For inhibition of production:

\[ \frac{dR}{dt}=k_{in}\left(1-\frac{I_{\max}C}{IC_{50}+C}\right)-k_{out}R \]

At steady state:

\[ R_{ss} = \frac{k_{in}}{k_{out}} \left( 1-\frac{I_{\max}C}{IC_{50}+C} \right) \]

Because \(R_0=k_{in}/k_{out}\), this can also be written:

\[ R_{ss} = R_0 \left( 1-\frac{I_{\max}C}{IC_{50}+C} \right) \]

This illustrates an important distinction between steady-state response and time to reach that response. The new steady-state value depends on the drug effect, while the speed of approach depends strongly on the turnover dynamics.

10 · Direct versus turnover

10. Turnover Models vs. Direct-Effect Models

A direct-effect model assumes that effect can be expressed directly as a function of concentration.

\[ E(t)=E_0+\frac{E_{\max}C(t)}{EC_{50}+C(t)} \]

A turnover model instead treats the response as a dynamic state variable:

\[ \frac{dR}{dt}=k_{in}F(C)-k_{out}R \]
FeatureDirect-effect modelTurnover model
Response variableCalculated directly from concentrationDynamic state variable
Production/loss processNot explicitly representedExplicitly represented
DelayUsually requires another mechanismCan arise naturally from turnover
BaselineOften represented as \(E_0\)Emerges from \(k_{in}/k_{out}\)
Biological interpretationConcentration-effect relationshipDynamic balance of biological processes

Neither structure is universally preferable. The appropriate model depends on the biological system, study design, sampling schedule, and scientific question.

11 · Worked example

11. Worked Example: An Inhibitory Turnover Model

Consider a hypothetical biomarker with a baseline response of 100 units. Suppose the response turnover rate constant is \(k_{out}=0.10\) h\(^{-1}\). A drug inhibits production with \(I_{\max}=0.80\) and \(IC_{50}=2\) mg/L.

Step 1: Determine the baseline production rate

At baseline:

\[ R_0=\frac{k_{in}}{k_{out}} \]

Therefore:

\[ k_{in}=R_0k_{out} =100(0.10) =10\text{ units/h} \]

Step 2: Calculate the drug effect at 4 mg/L

\[ I(C)=\frac{I_{\max}C}{IC_{50}+C} \]
\[ I(4) = \frac{0.80(4)}{2+4} = \frac{3.2}{6} \approx0.533 \]

Thus, production is reduced by approximately 53.3%.

Step 3: Calculate the new steady-state response

\[ R_{ss} = R_0(1-I(C)) \]
\[ R_{ss} = 100(1-0.533) \approx46.7\text{ units} \]

Step 4: Calculate the turnover half-life

\[ t_{1/2,\mathrm{turnover}} = \frac{0.693}{0.10} = 6.93\text{ h} \]

The model therefore predicts a new steady-state response of approximately 46.7 units, but the response will approach that value gradually over a time scale characterized by a turnover half-life of approximately 6.93 hours.

Interpretation: the drug concentration determines the magnitude of the production inhibition, while \(k_{out}\) determines how quickly the existing biomarker pool can turn over toward the new steady state.
12 · Dynamic behavior

12. The Time Course After a Drug Effect

For a constant concentration and a constant drug effect, the turnover equation has a simple exponential solution.

For a constant input \(k_{in}^\ast\):

\[ \frac{dR}{dt}=k_{in}^\ast-k_{out}R \]

The solution is:

\[ R(t) = R_{ss} + \left(R_0-R_{ss}\right)e^{-k_{out}t} \]

This equation shows that the response approaches the new steady state exponentially.

Time after changeApproximate fraction of the transition completed
0 half-lives0%
1 half-life50%
2 half-lives75%
3 half-lives87.5%
4 half-lives93.75%
5 half-lives96.875%

This is analogous to other first-order dynamic systems: the response approaches, but does not mathematically reach, its new steady state in finite time.

13 · Model estimation

13. What Data Are Needed to Estimate Turnover Models?

Turnover models require data that contain information about both the drug exposure and the response dynamics.

  1. Measure drug concentrations or obtain reliable PK predictions. The PD model needs an exposure input.
  2. Collect baseline response measurements. These help establish \(R_0\).
  3. Sample the response over time. Repeated measurements are needed to characterize turnover.
  4. Capture onset and recovery. The rising and falling portions of the response can provide information about the turnover rate.
  5. Include a useful range of concentrations. Concentration variation helps identify parameters such as \(IC_{50}\), \(EC_{50}\), \(I_{\max}\), or \(S_{\max}\).
  6. Evaluate the model with diagnostics. Observed-versus-predicted plots, residuals, parameter estimates, and simulation-based checks can help determine whether the model adequately describes the data.
Sampling principle: if response measurements are collected only after the system has already reached steady state, there may be little information about \(k_{out}\). The transient portion of the response is often particularly informative for turnover.
14 · Interpretation

14. Interpreting \(k_{in}\), \(k_{out}\), and Baseline Response

The parameters of a turnover model have different roles.

ParameterInterpretation
\(R_0\)Baseline response when production and loss are balanced
\(k_{in}\)Rate of response production or generation
\(k_{out}\)Fractional rate of response loss
\(t_{1/2,\mathrm{turnover}}\)Characteristic time scale for response turnover
\(I_{\max}\)Maximum fractional inhibition in an inhibitory production model
\(IC_{50}\)Concentration producing half of the maximum modeled inhibitory effect
\(S_{\max}\)Maximum fractional stimulation in an appropriate stimulation model
\(SC_{50}\)Concentration producing half of the maximum modeled stimulatory effect

Because \(R_0=k_{in}/k_{out}\), baseline observations alone generally identify their ratio rather than independently determining both rates. Information about the response time course is needed to provide information about the turnover rate.

15 · Interpretation cautions

15. What Turnover Models Do Not Tell Us Automatically

Turnover models are mechanistic representations, but their parameters should not automatically be interpreted as direct measurements of specific biological processes.

  • A turnover compartment is a mathematical representation. It may summarize several biological processes rather than correspond to one physical compartment.
  • Parameter interpretation depends on model structure. The same data can sometimes be represented by different turnover mechanisms.
  • Baseline data alone may not identify production and loss separately. Their ratio determines the baseline response.
  • Sampling frequency matters. Sparse response measurements can make turnover dynamics difficult to estimate.
  • Drug concentration and response delay can be confounded. Poor PK characterization can complicate interpretation of the PD model.
  • Steady-state observations may be insufficient. Transient response data can be important for identifying turnover dynamics.
  • Model fit does not establish mechanism. A model can adequately reproduce observed data without proving that its assumed biological mechanism is literally correct.
Modeling principle: turnover models are most useful when the data contain enough temporal information to distinguish the response dynamics from the drug concentration dynamics.
16 · Practical workflow

16. A Practical Turnover Modeling Workflow

  1. Define the pharmacodynamic response. Identify the biomarker or physiological quantity being modeled.
  2. Characterize baseline behavior. Determine whether the response has a stable baseline and whether \(R_0\) can be estimated reliably.
  3. Characterize the PK input. Obtain measured concentrations or an appropriate PK model.
  4. Inspect the temporal relationship. Determine whether the response appears delayed relative to concentration.
  5. Choose the turnover structure. Consider whether the drug is plausibly acting on production or loss.
  6. Specify the concentration-effect function. An \(E_{\max}\), inhibitory \(E_{\max}\), or related function may describe the drug effect.
  7. Estimate the model parameters. Estimate turnover and drug-effect parameters using an appropriate nonlinear modeling framework.
  8. Evaluate diagnostics. Examine observations versus predictions, residuals, parameter plausibility, and dynamic behavior.
  9. Simulate the model. Check whether the model reproduces important features of the observed response over time.
  10. Use the model for prediction or simulation. Clearly distinguish model-based predictions from directly observed data.
17 · Extensions

17. Extensions of Turnover Models

The basic turnover framework can be expanded when the biological system requires additional structure.

ExtensionPurpose
Time-varying productionRepresent endogenous rhythms or changing physiological inputs
Time-varying lossRepresent changes in response elimination or degradation
Indirect response modelsRepresent drug effects on production or loss of a response
Multiple turnover compartmentsRepresent sequential biological processes or delayed propagation
Transit compartmentsRepresent distributed delays between drug action and measured response
Nonlinear turnoverRepresent capacity-limited production or loss
Population PK/PD modelsDescribe typical turnover behavior and between-subject variability
Covariate modelsExplain differences in turnover parameters between individuals

These extensions should be introduced when they address a specific feature of the data or scientific question rather than simply increasing model complexity.

18 · The bigger picture

18. Turnover Models Within PK/PD Modeling

Turnover models occupy an important position between simple concentration-effect relationships and more elaborate mechanistic pharmacodynamic models.

\[ \text{Dose} \rightarrow \text{PK} \rightarrow C(t) \rightarrow \text{Drug effect} \rightarrow \text{Turnover dynamics} \rightarrow R(t) \]

The PK component determines exposure. The drug-effect component determines how exposure modifies production or loss. The turnover component determines how the biological response evolves through time.

This framework is especially useful when a biomarker has an intrinsic biological lifetime that is different from the lifetime of the drug in plasma.

Big picture: a drug may disappear relatively quickly from plasma while its pharmacodynamic effect persists because the biological response has a slower turnover process.

19. Key Takeaways

  • A turnover model describes a biological response that is continuously produced and removed.
  • The basic turnover equation is \(\frac{dR}{dt}=k_{in}-k_{out}R\).
  • At baseline steady state, \(R_0=k_{in}/k_{out}\).
  • The turnover half-life is \(t_{1/2}=\ln(2)/k_{out}\).
  • Drug effects can act on response production, response loss, or both.
  • Inhibition of production, stimulation of production, stimulation of loss, and inhibition of loss are four fundamental turnover structures.
  • Turnover dynamics can generate delayed pharmacodynamic responses even when the drug acts directly on a biological process.
  • The drug concentration determines the magnitude of the concentration-dependent effect, while turnover parameters determine how the response changes over time.
  • Baseline response identifies the ratio \(k_{in}/k_{out}\), while temporal response data provide information about turnover dynamics.
  • Turnover models can be linked directly to PK models to describe concentration-driven changes in biological responses.
  • A good model fit does not automatically establish that the assumed biological mechanism is literally correct.
  • The appropriate turnover model is the one that is adequate for the scientific question, available data, and intended predictions.
Next step

Where to Go Next

A natural progression is to study indirect-response PK/PD models in greater detail, including inhibition and stimulation of response production or loss, followed by transit-compartment models, effect-compartment models, and more complex mechanistic PK/PD systems.

The next tutorial can build directly on the turnover framework introduced here by deriving the four classic indirect-response models and showing how their predicted time courses differ after a drug dose.

← Back to Pharmacokinetics Tutorials