Tutorials › Pharmacometrics › Hill and Sigmoid Emax Models
Pharmacodynamics · PK/PD Modeling

Hill and Sigmoid Emax Models

Learn how Hill and sigmoid Emax models extend the Emax model to describe steep, nonlinear exposure-response relationships—and how to interpret Emax, EC50, and the Hill coefficient.

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

1. Why Use a Hill or Sigmoid Emax Model?

The standard Emax model is one of the most widely used models for describing a saturable exposure-response relationship. It assumes that the response rises toward a maximum as concentration increases.

For some pharmacodynamic systems, however, the transition from low response to high response is relatively gradual. In others, the response changes much more sharply over a limited concentration range.

The Hill model, also called the sigmoid Emax model, introduces an additional parameter that controls the steepness of the concentration-response relationship.

Core idea: the ordinary Emax model describes the magnitude and location of a saturable response, while the Hill coefficient adds control over how sharply the response approaches Emax.
02 · The equation

2. The Hill or Sigmoid Emax Equation

A common form of the sigmoid Emax model is:

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

Here:

ParameterMeaning
E0Baseline response when concentration approaches zero
EmaxMaximum drug-related change in response above baseline
EC50Concentration producing 50% of the maximum drug-related effect
CDrug concentration or another exposure measure used as the driver of effect
γHill coefficient controlling the steepness of the concentration-response relationship

When \(\gamma=1\), the equation reduces to the ordinary Emax model:

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

Thus, the standard Emax model can be viewed as a special case of the Hill model.

03 · Shape

3. How the Hill Coefficient Changes the Curve

The Hill coefficient is particularly important because it changes the shape of the exposure-response relationship without directly changing the asymptotic maximum effect.

γ < 1 γ = 1 γ > 1 Emax Concentration Response

Increasing the Hill coefficient makes the concentration-response transition progressively steeper around EC50. The illustration is conceptual rather than a fitted dataset.

In general:

  • γ = 1: the ordinary Emax relationship.
  • γ > 1: a steeper, more switch-like transition around EC50.
  • γ < 1: a more gradual transition.

The term sigmoid refers to the characteristic S-shaped relationship that can arise when the Hill coefficient produces sufficient curvature.

04 · EC50

4. Why EC50 Still Has the Same Interpretation

An important property of the Hill model is that EC50 retains a convenient interpretation.

At \(C=EC_{50}\):

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

Therefore:

\[ E(EC_{50})=E_0+\frac{E_{\max}}{2} \]

So EC50 is the concentration associated with half of the maximum drug-related effect, regardless of the value of the Hill coefficient.

Key distinction: EC50 controls where the response transition occurs, while γ controls how sharply that transition occurs.
05 · Parameters

5. Interpreting the Four Main Parameters

Baseline: E0

E0 represents the expected response when the drug concentration is zero or approaches zero under the model.

For a biomarker that already has a nonzero physiological value, E0 can be an important part of the model rather than a nuisance parameter.

Maximum effect: Emax

Emax represents the maximum additional response attributable to the modeled drug effect.

As concentration becomes very large:

\[ \lim_{C\rightarrow\infty}E(C)=E_0+E_{\max} \]

Thus, the asymptotic response is E0 + Emax when Emax is parameterized as the drug-related change from baseline.

Potency: EC50

EC50 describes the concentration scale at which half of the maximum drug-related effect is achieved.

A lower EC50 shifts the concentration-response relationship toward lower concentrations; a higher EC50 shifts it toward higher concentrations.

Steepness: γ

The Hill coefficient determines how rapidly the response changes around EC50.

It is important not to interpret γ automatically as a direct measure of a particular molecular mechanism. In pharmacometric modeling, it is often a phenomenological parameter used to capture the observed steepness of the exposure-response relationship.

06 · Limiting behavior

6. What Happens at Very Low and Very High Concentrations?

The Hill model has useful limiting behavior.

As C approaches zero

When \(C\rightarrow0\), assuming \(\gamma>0\):

\[ E(C)\rightarrow E_0 \]

The response therefore approaches baseline.

As C becomes very large

When \(C\rightarrow\infty\):

\[ E(C)\rightarrow E_0+E_{\max} \]

The response approaches an upper asymptote rather than increasing indefinitely.

Saturation: the Hill Emax model is appropriate when the drug-related response is expected to approach a finite maximum over increasing exposure.
07 · Steepness

7. A Mathematical View of Steepness

The Hill coefficient can be understood more precisely by examining the slope of the response curve.

For:

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

the derivative with respect to concentration is:

\[ \frac{dE}{dC} = \frac{E_{\max}\gamma EC_{50}^{\gamma}C^{\gamma-1}} {\left(EC_{50}^{\gamma}+C^\gamma\right)^2} \]

At \(C=EC_{50}\), the slope simplifies to:

\[ \left.\frac{dE}{dC}\right|_{C=EC_{50}} = \frac{E_{\max}\gamma}{4EC_{50}} \]

This makes the role of γ especially clear: holding Emax and EC50 fixed, increasing γ increases the slope at EC50.

08 · Log concentration

8. Why the Relationship Often Looks More Sigmoidal on a Log Scale

Pharmacodynamic concentrations can span several orders of magnitude. For that reason, concentration-response relationships are frequently examined using a logarithmic concentration axis.

The Hill model can be rewritten in terms of the log concentration ratio:

\[ \frac{E(C)-E_0}{E_{\max}} = \frac{1}{1+\left(\frac{EC_{50}}{C}\right)^\gamma} \]

or equivalently:

\[ \frac{E(C)-E_0}{E_{\max}} = \frac{1}{1+\exp\left[-\gamma\ln\left(\frac{C}{EC_{50}}\right)\right]} \]

This form shows the connection between the Hill model and a logistic-type function when response is plotted against log concentration.

Practical point: a response that appears highly curved on a linear concentration axis may become easier to interpret when concentration is displayed on a logarithmic scale.
09 · Worked example

9. Worked Example: Predicting Response From a Hill Model

Suppose a pharmacodynamic biomarker has a baseline value of 20 units. A drug is expected to produce a maximum additional effect of 80 units, with an EC50 of 10 mg/L. Suppose the fitted Hill coefficient is 2.

The model is therefore:

\[ E(C)=20+\frac{80C^2}{10^2+C^2} \]

Step 1: Response at C = 2 mg/L

\[ E(2) = 20+\frac{80(2^2)}{10^2+2^2} \] $$ = 20+\frac{320}{104} \approx23.08 $$ $$ \boxed{E(2)\approx23.1\text{ units}} $$

Step 2: Response at EC50 = 10 mg/L

\[ E(10) = 20+\frac{80(10^2)}{10^2+10^2} \] $$ = 20+40 = 60 $$ $$ \boxed{E(10)=60\text{ units}} $$

The maximum drug-related effect is 80 units, so half of that effect is 40 units above baseline. The result is therefore exactly what the EC50 definition predicts.

Step 3: Response at C = 20 mg/L

\[ E(20) = 20+\frac{80(20^2)}{10^2+20^2} \] $$ = 20+\frac{32000}{500} = 84 $$ $$ \boxed{E(20)=84\text{ units}} $$

Step 4: Response at C = 100 mg/L

\[ E(100) = 20+\frac{80(100^2)}{10^2+100^2} \] $$ = 20+\frac{800000}{10100} \approx99.21 $$ $$ \boxed{E(100)\approx99.2\text{ units}} $$

The predicted response is approaching the asymptotic maximum of \(20+80=100\) units.

ConcentrationPredicted responseInterpretation
2 mg/L23.1 unitsNear baseline
10 mg/L60.0 units50% of maximum drug-related effect
20 mg/L84.0 unitsRapidly approaching maximum
100 mg/L99.2 unitsVery close to the asymptote
10 · Model comparison

10. Emax Versus Hill Emax

The ordinary Emax model and Hill Emax model are closely related.

FeatureOrdinary EmaxHill Emax
BaselineE0E0
Maximum drug-related effectEmaxEmax
Potency parameterEC50EC50
Shape parameterFixed at γ = 1Estimated Hill coefficient γ
Curve steepnessFixedFlexible
Number of structural parametersUsually 3Usually 4

The additional flexibility of the Hill model comes at a cost: there is one more parameter to estimate. The data must therefore contain enough information to distinguish the Hill coefficient from the other parameters.

Modeling principle: a more flexible model is not automatically a better model. The Hill coefficient should be supported by the data and the scientific question.
11 · Inhibitory effects

11. Hill Models for Inhibition

The same basic framework can describe inhibitory responses. One common representation is:

\[ E(C)=E_0-\frac{I_{\max}C^\gamma}{IC_{50}^\gamma+C^\gamma} \]

Here, \(I_{\max}\) represents the maximum decrease from baseline and \(IC_{50}\) is the concentration associated with half of the maximum inhibitory effect.

At \(C=IC_{50}\):

\[ E(IC_{50})=E_0-\frac{I_{\max}}{2} \]

The mathematical structure is therefore closely related to the stimulatory Hill Emax model; the main difference is the direction of the effect.

13 · Time delay

13. When a Hill Model Alone Is Not Enough

A Hill Emax relationship assumes that the observed effect is directly related to the concentration used as the model input. This assumption may not hold when there is a meaningful delay between plasma concentration and pharmacodynamic effect.

For example, a drug may distribute slowly to the effect site, trigger a downstream biological process, or produce an effect that persists after plasma concentrations have fallen.

In such situations, a direct concentration-effect model may show hysteresis or systematic time-related deviations.

Important distinction: changing the Hill coefficient changes the shape of the concentration-response relationship. It does not by itself create a time-delay mechanism.

A common extension is an effect-compartment model in which the PD model uses an effect-site concentration \(C_e\) rather than plasma concentration \(C_p\):

\[ \frac{dC_e}{dt}=k_{e0}(C_p-C_e) \]

The Hill model can then be applied to \(C_e\):

\[ E(C_e) = E_0+ \frac{E_{\max}C_e^\gamma} {EC_{50}^\gamma+C_e^\gamma} \]
14 · Estimation

14. Estimating Hill Model Parameters

Hill model parameters are generally estimated by comparing observed pharmacodynamic responses with model-predicted responses.

  1. Define the response variable. Determine exactly what biomarker, endpoint, or physiological measurement is being modeled.
  2. Choose the exposure driver. This may be concentration, effect-site concentration, exposure, or another scientifically justified measure.
  3. Specify the structural model. Decide whether an ordinary Emax or Hill Emax relationship is appropriate.
  4. Estimate E0, Emax, EC50, and γ. The estimation method depends on the data structure and modeling framework.
  5. Evaluate diagnostics. Examine residuals, observed-versus-predicted plots, parameter precision, and the shape of the fitted relationship.
  6. Assess identifiability. Determine whether the available exposure range contains sufficient information to estimate all model parameters.
  7. Perform sensitivity or simulation analyses. Consider how conclusions change under alternative plausible parameter values or structural models.

The exposure range is particularly important. If nearly all observations are far below EC50, the maximum effect and Hill coefficient may be poorly identified. Similarly, if observations all lie near saturation, EC50 and γ may be difficult to estimate precisely.

15 · Identifiability

15. Why the Hill Coefficient Can Be Difficult to Estimate

The Hill coefficient is often the least well-informed parameter in a sigmoid Emax model because it primarily determines the shape of the transition region.

Consider three broad situations:

Observed exposure rangePotential information available
Mostly below EC50Information about baseline and the low-exposure portion of the curve, but limited information about saturation and steepness
Spans EC50Strongest opportunity to characterize the transition and estimate the Hill coefficient
Mostly above EC50Information about the upper asymptote, but potentially limited information about potency and transition shape

This is why dose-ranging or exposure-ranging studies are important when the scientific objective is to characterize the complete exposure-response relationship.

Practical lesson: a statistically estimated Hill coefficient is not necessarily a well-informed Hill coefficient. Precision and identifiability should be evaluated before giving the parameter a strong interpretation.
16 · Common mistakes

16. Common Interpretation Mistakes

Mistake 1: Treating γ as a molecular binding constant

The Hill coefficient can sometimes have mechanistic interpretations in specific biochemical models, but in pharmacometric exposure-response modeling it is often used as an empirical shape parameter.

Mistake 2: Assuming EC50 is the concentration of maximum effect

EC50 is not the concentration that produces Emax. It is the concentration associated with half of the maximum drug-related effect.

Mistake 3: Ignoring baseline

If E0 is nonzero, the total response at saturation is E0 + Emax, not simply Emax.

Mistake 4: Assuming a large γ proves a switch-like biological mechanism

A steep fitted relationship may be a useful empirical description without establishing the biological mechanism responsible for the observed response.

Mistake 5: Estimating γ without enough exposure range

The Hill coefficient requires information about the transition region. Sparse observations concentrated in only one portion of the curve may not support precise estimation.

Mistake 6: Using plasma concentration when an effect-site delay is important

A Hill model applied directly to plasma concentration cannot by itself account for delayed pharmacodynamic response. An effect-compartment or other dynamic model may be needed.

17 · Practical workflow

17. A Practical Hill Emax Modeling Workflow

  1. Plot response against exposure. Examine the relationship before selecting a nonlinear model.
  2. Use a log exposure axis when appropriate. This can make the shape of a wide exposure range easier to assess.
  3. Fit the ordinary Emax model first when scientifically reasonable. It provides a useful reference with γ fixed at 1.
  4. Consider the Hill extension. Estimate γ when the observed response pattern provides evidence for additional curvature.
  5. Check parameter precision. Examine standard errors, confidence intervals, profile likelihoods, or other appropriate uncertainty measures.
  6. Inspect observed-versus-predicted behavior. Look for systematic deviations across the exposure range.
  7. Assess time dependence. If response depends on prior exposure or shows hysteresis, consider a dynamic PK/PD model.
  8. Evaluate extrapolation. Be cautious when predicting beyond the exposure range used to estimate the model.
  9. Use simulation where appropriate. Simulate the fitted model to understand the implications of uncertainty in Emax, EC50, and γ.

18. Key Takeaways

  • The Hill model, also called the sigmoid Emax model, extends the ordinary Emax model by introducing a Hill coefficient.
  • The standard Emax model is a special case of the Hill model with \(\gamma=1\).
  • E0 represents baseline response, Emax represents the maximum drug-related effect, and EC50 controls the concentration scale of the response.
  • The Hill coefficient \(\gamma\) controls the steepness of the concentration-response relationship.
  • At \(C=EC_{50}\), the predicted response is exactly halfway between baseline and the upper asymptote.
  • As concentration approaches zero, the response approaches E0; as concentration becomes very large, it approaches E0 + Emax.
  • A Hill coefficient greater than 1 produces a steeper transition around EC50, whereas a coefficient below 1 produces a more gradual transition.
  • The Hill coefficient is often a phenomenological shape parameter and should not automatically be assigned a specific mechanistic interpretation.
  • Reliable estimation of the Hill coefficient requires observations that adequately characterize the exposure-response transition.
  • A Hill Emax model describes concentration-response shape; it does not by itself account for delayed pharmacodynamic effects.
  • When hysteresis or delayed response is present, an effect-compartment or other dynamic PK/PD model may be appropriate.
  • Model complexity should be supported by the scientific question, exposure range, and information contained in the data.
Next step

Where to Go Next

After understanding Hill and sigmoid Emax models, a natural progression is to study indirect-response PK/PD models, effect-compartment models, and time-delay mechanisms.

These models extend the concentration-effect framework when pharmacodynamic response is governed by turnover, delayed distribution, or downstream biological processes rather than an instantaneous concentration-effect relationship.

References

References

  1. Holford NHG, Sheiner LB. Understanding the dose-effect relationship: clinical application of pharmacokinetic-pharmacodynamic models. Clinical Pharmacokinetics.
  2. Gabrielsson J, Weiner D. Pharmacokinetic and Pharmacodynamic Data Analysis: Concepts and Applications.
  3. Rowland M, Tozer TN. Clinical Pharmacokinetics and Pharmacodynamics: Concepts and Applications.
  4. Jusko WJ, Ko HC. Physiologic indirect response models characterize diverse types of pharmacodynamic effects. Clinical Pharmacology & Therapeutics.

Related Tutorials

Clinical Biostats

← Back to Pharmacokinetics Tutorials