Hysteresis And Effect-Compartment Modeling

Why Concentration and Response May Not Move Together

A drug concentration can rise and fall faster than its pharmacodynamic response. When response is plotted against plasma concentration, observations at the same concentration may then have different responses depending on whether concentration is rising or falling. This pattern is called hysteresis. Its direction and timing can help distinguish a response that follows concentration immediately from one that is delayed.

For a stimulatory drug effect, a counterclockwise loop is consistent with response lagging behind plasma concentration: on the rising concentration limb, response is relatively low; on the falling limb, response remains relatively high. An effect-compartment model represents this delay by allowing a hypothetical effect-site concentration to approach plasma concentration over time. The effect-site concentration is a modeling construct, not necessarily a measurable concentration in a distinct anatomical compartment.

Definition: In a first-order effect-compartment model, $C_p(t)$ is plasma concentration and $C_e(t)$ is the modeled effect-compartment concentration. Their relationship is $$ \frac{dC_e(t)}{dt}=k_{e0}\left(C_p(t)-C_e(t)\right), $$ where $k_{e0}>0$ is the equilibration rate constant, usually expressed in reciprocal time units. A response model then relates response to $C_e(t)$, rather than directly to $C_p(t)$.

A common stimulatory response function is

E(t)=E_0+\frac{E_{\max}C_e(t)}{EC_{50}+C_e(t)}.

Here $E_0$ is baseline response, $E_{\max}$ is the maximum increase above baseline, and $EC_{50}$ is the effect-compartment concentration producing half of that maximum increase. Concentrations and $EC_{50}$ must use the same units; response uses the measurement's units. For an inhibitory response, the response function must instead represent inhibition, with a direction and parameterization appropriate to the endpoint. The equilibration equation itself is unchanged.

Reading the Hysteresis Loop

Consider plotting plasma concentration on the horizontal axis and a stimulatory response on the vertical axis, with observations connected in time order. If concentration rises first while response lags, the path moves rightward at comparatively low response. After the concentration peak, response can continue rising or remain elevated while concentration falls, so the path returns leftward at higher response. This is the counterclockwise direction.

The loop is evidence of a time-dependent concentration–response relationship, not proof of a particular mechanism. A direct-response model assumes that response at a given time is determined by the plasma concentration at that time and other modeled covariates. Under the same conditions, it therefore predicts the same response at the same plasma concentration. An effect-compartment model can predict different responses at that concentration because $C_e(t)$ depends on the preceding concentration history.

Worked Example: Recognizing Counterclockwise Hysteresis

Suppose a hypothetical study records plasma concentration in mg/L and a stimulatory response in response units. The baseline response is approximately 10 units.

Time (h)	Plasma concentration (mg/L)	Response (units)
0	0	10
1	2	30
2	4	50
4	2	42
6	0	20

At $C_p=2$ mg/L, response is 30 units on the rising limb and 42 units on the falling limb. The higher response on the falling limb shows that response does not depend on current plasma concentration alone. With concentration on the horizontal axis and response on the vertical axis, the time-ordered path moves right at lower response and returns left at higher response, indicating counterclockwise hysteresis. The pattern is compatible with delayed drug effect, although other time-dependent processes could also produce it.

Solving and Interpreting the Effect-Compartment Model

If $C_p$ is constant over an interval of length $\Delta t$, the differential equation has a closed-form solution. Starting from effect-compartment concentration $C_{e,0}$, the solution is

C_e(\Delta t)=C_p+(C_{e,0}-C_p)e^{-k_{e0}\Delta t}.

This says that the difference $C_p-C_e$ shrinks exponentially. It also gives a convenient calculation when plasma concentrations are treated as constant between sampling times. For a time-varying concentration profile, the equation can instead be integrated numerically using the concentration predictions from the pharmacokinetic model.

Key result: The effect-compartment equilibration half-life is the time required for the difference between $C_p$ and $C_e$ to decrease by half when $C_p$ is held constant: $$ t_{1/2,e0}=\frac{\ln(2)}{k_{e0}}. $$ It has units of time. It describes equilibration in this model; it is not the plasma elimination half-life.

The half-life formula follows directly from the solution. With constant $C_p$, the distance from equilibrium is

C_p-C_e(t)=(C_p-C_{e,0})e^{-k_{e0}t}.

At the half-life, this distance is half its initial value. Therefore $e^{-k_{e0}t_{1/2,e0}}=1/2$. Taking natural logarithms gives $-k_{e0}t_{1/2,e0}=-\ln(2)$, and hence $t_{1/2,e0}=\ln(2)/k_{e0}$. A larger $k_{e0}$ means faster modeled equilibration and a shorter half-life.

Worked Example: Calculating Effect-Compartment Concentration

In a hypothetical calculation, plasma concentration is 4 mg/L for one hour, then 1 mg/L for the next hour. Let $C_e(0)=0$ mg/L and $k_{e0}=0.5$ h$^{-1}$. Treat plasma concentration as constant within each one-hour interval.

For the first hour,

C_e(1)=4+(0-4)e^{-0.5(1)} =4(1-e^{-0.5}) =1.574\ \text{mg/L}.

For the second hour, the new constant plasma concentration is 1 mg/L, and the starting effect-compartment concentration is 1.574 mg/L:

C_e(2)=1+(1.574-1)e^{-0.5(1)} =1+0.574(0.6065) =1.348\ \text{mg/L}.

Thus $C_p$ has fallen from 4 to 1 mg/L, while $C_e$ remains above 1 mg/L because it has not yet equilibrated. For illustration, if $E_0=10$ response units, $E_{\max}=80$ response units, and $EC_{50}=2$ mg/L, then at two hours the predicted response is

E(2)=10+\frac{80(1.348)}{2+1.348} =42.21\ \text{response units}.

The calculation assumes the stated piecewise-constant plasma profile and a single first-order equilibration process. With measured or model-predicted concentrations that change substantially within an interval, a finer time grid or numerical integration may be needed.

Estimating the Delay and Comparing Models

In analysis, $k_{e0}$ is typically estimated along with the parameters of the response model by fitting predictions to the observed responses. The pharmacokinetic concentration profile is supplied by measured concentrations or by a pharmacokinetic model. At each proposed $k_{e0}$, the effect-compartment equation produces $C_e(t)$, which is then passed to the response function. Estimation should account for the response data's measurement-error structure, and the model should specify how the initial $C_e$ is set. For a subject starting at baseline before drug exposure, a common condition is $C_e(0)=0$; other study designs may require a different initial condition.

The main comparison is with a direct-response model, in which the response function uses $C_p(t)$ instead of $C_e(t)$. The models should be compared using the same observations and a suitable likelihood or residual-based criterion, while examining the fit over time and against concentration. A lower objective value alone does not establish that the effect-compartment interpretation is scientifically adequate. In particular, an effect compartment is not interchangeable with the turnover mechanism represented in the earlier material on Indirect-Response Pharmacodynamics. The two structures encode different explanations for delayed response.

Worked Example: Estimating an Equilibration Half-Life

Consider a hypothetical, idealized concentration profile: $C_p=4$ mg/L from time 0 to 2 h, followed by $C_p=0$ mg/L. Assume $C_e(0)=0$, and treat $E_0=10$ response units, $E_{\max}=80$ response units, and $EC_{50}=2$ mg/L as fixed. At one hour, suppose the observed response is 50 units.

First infer the effect-compartment concentration corresponding to that response. The fraction of the maximum increase is $(50-10)/80=0.5$. Thus

0.5=\frac{C_e}{2+C_e} \quad\Longrightarrow\quad C_e=2\ \text{mg/L}.

During the first hour, plasma concentration is constant at 4 mg/L, so

2=4(1-e^{-k_{e0}(1)}) \quad\Longrightarrow\quad e^{-k_{e0}}=0.5 \quad\Longrightarrow\quad k_{e0}=\ln(2)=0.6931\ \text{h}^{-1}.

The estimated equilibration half-life is therefore

t_{1/2,e0}=\frac{\ln(2)}{0.6931\ \text{h}^{-1}}=1.000\ \text{h}.

At two hours, the concentration profile gives $C_e(2)=4(1-e^{-0.6931(2)})=3$ mg/L. After plasma concentration falls to zero, one further hour gives $C_e(3)=3e^{-0.6931(1)}=1.5$ mg/L. The corresponding effect-compartment response prediction is $10+80(1.5)/(2+1.5)=44.29$ units. A direct-response model with these same fixed response parameters predicts $10+80(4)/(2+4)=63.33$ units at one hour, and 10 units at three hours when plasma concentration is zero. The delayed model matches the one-hour value and predicts a residual response after plasma concentration has fallen; the direct predictions do not.

This algebra illustrates how a rate constant translates into a half-life, but it is not a reliable general estimation method by itself. Real data are noisy, response parameters are often estimated rather than known, and a single observation cannot establish that the assumed model is correct. In practice, estimate parameters from the full time course and assess uncertainty and fit diagnostics.

Assumptions, Diagnostics, and Limitations

An effect-compartment interpretation rests on several assumptions. The specified plasma concentration profile must be adequate; the modeled effect site is assumed to equilibrate with plasma by a first-order process; and the response must be adequately described as a function of $C_e$ and the other included predictors. The model also assumes that one $k_{e0}$ adequately describes the delay over the analyzed range and time period. These are modeling choices, not conclusions guaranteed by observing a loop.

Check the time course as well as the hysteresis plot. Plot observed responses and model predictions against time, inspect residuals for systematic patterns, and compare observed and predicted responses against both $C_p$ and $C_e$. A persistent pattern in residuals after the concentration peak may signal that the delay model, the response function, or the concentration profile is inadequate. Check that sampling times are correctly aligned and that the baseline is handled consistently. Sparse samples near the concentration or response peak may make loop direction and timing difficult to determine.

A fitted delay can be hard to distinguish from other explanations. Active metabolites, tolerance, disease progression, changes in baseline response, or an indirect response mechanism can all create time-dependent concentration–response patterns. A single effect compartment may also be too simple if equilibration has multiple time scales. A fitted $k_{e0}$ may be weakly identified when observations do not cover both rising and falling concentration phases, when response data are sparse, or when changes in $C_e$ are confounded with changes in response-model parameters. A very large estimated $k_{e0}$ corresponds to near-immediate equilibration; in that limit, the effect-compartment model approaches a direct-response model.

Common pitfall: “The response is higher at the same plasma concentration on the falling limb, so a counterclockwise loop proves effect-site equilibration.” The observation supports a time-dependent relationship, but does not identify its cause. Check the time ordering, model the concentration history, compare plausible response structures, and inspect residuals before attributing the delay to a single effect compartment.

The estimated half-life should be reported with its units and interpreted narrowly: it is the half-life implied by the fitted effect-compartment rate constant under the model assumptions. It is not automatically the time for the clinical response to halve, the duration of a pharmacodynamic effect, or the terminal half-life of drug in plasma. If the fitted rate is imprecise or poorly identified, that uncertainty limits how specifically the delay can be interpreted.

Check Your Understanding

Answer each question, then check the short solutions below.

On a plot with plasma concentration on the horizontal axis and a stimulatory response on the vertical axis, response is lower on the rising limb than on the falling limb at the same concentration. What direction of hysteresis is consistent with this pattern?
If $k_{e0}=0.3466$ h$^{-1}$, calculate the effect-compartment equilibration half-life.
With constant $C_p=3$ mg/L, $C_e(0)=1$ mg/L, $k_{e0}=1$ h$^{-1}$, and an interval of one hour, calculate $C_e(1)$.
Why can a direct-response model fail when two observations have the same plasma concentration but different responses?
Name one reason, other than effect-site equilibration, that could produce a time-dependent concentration–response pattern.

Solutions. (1) Counterclockwise hysteresis is consistent with the described pattern: the path returns at higher response as concentration falls. (2) $t_{1/2,e0}=\ln(2)/0.3466=2.000$ h. (3) $C_e(1)=3+(1-3)e^{-1}=3-2(0.3679)=2.264$ mg/L. (4) Under otherwise identical conditions, a direct-response model predicts the same response at the same current plasma concentration, so it cannot represent both observations without residual error or additional predictors. (5) Examples include an active metabolite, tolerance, a changing disease state, or an indirect response mechanism.