Distinguishing Stimulation of Input From Stimulation of Loss

Why the Site of Stimulation Matters

A drug can stimulate the process that supplies a measured response, or it can stimulate the process that removes that response. These mechanisms have opposite implications for the response itself: more input tends to raise the response, while faster loss tends to lower it. Both mechanisms are indirect because the drug changes a turnover rate rather than adding a direct response term.

The distinction matters beyond the direction of the steady-state change. During constant exposure, input stimulation and loss stimulation have different response time constants. After exposure ends, the return toward baseline can look similar if both mechanisms have the same drug-free turnover rate. Interpretation therefore benefits from examining onset, the on-drug equilibrium, and the transition after drug concentrations or drug effects decline.

Definition: In the basic indirect-response model, $R(t)$ is the measured response, $k_{\mathrm{in}}$ is its zero-order input rate, and $k_{\mathrm{out}}$ is its first-order loss-rate constant. Without drug effect, the model is $$ \frac{dR}{dt}=k_{\mathrm{in}}-k_{\mathrm{out}}R. $$ At baseline steady state, $R_0=k_{\mathrm{in}}/k_{\mathrm{out}}$. Let $s(t)\geq 0$ denote the drug effect on a rate. Stimulation of input multiplies $k_{\mathrm{in}}$ by $1+s(t)$; stimulation of loss multiplies $k_{\mathrm{out}}$ by $1+s(t)$.

For the comparisons below, assume the response begins at the drug-free baseline $R_0$, the baseline rates are constant, and the stated drug effect acts only on the specified turnover rate. Unless stated otherwise, $s$ is a nonnegative, fixed effect during constant exposure. These assumptions make the mechanistic predictions clear; if they are not reasonable, a mismatch with observed data need not identify which of the two mechanisms is correct.

Two Models, Opposite Response Directions

With input stimulation, the response equation is $dR/dt=k_{\mathrm{in}}[1+s(t)]-k_{\mathrm{out}}R$. With loss stimulation, it is $dR/dt=k_{\mathrm{in}}-k_{\mathrm{out}}[1+s(t)]R$. Normalize the response as $r=R/R_0$, and use $k_{\mathrm{in}}=k_{\mathrm{out}}R_0$. Then the two equations become

\begin{aligned} \text{Input stimulation:}\qquad &\frac{dr}{dt}=k_{\mathrm{out}}\{1+s(t)-r\},\\ \text{Loss stimulation:}\qquad &\frac{dr}{dt}=k_{\mathrm{out}}\{1-[1+s(t)]r\}. \end{aligned}

At the instant a constant effect $s$ begins, if $r=1$, the input-stimulation model predicts $dr/dt=k_{\mathrm{out}}s>0$; the loss-stimulation model predicts $dr/dt=-k_{\mathrm{out}}s<0$. Thus the initial direction is a direct mechanistic clue when baseline is stable and the exposure-related effect is known to have begun.

Key result: Under a fixed effect $s>0$, input stimulation has equilibrium $r_{\mathrm{ss}}=1+s$ and response time constant $1/k_{\mathrm{out}}$. Loss stimulation has equilibrium $r_{\mathrm{ss}}=1/(1+s)$ and response time constant $1/[k_{\mathrm{out}}(1+s)]$. Input stimulation raises the equilibrium; loss stimulation lowers it.

These results follow by setting the corresponding derivative to zero. For input stimulation, $0=k_{\mathrm{out}}(1+s-r_{\mathrm{ss}})$, giving $r_{\mathrm{ss}}=1+s$. For loss stimulation, $0=k_{\mathrm{out}}[1-(1+s)r_{\mathrm{ss}}]$, giving $r_{\mathrm{ss}}=1/(1+s)$. To find the time constant, subtract the equilibrium from the response. In the input model, $d(r-r_{\mathrm{ss}})/dt=-k_{\mathrm{out}}(r-r_{\mathrm{ss}})$. In the loss model, $d(r-r_{\mathrm{ss}})/dt=-k_{\mathrm{out}}(1+s)(r-r_{\mathrm{ss}})$. A deviation governed by $dy/dt=-ay$ decays as $y(t)=y(0)e^{-at}$; its time constant is $1/a$.

Worked Trajectories During Constant Exposure

Worked Example: Onset With Input Stimulation

Suppose the hypothetical baseline response is $R_0=30$ units and $k_{\mathrm{out}}=0.40\ \mathrm{h}^{-1}$. Then $k_{\mathrm{in}}=k_{\mathrm{out}}R_0=12$ units/h. A constant exposure produces input stimulation $s=0.60$, beginning at $t=0$ from baseline. Find the response at 2 h.

For a constant effect, the input-stimulation equation is $dR/dt=k_{\mathrm{in}}(1+s)-k_{\mathrm{out}}R$. Its equilibrium is $R_{\mathrm{ss}}=R_0(1+s)=30(1.60)=48$ units. The solution starting from $R(0)=30$ is

R(t)=R_{\mathrm{ss}}+[R(0)-R_{\mathrm{ss}}]e^{-k_{\mathrm{out}}t} =30[1.60-0.60e^{-0.40t}].

At $t=2$ h, $e^{-0.8}\approx0.4493$, so $R(2)=30[1.60-0.60(0.4493)]=30(1.33042)=39.91$ units. The response has risen from 30 toward the on-drug equilibrium of 48 units. Its initial slope is $k_{\mathrm{in}}s=12(0.60)=7.20$ units/h.

Worked Example: Onset With Loss Stimulation

Keep the same hypothetical baseline, $R_0=30$ units and $k_{\mathrm{out}}=0.40\ \mathrm{h}^{-1}$, but now let the effect $s=0.60$ stimulate loss. Starting from baseline, calculate the response after 2 h and compare its time course with input stimulation.

The on-drug loss rate is $k_{\mathrm{out}}(1+s)=0.40(1.60)=0.64\ \mathrm{h}^{-1}$. The equilibrium is $R_{\mathrm{ss}}=R_0/(1+s)=30/1.60=18.75$ units. Therefore

R(t)=18.75+(30-18.75)e^{-0.64t} =30[0.625+0.375e^{-0.64t}].

At $t=2$ h, $e^{-1.28}\approx0.2780$, so $R(2)=30[0.625+0.375(0.2780)]=30(0.72925)=21.88$ units. The response falls toward 18.75 units. Its initial slope is $-k_{\mathrm{out}}sR_0=-0.40(0.60)(30)=-7.20$ units/h. The absolute initial slope has the same magnitude as in the preceding example because the baseline and $s$ are the same, but its direction is opposite. The loss-stimulation approach to its equilibrium is faster: its time constant is $1/0.64=1.5625$ h rather than $1/0.40=2.50$ h.

What Washout Can and Cannot Show

If the drug effect returns immediately to zero after washout, either model reverts to $dR/dt=k_{\mathrm{in}}-k_{\mathrm{out}}R$. From a response value $R_w$ at washout time, the post-washout trajectory is

R(t)=R_0+(R_w-R_0)e^{-k_{\mathrm{out}}(t-t_w)},\qquad t\geq t_w.

Both mechanisms therefore share the same drug-free recovery time constant when their baseline $k_{\mathrm{out}}$ is the same. Their recovery directions differ because they typically begin washout on opposite sides of baseline: an elevated response after input stimulation falls toward baseline, while a depressed response after loss stimulation rises toward baseline. The common exponential rate is not, by itself, proof of either mechanism.

Worked Example: Comparing Post-Washout Recovery

Use the preceding hypothetical parameters and suppose each mechanism reaches its constant-effect equilibrium before immediate washout: 48 units for input stimulation and 18.75 units for loss stimulation. What are the predicted responses 2 h after washout?

The drug-free rate constant is $0.40\ \mathrm{h}^{-1}$ in both cases, so the fraction of the deviation from baseline that remains after 2 h is $e^{-0.40(2)}=e^{-0.8}\approx0.4493$. After input stimulation, $R=30+(48-30)(0.4493)=30+8.0874=38.09$ units. After loss stimulation, $R=30+(18.75-30)(0.4493)=30-5.0546=24.95$ units. The first response is declining toward baseline; the second is increasing toward baseline. In both cases, 44.93% of the washout-time deviation remains.

Real washout is often not an instantaneous disappearance of pharmacologic effect. If $s(t)$ declines gradually, the response may keep rising briefly under input stimulation or keep falling briefly under loss stimulation, depending on the remaining effect and the response state. A delayed concentration-effect relationship can also separate the plasma concentration profile from the effective $s(t)$. In those circumstances, the simple post-washout expression applies only once the effect is negligible; the full time-varying model is needed before attributing a lag or turning point to the mechanism.

Observations That Discriminate the Mechanisms

The most useful evidence combines direction and timing rather than relying on one response measurement. Under the assumptions stated above, the following predictions differ:

Feature	Input stimulation	Loss stimulation
Initial response direction from baseline	Increases	Decreases
Equilibrium under fixed positive effect	Above baseline by factor $1+s$	Below baseline by factor $1/(1+s)$
On-drug response time constant	$1/k_{\mathrm{out}}$	$1/[k_{\mathrm{out}}(1+s)]$, shorter for $s>0$
Direction after effect disappears	Falls toward baseline from above	Rises toward baseline from below
Drug-free recovery time constant	$1/k_{\mathrm{out}}$	$1/k_{\mathrm{out}}$

For example, a well-measured early increase during confirmed exposure supports input stimulation over loss stimulation in this model, while a sustained decrease supports loss stimulation. A measured equilibrium can also be informative if the baseline is established and the exposure effect is approximately constant. Sampling during onset helps estimate direction and the on-drug approach rate; sampling near equilibrium helps establish the response level; sampling after exposure has declined helps assess recovery. No single feature is decisive if the effect size, exposure history, baseline, or response mechanism is uncertain.

Assumptions, Diagnostics, and Common Mistakes

A mechanistic interpretation of these signatures requires more than a curve that happens to rise or fall. Check that the response was near a stable baseline before exposure, that the exposure timing is known, and that the response scale and units are consistent. Check whether an observed plateau is plausibly an equilibrium rather than a transient observed over a short interval. If exposure is changing substantially, a constant-effect equilibrium or time constant should not be applied as though exposure were fixed.

Systematic disagreement with the predicted direction or curvature is a warning. For instance, an initial rise followed by a fall during apparently maintained exposure may reflect changing exposure, a delayed effect, baseline drift, feedback, or a mechanism not represented by either simple model. An apparent post-dose lag can arise from drug-distribution or effect delay rather than unusually slow response turnover. Measurements with substantial noise or sparse timing can also obscure the initial slope and conceal a transient. These alternatives should be examined before interpreting a fitted stimulation term literally.

Common mistake: “The drug stimulates loss, so the response should be stimulated upward.” The correction is to distinguish the stimulated process from the measured response. Increasing the loss-rate constant removes response faster; with input unchanged, it lowers the equilibrium to $R_0/(1+s)$.
Using washout speed alone to assign mechanism: once the effect is gone, both models recover with the same $k_{\mathrm{out}}$ under the assumptions here. The side of baseline and the on-drug trajectory provide the distinguishing evidence.
Calling a transient a steady state: a response that has not settled cannot be compared directly with the equilibrium formulas. Use the time-course solution or continue sampling long enough to assess whether a plateau is supported.

The simple comparison has important limits. It assumes stimulation is represented by the same nonnegative multiplier in the specified rate, a stable baseline, and no additional drug effects or response feedback. A response can be influenced by both input and loss, and a real concentration-effect relationship can vary over time. The signatures provide model-based predictions, not a guarantee that one mechanism is uniquely established by a finite set of observations.

Check Your Understanding

Answer each question using the assumptions of the basic models above. Solutions follow the questions.

At baseline, what is the initial direction of change when $s=0.25$ begins for input stimulation and for loss stimulation?
If $R_0=40$ units, what are the predicted equilibria for input and loss stimulation when $s=0.50$?
For $k_{\mathrm{out}}=0.30\ \mathrm{h}^{-1}$ and $s=0.50$, calculate the on-drug time constant under each mechanism.
Why does the post-washout recovery time constant not distinguish the mechanisms in the basic model?
Name two observations or conditions that would make a mechanistic interpretation less secure.

Solutions. 1. Input stimulation initially increases the response because $dr/dt=k_{\mathrm{out}}s>0$ at baseline; loss stimulation initially decreases it because $dr/dt=-k_{\mathrm{out}}s<0$. 2. Input stimulation predicts $40(1.50)=60$ units; loss stimulation predicts $40/1.50=26.67$ units. 3. For input stimulation, the time constant is $1/0.30=3.333$ h. For loss stimulation, it is $1/[0.30(1.50)]=2.222$ h. 4. When the drug effect is zero, both models reduce to the same baseline equation with loss-rate constant $k_{\mathrm{out}}$, so both recover with time constant $1/k_{\mathrm{out}}$. 5. Examples include an unstable or drifting baseline, unknown exposure timing, a response that has not reached equilibrium but is treated as if it had, sparse or noisy early measurements, or evidence of a delayed effect or feedback not included in the basic models.