Deriving the Baseline Response at Equilibrium

Why the Baseline Is a Rate Balance

In an indirect response model, the response is maintained by production and reduced by loss. Even when no drug is present, these processes continue. The untreated baseline is therefore not a response with no underlying activity; it is the value at which the ongoing production and loss balance. This balance gives baseline observations a mechanistic interpretation, while also limiting what those observations can tell us about the separate rates.

The earlier material on state equations for indirect response models establishes the production-and-loss description. Here we derive its untreated equilibrium and examine what that equation implies for interpreting baseline data. The derivation is conditional: it applies when production and loss follow the stated model and the untreated response has reached equilibrium.

The corresponding untreated state equation is

The production term \(k_{\mathrm{in}}\) adds response per unit time. The loss term \(k_{\mathrm{out}}R(t)\) removes response at a rate proportional to the amount currently present. Thus \(k_{\mathrm{out}}\) has units of inverse time, while \(k_{\mathrm{in}}\) has units of response per time. Their ratio has units of response, as a baseline response must.

Deriving the Equilibrium Baseline

An equilibrium is a state that does not change over time. For the untreated response, this requires \(dR(t)/dt=0\). Setting the state equation's right-hand side to zero gives the rate-balance equation:

If \(k_{\mathrm{out}}>0\), rearranging yields the unique equilibrium:

This result does not say that either rate is zero at baseline. Instead, at equilibrium the production rate equals the loss rate:

For example, if \(R_0\) is 10 response units and \(k_{\mathrm{out}}\) is \(0.2\ \mathrm{h}^{-1}\), the implied production rate is \(2\) response units per hour. Both processes are active, but their net contribution to the response is zero.

The equilibrium is stable under the model when \(k_{\mathrm{out}}>0\). To see why, define the deviation from equilibrium as \(D(t)=R(t)-R_0\). Subtract the equilibrium balance \(k_{\mathrm{in}}=k_{\mathrm{out}}R_0\) from the state equation:

The solution is \(D(t)=D(0)e^{-k_{\mathrm{out}}t}\). Because \(e^{-k_{\mathrm{out}}t}\) decreases toward zero when \(k_{\mathrm{out}}>0\), deviations diminish and the response approaches \(R_0\). If \(k_{\mathrm{out}}=0\), a positive constant production rate has no finite equilibrium; if \(k_{\mathrm{out}}<0\), deviations do not decay. These cases do not support the stable baseline interpretation used here.

What Baseline Observations Constrain

The equilibrium equation directly constrains a ratio, not both rates separately. Rearranging it gives

If a baseline is known exactly, any positive pair of rates with this ratio gives the same equilibrium value. Multiplying both rates by the same positive constant changes the speed at which deviations return toward baseline, but leaves the ratio unchanged. A single equilibrium observation therefore cannot, by itself, determine the two rates separately.

With measured rather than exact baselines, the inference is less direct. Repeated untreated observations can help estimate a typical baseline, but their spread may reflect measurement error, within-person fluctuation, or genuine change over time. A sample average is not automatically the equilibrium value: it is a useful estimate of \(R_0\) only to the extent that the observations represent the same stable untreated condition and the measurement process is adequately characterized.

Assumptions and Practical Checks

The ratio \(R_0=k_{\mathrm{in}}/k_{\mathrm{out}}\) is valid when the following conditions are reasonable over the baseline period:

• The response follows constant production and first-order loss, so its rate is \(k_{\mathrm{in}}-k_{\mathrm{out}}R(t)\).
• The production and loss rates are stable over the period being interpreted; important time-varying influences are absent or accounted for.
• The observations represent an untreated condition, without a relevant prior intervention still affecting the response.
• The response is at, or sufficiently close to, equilibrium that its mean rate of change is approximately zero.
• \(k_{\mathrm{out}}>0\), giving a finite and stable equilibrium for nonnegative production.

These are modeling assumptions, not facts established merely because a value is labelled “baseline.” Useful checks include plotting serial untreated observations against time, checking for persistent upward or downward trends, comparing measurements across relevant times or conditions, and examining whether replicate measurements vary more than expected from the observation process. A marked trend, systematic cycling, or abrupt shift warns that a single constant \(R_0\) may not describe the data.

A baseline taken soon after a prior intervention may also fail to represent the untreated equilibrium. In that situation, the response can still be moving even if no intervention is being administered at the time of measurement. The relevant question is whether the system has returned to the stable untreated condition, not simply whether treatment is currently absent.

Mechanistic Baseline Versus Direct Estimation

There are two useful but different approaches. A direct empirical approach estimates baseline from untreated response measurements, for example by summarizing stable repeated observations. A mechanistic approach uses \(R_0=k_{\mathrm{in}}/k_{\mathrm{out}}\) to relate that baseline to production and loss. The direct approach can provide a clear data-based summary without requiring the rate model to be correct. The mechanistic approach makes the rate balance explicit and can connect baseline information to a dynamic model, but depends on the stated assumptions.

The approaches need not compete. An empirical baseline estimate can be compared with the ratio implied by proposed rates. Disagreement may indicate parameter inconsistency, measurement or sampling issues, failure to reach equilibrium, or a misspecified production-and-loss structure. Agreement is a useful consistency check, but does not prove that the two rates are separately correct: many rate pairs can have the same ratio.

The equation also has limits. If production varies with time, loss is not proportional to response, or other processes contribute to the state equation, the equilibrium must be derived from that expanded model's zero-rate condition. The simple ratio should not be carried over automatically. Even within the simple model, uncertain baseline observations constrain the ratio with uncertainty rather than fixing it exactly.

Common Mistake

Wrong reasoning: “The baseline is \(R_0\), so the production rate must equal \(R_0\), and the loss rate must be zero.”

Correction: \(R_0\) is a response amount, while \(k_{\mathrm{in}}\) is a production rate and \(k_{\mathrm{out}}\) is a loss rate constant. Their units differ. At equilibrium, the production rate equals the loss rate evaluated at the baseline: \(k_{\mathrm{in}}=k_{\mathrm{out}}R_0\). Neither process must be zero.