Why a Baseline Value Does Not Reveal Both Turnover Rates
An untreated baseline is often the first measurement available in an indirect response model. It is useful for anchoring the response scale, but it does not, by itself, reveal how quickly the response is produced or removed. A high baseline could reflect rapid production paired with rapid loss, or slower production paired with slower loss. The equilibrium value alone cannot distinguish those explanations.
Consider a response \(R(t)\), such as a biomarker concentration, governed in the absence of drug by constant input and first-order loss. The input rate is \(k_{\mathrm{in}}\), and the loss rate constant is \(k_{\mathrm{out}}\). The model and its units are:
If \(R\) is measured in concentration units, \(k_{\mathrm{in}}\) has units of concentration per time, while \(k_{\mathrm{out}}\) has units of inverse time. The term \(k_{\mathrm{out}}R(t)\) is therefore a loss rate in concentration per time.
This relationship is the central identification result: an exact untreated equilibrium value determines the ratio \(k_{\mathrm{in}}/k_{\mathrm{out}}\), but not either rate separately. “Identifies” here means that the observation uniquely determines a parameter or function of parameters under the model. The baseline identifies one function of two unknown rates, leaving multiple rate pairs compatible with the same value.
What the Equilibrium Equation Can and Cannot Determine
The equilibrium relationship follows directly from the model. At a stable untreated baseline, set the derivative to zero and solve:
For any positive choice of \(k_{\mathrm{out}}\), choosing \(k_{\mathrm{in}}=R_0k_{\mathrm{out}}\) produces the same equilibrium. Thus, the set of compatible rates lies along a curve, not at a unique pair. The equilibrium equation supplies one constraint for two unknowns.
The same conclusion can be expressed through the sensitivity of the model prediction to the parameters. If the only predicted observation is \(R_0=k_{\mathrm{in}}/k_{\mathrm{out}}\), its derivatives with respect to the two rates are \(1/k_{\mathrm{out}}\) and \(-k_{\mathrm{in}}/k_{\mathrm{out}}^2\). This is one sensitivity row: the baseline prediction changes along only one parameter direction. It cannot provide two independent pieces of information about the two rates.
This is a structural property of the model and observation design, not a problem that can be fixed merely by collecting more measurements of the same kind. Repeated untreated measurements can improve precision for the equilibrium response, but if they all represent the same equilibrium, they still inform only the ratio. To separate the rates, the data must include information that responds differently to \(k_{\mathrm{in}}\) and \(k_{\mathrm{out}}\).
Worked Example: Many Rate Pairs, One Baseline
Suppose a hypothetical biomarker has a stable untreated baseline of \(12.0\) concentration units. Find several production-loss rate pairs compatible with that baseline.
At equilibrium, \(k_{\mathrm{in}}=12.0k_{\mathrm{out}}\). If \(k_{\mathrm{out}}=0.10\ \mathrm{day}^{-1}\), then
The pair \((1.20\ \mathrm{concentration\ units/day},\,0.10\ \mathrm{day}^{-1})\) predicts \(R_0=1.20/0.10=12.0\) concentration units. Choosing instead \(k_{\mathrm{out}}=0.25\ \mathrm{day}^{-1}\) gives \(k_{\mathrm{in}}=3.00\ \mathrm{concentration\ units/day}\), and \(3.00/0.25=12.0\) concentration units as well.
Both pairs, and infinitely many others, fit the same baseline exactly. The baseline supports the rate ratio \(12.0\) concentration units; it does not tell us which pair is correct.
How Additional Observations Resolve the Ambiguity
A useful additional observation must contribute information beyond the equilibrium ratio. One option is a response measured away from equilibrium, such as during recovery from a known perturbation. Under the untreated model, the rate of return to baseline depends on \(k_{\mathrm{out}}\). Once \(k_{\mathrm{out}}\) is estimated from that time course, \(k_{\mathrm{in}}\) follows from the baseline relation \(k_{\mathrm{in}}=k_{\mathrm{out}}R_0\).
To see why, suppose the response begins at \(R(0)\) and then follows the untreated model. Subtract the equilibrium value from both sides of the differential equation. Since \(k_{\mathrm{in}}=k_{\mathrm{out}}R_0\), the deviation \(R(t)-R_0\) satisfies a first-order decay equation. Separating variables and integrating from time \(0\) to \(t\) gives:
For a nonzero initial deviation, a time-course observation therefore carries information about the exponential rate \(k_{\mathrm{out}}\), unlike another observation made only at equilibrium. This inference requires the specified untreated dynamics to apply during the observed interval and requires the baseline \(R_0\) to be known or estimable.
Worked Example: Estimating Loss From a Recovery Time Course
In a hypothetical experiment, the untreated baseline is \(R_0=12.0\ \mathrm{mg/L}\). After a transient perturbation ends, the response is \(18.0\ \mathrm{mg/L}\) at time zero and \(14.7\ \mathrm{mg/L}\) four days later. Assume the response follows the constant-input, first-order-loss model during recovery.
The recovery solution gives \(R(4)-R_0=\{R(0)-R_0\}e^{-4k_{\mathrm{out}}}\). Substituting the values:
Taking natural logarithms,
The corresponding input rate is obtained from the baseline:
These estimates use both the equilibrium level and a nonequilibrium recovery measurement. The recovery data add rate information because the decline in the deviation from baseline depends on \(k_{\mathrm{out}}\). The calculation would not be justified if the response were still affected by drug, the input changed during recovery, or the one-compartment first-order-loss description were inadequate.
A second route is an independent measurement of one rate. For example, if an appropriate experiment or external measurement supplies \(k_{\mathrm{out}}\), the baseline determines \(k_{\mathrm{in}}\) by multiplication. Conversely, knowing \(k_{\mathrm{in}}\) permits calculation of \(k_{\mathrm{out}}=k_{\mathrm{in}}/R_0\). The independent measurement must correspond to the same biological system and conditions as the baseline; an unrelated rate value does not resolve the model’s ambiguity.
Worked Example: Combining Baseline With an Independent Loss Estimate
Suppose a hypothetical analyte has a stable untreated baseline of \(8.00\ \mathrm{ng/mL}\). A separate, suitable turnover measurement estimates its loss rate as \(0.150\ \mathrm{day}^{-1}\). Calculate the input rate and check the implied equilibrium.
Using \(k_{\mathrm{in}}=k_{\mathrm{out}}R_0\):
Substitution back into the equilibrium formula gives \(R_0=k_{\mathrm{in}}/k_{\mathrm{out}}=(1.20\ \mathrm{ng/mL/day})/(0.150\ \mathrm{day}^{-1})=8.00\ \mathrm{ng/mL}\), matching the observed baseline. In this case, the independent loss-rate information supplies the second constraint needed to calculate both rates.
Assumptions, Diagnostics, and Limits
The baseline relationship is valid only when its assumptions are credible. The observed level must represent an untreated equilibrium for the system being modeled. Input must be constant over the relevant period, loss must be proportional to the response, and there must not be an omitted process that changes the equilibrium. The response measurement must also be on the modeled scale, with units and assay behavior understood.
A single baseline value does not establish that equilibrium has been reached. Repeated measurements over time can help assess stability: a persistent trend suggests that a constant baseline may be inappropriate. Such measurements can also reveal measurement variability, but repeated values near a stable level still do not separate the two rates unless they include informative dynamics or are paired with independent rate information.
For dynamic data, examine whether deviations from baseline follow the predicted exponential pattern. For example, under the model, the logarithm of the absolute deviation, \(\log|R(t)-R_0|\), should change linearly with time at slope \(-k_{\mathrm{out}}\), when the deviation is nonzero and measurements are sufficiently informative. Curvature, systematic residual patterns, or different apparent slopes across intervals may indicate changing input, drug effects, non-first-order loss, multiple response processes, or an incorrect baseline. A fitted rate should not be interpreted mechanistically without checking such departures.
The main alternatives provide different kinds of information. More equilibrium observations improve the estimate of \(R_0\); a nonequilibrium time course can inform a turnover rate through its temporal shape; an independent measurement can supply one rate directly. These approaches are complementary, but their validity depends on design and assumptions. A time course with measurements too close together, too noisy, or all near equilibrium may contain little practical information about the recovery rate even though the rate is identifiable in principle from the model. Likewise, an independent rate estimate with substantial uncertainty will pass that uncertainty into the calculated second rate.
A common mistake is to treat a precise baseline estimate as if it were a precise estimate of both \(k_{\mathrm{in}}\) and \(k_{\mathrm{out}}\). Precision about \(R_0\) only narrows the ratio. It does not select one point from the continuum \(k_{\mathrm{in}}=R_0k_{\mathrm{out}}\). The correction is to distinguish the equilibrium quantity that the data directly identify from the separate rates that require additional information.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions.
- A stable baseline is \(5.0\ \mathrm{units}\). What relationship must hold between \(k_{\mathrm{in}}\) and \(k_{\mathrm{out}}\), and can the baseline alone determine both?
- Why do ten repeated measurements at the same untreated equilibrium not, by themselves, resolve the two rates?
- A response returns toward baseline after a perturbation. Under the stated model, which rate governs the exponential decline of its deviation from baseline?
- Suppose \(R_0=10.0\ \mathrm{mg/L}\) and an independent measurement gives \(k_{\mathrm{out}}=0.0800\ \mathrm{day}^{-1}\). Calculate \(k_{\mathrm{in}}\) with units.
- During a proposed recovery period, the response continues to drift in one direction and does not show a stable reference level. What assumption should be questioned before using the baseline formula?
Solutions. (1) \(k_{\mathrm{in}}=5.0k_{\mathrm{out}}\), with compatible units. This is one constraint on two unknown rates, so the baseline does not determine both. (2) Repeated equilibrium measurements can improve estimation of the common baseline but still measure only \(k_{\mathrm{in}}/k_{\mathrm{out}}\); they do not add a distinct dynamic or independent rate constraint. (3) The deviation follows \(R(t)-R_0=\{R(0)-R_0\}e^{-k_{\mathrm{out}}t}\), so its exponential rate is \(k_{\mathrm{out}}\). (4) \(k_{\mathrm{in}}=(10.0\ \mathrm{mg/L})(0.0800\ \mathrm{day}^{-1})=0.800\ \mathrm{mg/L/day}\). (5) The assumption that the measured level is an untreated equilibrium should be questioned. A changing baseline or input, residual drug effect, or other evolving process could invalidate the equilibrium relationship.