Why a Two-Compartment Curve Has Two Phases
After an intravenous bolus in a linear two-compartment model, concentration in the central compartment is commonly described by the sum of two exponential terms. On a concentration-versus-time plot, the curve falls rapidly at first and then declines more slowly. The early fall is associated with distribution between compartments as well as elimination; the later, shallower decline is the terminal phase. Interpreting these phases helps describe the curve, but it does not mean that the drug first distributes and only later begins to be eliminated.
The phases are features of the combined concentration-time curve. Both exponential terms contribute at positive times, and their relative contributions change continuously. A phase name is therefore a useful description of the curve over a time range, not a claim that one isolated biological process operates in that range.
The exponential rate constants have units of inverse time, such as \(\mathrm{h}^{-1}\), and the amplitudes have the same units as concentration, such as \(\mathrm{mg/L}\). At \(t=0\), the two terms sum to the initial model concentration, \(C(0)=A+B\). The amplitudes are the extrapolated contributions of the two exponentials at time zero. They are not, in general, concentrations in the central and peripheral compartments.
The interpretation assumes a linear, time-invariant two-compartment disposition model over the times considered: the bolus is treated as instantaneous, transfer and elimination follow first-order processes, and the observed concentration is the central-compartment concentration. It also assumes that the fitted curve adequately describes the data. Nonlinear disposition, time-varying clearance, an incorrect structural model, or measurement problems can make a two-exponential interpretation misleading.
What the Slopes Mean
If a single exponential \(C(t)=K e^{-kt}\) is plotted on a logarithmic concentration axis, its logarithm is a straight line: \(\ln C(t)=\ln K-kt\). Its slope is \(-k\), and its half-life is \(\ln(2)/k\). A two-exponential curve is not a straight line on a semilog plot over its entire course, because both terms contribute. Early in the curve, the fast component may make the local slope steep; late in the curve, the slow component dominates and the slope approaches \(-\beta\).
This distinction matters: the instantaneous negative slope of the log concentration is generally neither \(\alpha\) nor \(\beta\). Differentiating the two-exponential expression gives a weighted average of the two rate constants:
The weights are the components’ shares of the total concentration at that time. The local rate is therefore between \(\beta\) and \(\alpha\) when both components are positive. As time increases, the fast component loses relative weight, so this local rate tends toward \(\beta\). This is why the terminal semilog slope, when it is genuinely supported by late observations, can be interpreted as \(-\beta\).
Worked Examples
Worked Example: Identifying a Terminal Contribution
Consider a hypothetical fitted curve, with time in hours and concentration in \(\mathrm{mg/L}\):
The fast rate is \(\alpha=0.8\ \mathrm{h}^{-1}\), and the slow rate is \(\beta=0.170059\ \mathrm{h}^{-1}\). Since \(\alpha>\beta\), the first term is the fast, distribution-associated component and the second is the slow, terminal component. At time zero the model concentration is \(6+4=10\ \mathrm{mg/L}\).
At \(t=14\) hours, the fast contribution is \(6e^{-0.8(14)}=6e^{-11.2}\approx0.00008205\ \mathrm{mg/L}\). The slow contribution is \(4e^{-0.170059(14)}=4e^{-2.380826}\approx0.369897\ \mathrm{mg/L}\). Thus, \(C(14)\approx0.369979\ \mathrm{mg/L}\). The fast contribution is negligible at this time, so the total is effectively determined by the slow terminal component.
The terminal half-life is \(t_{1/2,\mathrm{terminal}}=\ln(2)/0.170059=0.693147/0.170059\approx4.076\) hours. This is the half-life of the slow exponential. It would not describe the initial fall of the complete curve, when the fast component contributes substantially.
Worked Example: Finding When the Components Are Equal
For the hypothetical curve \(C(t)=8e^{-0.50t}+2e^{-0.15t}\), with time in hours and concentration in \(\mathrm{mg/L}\), find when the fast and slow contributions are equal. Set the two terms equal:
At this exact crossover time, each contribution is \(8e^{-0.50t}=8e^{-0.50[\ln(4)/0.35]}\approx1.104\ \mathrm{mg/L}\). The slow term has the same value by the crossover equation, so the total concentration is approximately \(2(1.104)=2.208\ \mathrm{mg/L}\). The crossover is a point at which neither component is negligible; it does not mark an abrupt biological change from distribution to terminal decline.
The half-lives of the individual exponentials are \(0.693147/0.50=1.386\) hours for the fast component and \(0.693147/0.15=4.621\) hours for the slow component. At the crossover, the total concentration does not follow either half-life alone, because both terms contribute equally.
Worked Example: Interpreting the Local Semilog Slope
Use \(C(t)=8e^{-0.50t}+2e^{-0.15t}\) again. At \(t=4\) hours, the fast contribution is \(8e^{-2}=1.082682\ \mathrm{mg/L}\), and the slow contribution is \(2e^{-0.6}=1.097623\ \mathrm{mg/L}\). Therefore, \(C(4)\approx2.180306\ \mathrm{mg/L}\).
The local negative slope of the log concentration is
This value lies between \(0.15\) and \(0.50\ \mathrm{h}^{-1}\), as expected. The components are almost equal at four hours, so the local slope is not yet the terminal rate \(0.15\ \mathrm{h}^{-1}\). A single straight-line fit to a narrow part of this transition could give a slope near \(0.3238\ \mathrm{h}^{-1}\), but that would not establish the terminal rate.
Why the Two Phases Matter in Practice
The fast phase describes the steep early portion of the central concentration curve, where redistribution between central and peripheral spaces has an important influence. It is often called the distribution phase, but the fast exponential is not a direct measurement of a transfer rate or a separate “distribution concentration.” The slow phase describes the tail after the fast contribution has become small. Its rate constant \(\beta\) is the terminal rate constant for the biexponential curve.
Both phases reflect the joint behavior of the two-compartment system. In particular, elimination does not stop during the distribution phase, nor does distribution necessarily stop during terminal decline. The rate constants and amplitudes are curve-level quantities that arise from the coupled model. Their relationship to individual transfer and elimination parameters is a separate parameterization question.
The component crossover can be useful when describing where one term becomes larger than the other. For positive amplitudes, set \(A e^{-\alpha t}=B e^{-\beta t}\). If \(A>B\), the crossover time is
This expression is meaningful as a nonnegative time under the stated condition \(A\geq B\). If \(A=B\), the terms are equal at time zero; if \(A<B\), the slow term already exceeds the fast term at time zero and there is no positive-time crossover. In either case, component equality is a mathematical feature of the fitted curve, not a biological boundary.
Diagnostics, Alternatives, and Limitations
A semilog plot is a useful diagnostic, but it should be read with the sampling schedule and model assumptions in mind. A terminal rate is poorly supported if late samples are too sparse, too close to the assay’s quantification limit, or do not show a stable approximately straight tail. A distribution-associated phase may also be difficult to identify if there are few sufficiently early samples. In these situations, the fitted rates and amplitudes can be highly uncertain, even if the overall curve appears to fit reasonably well.
Curvature in a region expected to be approximately linear on the semilog plot can signal that both components still contribute, that observations are noisy, or that the assumed model does not describe the data well. It is not, by itself, proof of any one explanation. Check residual patterns and the broader model diagnostics, and avoid assigning a phase from a single observation or an arbitrary time cut-off. The appropriate interpretation depends on the full time course and the adequacy of the model.
| Approach | What it describes | Main caution |
|---|---|---|
| Interpret the full biexponential curve | Both fitted components and their changing contributions over time | Requires data that support estimating both components and a suitable two-compartment model |
| Fit a terminal straight line on a semilog plot | The late slope, interpreted as \(-\beta\) when the fast contribution is negligible | A short, noisy, or poorly sampled tail can give an unreliable terminal rate |
| Use a single-exponential description | One overall decline rate over a specified range | Cannot represent a genuine two-phase curve across the full time course |
A terminal straight-line estimate can be helpful as a descriptive check, but it should not replace interpretation of the full curve when both phases matter. Conversely, a two-exponential model is not automatically justified simply because it can fit a data set: the observations need to support the extra component, and diagnostics should not show systematic lack of fit. Measurements restricted to late times may inform the terminal decline while revealing little about the distribution phase.
Common Mistakes
- Calling \(\alpha\) the elimination rate. The fast rate describes an exponential component of the coupled system; it is not generally an individual elimination rate.
- Equating the slow phase with elimination alone. The terminal rate is the rate of the slow component in the two-compartment curve, not proof that distribution has ceased or that only one biological process remains.
- Reading amplitudes as compartment concentrations. \(A\) and \(B\) are exponential amplitudes; their sum gives the model’s time-zero central concentration, but neither amplitude is automatically a physical compartment concentration.
- Using the terminal half-life for the whole curve. The total curve has a changing local slope. The terminal half-life applies when the slow term dominates.
- Assuming the semilog plot must be straight throughout. A biexponential curve generally has a changing slope while both terms contribute; its late tail approaches a straight line only under suitable conditions.
Check Your Understanding
Try each question, then compare your reasoning with the brief solutions.
- For \(C(t)=5e^{-0.60t}+3e^{-0.12t}\), which component is fast, and what is the terminal half-life?
- For \(C(t)=6e^{-0.40t}+2e^{-0.10t}\), calculate the component crossover time, if it occurs at a positive time.
- Why is the instantaneous negative slope of \(\ln C(t)\) generally not equal to either exponential rate constant while both components contribute?
- What would make an estimate of the terminal rate questionable when examining a semilog concentration-time plot?
Solutions. (1) The \(5e^{-0.60t}\) term is fast because \(0.60>0.12\). The terminal half-life is \(\ln(2)/0.12=0.693147/0.12\approx5.776\) time units. (2) Since \(6>2\), a positive-time crossover occurs: \(t=\ln(6/2)/(0.40-0.10)=\ln(3)/0.30\approx3.662\) time units. (3) Differentiation gives a contribution-weighted average of the two rates; when both terms are positive, that average lies between them. It approaches the slow rate only as the fast contribution becomes negligible. (4) Sparse late observations, substantial noise, measurements near the quantification limit, or a tail that does not show a stable approximately straight segment all weaken the terminal-rate interpretation.