From Compartments to an Observable Concentration Curve
A two-compartment model describes drug in a central compartment, where concentration is measured and elimination occurs, and a peripheral compartment that exchanges drug with the central one. Its transfer and elimination rate constants are often called microconstants: they describe individual processes in the compartment model. After an IV bolus, however, the central concentration is observed as a sum of two exponential terms. The rates and amplitudes in that sum are the macroconstants. Connecting these parameterizations helps explain why the observed distribution and terminal phases do not directly equal the central elimination rate.
The derivation below assumes linear, first-order transfer and elimination, constant compartment volumes, and an instantaneous IV bolus placed entirely in the central compartment. There is no drug in the peripheral compartment at time zero, and elimination occurs only from the central compartment. These assumptions make the amount equations a linear, time-invariant system.
The amount equations and the measured central concentration are
The terms have direct interpretations. Central amount is lost through elimination and outward transfer, and gained through return transfer. Peripheral amount is gained through outward transfer and lost through return transfer. The two transfer flows are consistent with a common intercompartmental clearance because \(k_{12}V_1=k_{21}V_2=Q\).
Deriving the Macroconstants
For a solution proportional to \(e^{-\lambda t}\), the amount equations have a nonzero solution only when the determinant of their coefficient matrix is zero. This condition gives a quadratic equation for the two decay rates:
The roots of this quadratic are \(\alpha\) and \(\beta\). Therefore, their sum and product are
These identities link the observable decay rates to the underlying microconstants. They do not say that either observed rate equals \(k_{10}\). Drug can leave the central compartment by transfer as well as elimination, and returning drug can contribute to the later concentration.
For a dose \(D\) given as an IV bolus, the initial conditions are \(A_1(0)=D\) and \(A_2(0)=0\). Thus \(C(0)=D/V_1\). The initial slope follows directly from the central amount equation: \(dC/dt\) at time zero is \(-(k_{10}+k_{12})C(0)\). Matching these two conditions to the biexponential curve gives the amplitudes:
For positive \(k_{10}\), \(k_{12}\), and \(k_{21}\), the roots are distinct and ordered \(\alpha>k_{21}>\beta>0\). To see why \(k_{21}\) lies between the roots, evaluate the quadratic at \(k_{21}\): the result is \(-k_{12}k_{21}\), which is negative. A quadratic with positive leading coefficient is negative between its two real roots, so \(k_{21}\) lies between them. The product of the roots is positive and their sum is positive, so both roots are positive. Consequently, both amplitudes are positive under these assumptions.
Worked Example: Calculating Macroconstants From Microconstants
Hypothetical data. Suppose \(V_1=10\) L, \(V_2=20\) L, \(CL=3\) L/h, and \(Q=4\) L/h. A hypothetical \(100\)-mg IV bolus is given into the central compartment. First calculate the microconstants:
The roots have sum \(0.3000+0.4000+0.2000=0.9000\ \mathrm{h}^{-1}\) and product \(0.3000(0.2000)=0.06000\ \mathrm{h}^{-2}\). Thus
which gives \(\alpha=0.8275\ \mathrm{h}^{-1}\) and \(\beta=0.07251\ \mathrm{h}^{-1}\). The initial concentration is \(D/V_1=100/10=10.00\) mg/L. The amplitudes are
The resulting curve is \(C(t)=8.311e^{-0.8275t}+1.689e^{-0.07251t}\) mg/L when \(t\) is in hours. At time zero, its components sum to \(10.00\) mg/L, as required by the dose and central volume. The faster macro-rate is not \(k_{10}\): it reflects the coupled transfer-and-elimination system.
Interpreting the Two Exponential Phases
The larger rate \(\alpha\) produces the faster-decaying exponential and is associated with the early distribution phase. The smaller rate \(\beta\) produces the slower-decaying exponential and governs the late terminal phase. These labels describe the observed curve, not separate processes that occur in isolation. Each macro-rate depends on multiple microconstants.
When time is sufficiently late that the fast component is negligible, \(C(t)\) is approximately \(B e^{-\beta t}\). The terminal half-life is therefore \(\ln(2)/\beta\), not generally \(\ln(2)/k_{10}\). In the example above, the terminal half-life is \(0.6931/0.07251=9.559\) h, whereas the half-life associated with central elimination alone is \(0.6931/0.3000=2.310\) h. These are different quantities: the terminal behavior reflects the coupled system, including drug returning from the peripheral compartment.
Worked Example: Comparing Early and Late Contributions
Use the hypothetical curve from the previous example to compare its two components at \(t=1\) h and \(t=12\) h. At 1 h, the fast component is \(8.311e^{-0.8275(1)}=3.633\) mg/L, while the slow component is \(1.689e^{-0.07251(1)}=1.571\) mg/L. The total is \(3.633+1.571=5.204\) mg/L.
At 12 h, the fast component is \(8.311e^{-0.8275(12)}=0.000405\) mg/L, and the slow component is \(1.689e^{-0.07251(12)}=0.7076\) mg/L. The total is approximately \(0.7080\) mg/L. The slow component dominates at this late time, which is why the observed terminal slope approaches \(-\beta\). The calculation also illustrates that “terminal” is an approximation: the fast component is small, not mathematically absent.
Recovering Microconstants From Macroparameters
Under the same ideal model, the mapping can be reversed. If \(A\), \(B\), \(\alpha\), and \(\beta\) are known, then \(C(0)=A+B\). The amplitude equations show that \(k_{21}\) is the amplitude-weighted location between the two macro-rates:
Once \(k_{21}\) is found, the sum-and-product identities give \(k_{10}\) and \(k_{12}\). With a known IV bolus dose, the initial concentration also gives \(V_1\). The remaining volume and clearances follow from their definitions.
Worked Example: Recovering Microconstants From the Biexponential Curve
Hypothetical data. Suppose an ideal fitted curve after a known \(100\)-mg IV bolus is \(C(t)=8.311e^{-0.8275t}+1.689e^{-0.07251t}\) mg/L, with time in hours. The initial concentration is \(8.311+1.689=10.00\) mg/L, so \(V_1=100/10.00=10.00\) L. Using the amplitude relationship,
Then \(k_{10}=\alpha\beta/k_{21}=(0.8275)(0.07251)/0.2000=0.3000\ \mathrm{h}^{-1}\). The sum identity gives \(k_{12}=0.8275+0.07251-0.3000-0.2000=0.4000\ \mathrm{h}^{-1}\), allowing for rounding. Therefore \(CL=k_{10}V_1=3.000\) L/h. If the model also assumes common intercompartmental clearance, \(Q=k_{12}V_1=4.000\) L/h and \(V_2=Q/k_{21}=20.00\) L.
This recovery depends on the stated model and ideal macroparameters. It is not a claim that noisy concentration data will determine these values precisely; in practice, sparse or poorly timed observations can make estimates uncertain.
Assumptions, Diagnostics, and Alternatives
The algebra requires a linear two-compartment system with positive, constant microconstants and an IV bolus as specified. The formula for \(V_1\) from \(D/C(0)\) additionally requires a known dose and a meaningful concentration at time zero. A concentration curve that includes absorption, nonlinear elimination, time-varying parameters, or a different initial amount in the peripheral compartment need not have these amplitudes or this interpretation. Measurement error and sampling design affect the reliability of fitted macroparameters even when the structural model is appropriate.
Useful checks include confirming that the fitted curve has a physically sensible positive concentration, that its two components reproduce the initial concentration, and that the inferred roots and amplitudes have the expected ordering and signs. Examine whether the early samples capture the fast component and whether late samples support a terminal slope. If the fast component is poorly observed, its rate and amplitude may be weakly determined; if late observations are absent, the terminal rate may be uncertain. Structured residual patterns or systematic disagreement between observed and predicted concentrations can indicate that the assumed model does not describe the data adequately. A good numerical fit alone does not establish that the compartment interpretation is uniquely supported.
| Representation | What its parameters describe | Strength and limitation |
|---|---|---|
| Microconstant compartment model | Elimination and transfer between specified compartments | Mechanistically interpretable, but depends on the compartment structure and assumptions. |
| Biexponential macroconstant curve | Observed concentration amplitudes and exponential decay rates | Directly describes the curve, but its parameters do not individually name transfer or elimination processes. |
The macroconstant curve can be convenient for describing an observed biexponential profile, while the microconstant model makes explicit the assumed flows between compartments. Under the conditions derived here, they are related parameterizations of the same central concentration-time curve. The equivalence does not make them interchangeable for every task: the compartment model specifies how amounts move, whereas a sum of exponentials by itself does not specify a unique biological mechanism outside the stated model.
Common Pitfalls and Limits
- Equating \(\beta\) with \(k_{10}\). The terminal macro-rate is an eigenvalue of the coupled system. It is generally not the central elimination microconstant.
- Calling \(k_{12}\) and \(k_{21}\) the observed phase rates. These are directional transfer constants; \(\alpha\) and \(\beta\) depend jointly on the transfer and elimination constants.
- Ignoring units. Microconstants and macro-rates have units of inverse time, while amplitudes have concentration units. The exponential arguments, such as \(\alpha t\), must be dimensionless.
- Assuming two fitted exponentials prove two physical compartments. The curve is consistent with the specified two-compartment model under its assumptions, but the shape alone does not establish a unique mechanism.
- Overlooking degenerate cases. If outward transfer is absent, \(k_{12}=0\), the peripheral compartment is not populated from a central bolus and its mode has zero amplitude. The positive two-amplitude interpretation above assumes positive transfer and elimination rates.
These formulas are exact for the stated linear model. Their interpretation is limited by whether the model, initial conditions, and observed sampling times are appropriate for the data at hand.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions below.
- For \(k_{10}=0.20\ \mathrm{h}^{-1}\), \(k_{12}=0.30\ \mathrm{h}^{-1}\), and \(k_{21}=0.10\ \mathrm{h}^{-1}\), calculate \(\alpha\) and \(\beta\).
- In a positive two-compartment model, why does \(k_{21}\) lie between \(\beta\) and \(\alpha\)?
- A curve has a terminal macro-rate of \(0.05\ \mathrm{h}^{-1}\). Is this sufficient to conclude that \(k_{10}=0.05\ \mathrm{h}^{-1}\)? Explain.
- For the hypothetical curve \(C(t)=6e^{-0.8t}+4e^{-0.1t}\) mg/L after a known \(100\)-mg IV bolus, calculate \(V_1\) and \(k_{21}\), assuming the stated two-compartment model.
Solutions. 1. The root sum is \(0.20+0.30+0.10=0.60\ \mathrm{h}^{-1}\), and the product is \(0.20(0.10)=0.020\ \mathrm{h}^{-2}\). Thus \(\alpha,\beta=[0.60\pm\sqrt{0.60^2-4(0.020)}]/2=[0.60\pm\sqrt{0.28}]/2\), giving \(\alpha=0.5646\ \mathrm{h}^{-1}\) and \(\beta=0.03542\ \mathrm{h}^{-1}\). 2. The characteristic quadratic evaluated at \(k_{21}\) equals \(-k_{12}k_{21}<0\); therefore \(k_{21}\) lies between its two roots. 3. No. The terminal rate is \(\beta\), a rate of the coupled system; \(k_{10}\) also depends on the other model parameters and cannot be identified from \(\beta\) alone. 4. The initial concentration is \(6+4=10\) mg/L, so \(V_1=100/10=10\) L. Then \(k_{21}=0.1+[4/(6+4)](0.8-0.1)=0.3800\ \mathrm{h}^{-1}\).