Starting Values for Multicompartment Models

Why Two-Compartment Starting Values Need More Than a Half-Life

A two-compartment model has more disposition parameters than a one-compartment model, and its concentration-time profile changes shape over time. Early concentrations mainly inform the rapid distribution phase; later concentrations show the slower terminal phase. Exposure informs clearance, while the profile’s scale and shape help separate central volume, peripheral volume, and intercompartmental clearance. A terminal half-life by itself cannot provide that separation.

The aim here is to build plausible starting values for a linear two-compartment model, not to claim that hand calculations replace fitting the model to individual data. The calculations are most direct for an intravenous (IV) bolus with known dose and complete concentration-time information. For an infusion, extravascular dose, or sparse sampling, the same ideas can guide initialization, but the observation process and available information must be considered explicitly.

Model and Assumptions

Consider a mammillary model with a central compartment, where concentration is measured, and a peripheral compartment. Drug moves between compartments with clearance \(Q\); elimination occurs from the central compartment with clearance \(CL\). Let \(V_1\) and \(V_2\) denote central and peripheral volumes. The model assumes linear disposition, constant parameters, and a single dose entering the central compartment for the IV bolus formulas below.

Definition: The micro-rate constants are \(k_{10}=CL/V_1\), \(k_{12}=Q/V_1\), and \(k_{21}=Q/V_2\). Following an IV bolus of dose \(D\), the central concentration has the form \(C(t)=A e^{-\alpha t}+B e^{-\beta t}\), where \(\alpha>\beta>0\) are hybrid (macro) rate constants and \(A\) and \(B\) are their concentration coefficients. In the usual positive-parameter model, \(A\) and \(B\) are positive.

The hybrid rates are not the micro-rate constants. They summarize the two exponential phases of the observable concentration curve. Their relationships to the micro-rate constants follow from the model’s two-compartment differential equations:

\alpha+\beta=k_{10}+k_{12}+k_{21}, \qquad \alpha\beta=k_{10}k_{21}, \qquad A+B=\frac{D}{V_1}.

The last equality is the concentration immediately after the bolus: \(C(0)=A+B=D/V_1\). These relationships assume the bolus is delivered into the central compartment and that the measured concentration represents that compartment. A delay in sampling after the bolus can make \(C(0)\) an extrapolation rather than an observed value.

From Biexponential Shape to Model Parameters

If a biexponential curve has been estimated from a sufficiently informative IV bolus profile, its coefficients and rates can be converted to micro-rate constants. The coefficient-weighted rate identity is

k_{21}=\frac{A\beta+B\alpha}{A+B}.

It follows by differentiating the concentration at time zero. From the model, the initial decline is \(C'(0)=-(k_{10}+k_{12})C(0)\); from the biexponential expression it is \(C'(0)=-(A\alpha+B\beta)\). Equating these expressions and using the sum and product relationships above gives the stated identity. Then \(k_{10}=\alpha\beta/k_{21}\), and \(k_{12}=\alpha+\beta-k_{10}-k_{21}\). With \(V_1=D/(A+B)\), the clearances and peripheral volume follow from \(CL=k_{10}V_1\), \(Q=k_{12}V_1\), and \(V_2=Q/k_{21}\).

Worked Example: Converting a Biexponential Profile

Suppose a hypothetical IV bolus dose of 100 mg has a well-sampled profile summarized by \(A=8.00\) mg/L, \(B=2.00\) mg/L, \(\alpha=1.20\) h\(^{-1}\), and \(\beta=0.120\) h\(^{-1}\). Assume the linear two-compartment model is an appropriate description.

First, \(V_1=D/(A+B)=100\text{ mg}/(10.00\text{ mg/L})=10.0\) L. The rate constants are

k_{21}=\frac{(8.00)(0.120)+(2.00)(1.20)}{8.00+2.00} =\frac{3.36}{10.00}=0.336\text{ h}^{-1},

and

k_{10}=\frac{(1.20)(0.120)}{0.336} =0.4286\text{ h}^{-1}, \qquad k_{12}=1.20+0.120-0.4286-0.336 =0.5554\text{ h}^{-1}.

Therefore, \(CL=(0.4286\text{ h}^{-1})(10.0\text{ L})=4.286\) L/h, \(Q=(0.5554\text{ h}^{-1})(10.0\text{ L})=5.554\) L/h, and \(V_2=5.554\text{ L/h}/0.336\text{ h}^{-1}=16.53\) L. As a consistency check, the model-implied AUC is \(A/\alpha+B/\beta=8.00/1.20+2.00/0.120=23.33\) mg·h/L. Thus \(D/\mathrm{AUC}=100/23.33=4.286\) L/h, matching the clearance calculated from the rate constants.

Exposure, Terminal Slope, and What They Identify

For a complete IV concentration-time profile under linear disposition, exposure provides an independent clearance estimate: \(CL=D/\mathrm{AUC}\). The terminal slope can be estimated from two concentrations known to lie in the terminal phase. If the times are \(t_1\) and \(t_2\), with \(t_2>t_1\), then \(\beta\) is estimated as the negative slope of log concentration against time:

\hat{\beta}=\frac{\ln C(t_1)-\ln C(t_2)}{t_2-t_1}, \qquad t_{1/2,\mathrm{terminal}}=\frac{\ln 2}{\hat{\beta}}.

These estimates can help set a terminal rate and clearance, but they do not, on their own, determine \(V_1\), \(V_2\), and \(Q\). Many combinations of compartmental parameters can produce a similar terminal slope. Early profile information or additional exposure summaries are needed to constrain the distribution and volume terms.

Worked Example: Estimating Clearance and the Terminal Rate

In a hypothetical IV study, a 120 mg dose yields an estimated complete-profile AUC of 30.0 mg·h/L. Two late concentrations are 0.600 mg/L at 18 h and 0.150 mg/L at 30 h. Assume both observations are in the terminal log-linear phase and that the dose and AUC refer to the same linear disposition process.

The clearance estimate is \(CL=120\text{ mg}/(30.0\text{ mg·h/L})=4.00\) L/h. The terminal rate estimate is

\hat{\beta}=\frac{\ln(0.600)-\ln(0.150)}{30-18} =\frac{\ln(4)}{12} =0.1155\text{ h}^{-1}.

The associated terminal half-life is \(0.6931/0.1155=6.00\) h. These calculations provide useful starting information for clearance and the slow phase. They do not determine the central volume or the exchange clearance: no early distribution shape or initial concentration was supplied. Assigning \(V_1\), \(V_2\), and \(Q\) from these two summaries alone would add assumptions not supported by the stated data.

Using Concentration-Time Moments to Seed Volumes and Q

When a complete IV profile is available, the area under the first moment curve, AUMC, can provide another useful summary. For the biexponential profile, \(\mathrm{AUC}=A/\alpha+B/\beta\) and \(\mathrm{AUMC}=A/\alpha^2+B/\beta^2\). The mean residence time is \(\mathrm{MRT}=\mathrm{AUMC}/\mathrm{AUC}\), and for this linear IV bolus model the steady-state volume of distribution is \(V_{ss}=CL\cdot\mathrm{MRT}=V_1+V_2\). This identity is a moment relationship, not a claim that all drug resides in the central compartment.

A practical route is to estimate \(V_1=D/C(0)\), obtain \(CL\) from dose divided by AUC, calculate \(V_{ss}\), and then set \(V_2=V_{ss}-V_1\). If \(\beta\), \(V_1\), \(V_2\), and \(CL\) are available, \(Q\) can be calculated from the characteristic equation for the two hybrid rates. Substituting \(k_{10}=CL/V_1\), \(k_{12}=Q/V_1\), and \(k_{21}=Q/V_2\) into the equation for \(\beta\) and rearranging gives:

\beta^2-\beta\left(\frac{CL}{V_1}+ Q\left(\frac{1}{V_1}+\frac{1}{V_2}\right)\right) +\frac{CL}{V_1}\frac{Q}{V_2}=0,

so that

Q= \frac{\beta^2-\beta(CL/V_1)} {\beta(1/V_1+1/V_2)-(CL/V_1)/V_2}.

This calculation is useful only when the required summaries are credible and the resulting parameter values are positive and coherent. It can be numerically unstable when the denominator is near zero, and moment estimates can be sensitive to inadequate sampling or extrapolation of the terminal tail.

Worked Example: Combining Exposure, Moments, and the Terminal Phase

Consider a hypothetical 120 mg IV bolus profile adequately summarized by \(C(0)=12.0\) mg/L, AUC \(=30.0\) mg·h/L, and AUMC \(=225\) mg·h²/L. Suppose the terminal portion is consistent with concentrations 0.2953 mg/L at 20 h and 0.0400 mg/L at 40 h. Assume a linear two-compartment model and sufficiently informative sampling for these summaries to be meaningful.

The central volume is \(V_1=120/12.0=10.0\) L, and \(CL=120/30.0=4.00\) L/h. The mean residence time is \(225/30.0=7.50\) h, giving \(V_{ss}=(4.00\text{ L/h})(7.50\text{ h})=30.0\) L. Thus \(V_2=30.0-10.0=20.0\) L.

The terminal rate estimated from the two late concentrations is

\hat{\beta}=\frac{\ln(0.2953/0.0400)}{40-20} \approx\frac{\ln(7.383)}{20} \approx0.1000\text{ h}^{-1}.

Using \(CL/V_1=4.00/10.0=0.400\) h\(^{-1}\), the rearranged characteristic equation gives

Q= \frac{(0.1000)^2-(0.1000)(0.400)} {(0.1000)(1/10.0+1/20.0)-0.400/20.0} = \frac{-0.0300}{-0.00500} =6.00\text{ L/h}.

These starting values imply \(k_{21}=Q/V_2=0.300\) h\(^{-1}\) and \(k_{12}=Q/V_1=0.600\) h\(^{-1}\). They are internally consistent with the supplied terminal rate: the two hybrid rates have sum \(0.400+0.600+0.300=1.300\) h\(^{-1}\) and product \(0.400\times0.300=0.120\) h\(^{-2}\), whose roots are 1.20 and 0.100 h\(^{-1}\). The initial concentration, exposure, moments, and terminal slope therefore support one coherent set of plausible values: \(CL=4.00\) L/h, \(V_1=10.0\) L, \(V_2=20.0\) L, and \(Q=6.00\) L/h.

Diagnostics, Alternatives, and Limitations

Treat calculated values as a consistency check and a way to choose a reasonable neighborhood for initialization. Before using them, check units, dose and concentration alignment, sampling times, and whether terminal observations actually follow a log-linear phase. A very early sample is important for estimating \(C(0)\) and distribution behavior; without one, \(V_1\) may depend heavily on extrapolation. AUC and AUMC require adequate coverage of the profile and a defensible terminal-tail extrapolation.

Check positivity and identities. A negative \(Q\), \(V_2\), or rate constant is incompatible with the intended positive-parameter model. Recheck the summaries, units, and algebra rather than forcing an invalid value into an optimizer.
Inspect the profile shape. Curvature in a proposed terminal log-concentration plot suggests the selected points may still include distribution, or may be dominated by noise. A slope from that region is not a reliable \(\beta\).
Account for the route and input process. With extravascular dosing, absorption can influence the observed slope; the IV bolus formulas cannot be applied to the observed profile as if it began in the central compartment. Unknown bioavailability also affects interpretation of dose-based exposure calculations.
Do not overinterpret sparse data. If there are few early observations, the distribution phase and central volume may be weakly informed. A plausible numerical answer from a formula does not make the underlying parameters identifiable from the data.

A one-compartment approximation is simpler and may be a useful comparison when the observed profile does not support a distinct distribution phase. It cannot represent a clearly biexponential decline. Directly fitting the full two-compartment model uses the concentration-time observations rather than only summary quantities, but remains sensitive to poor initial values and inadequate sampling. The calculation routes in this tutorial are complementary: biexponential coefficients use the profile’s shape directly, while AUC and moments provide exposure and volume constraints. Neither route removes the need to assess model fit, data quality, and plausibility.

Common mistake: treating the terminal half-life as though it were \(0.693\,V_1/CL\), the one-compartment relationship. In a two-compartment model, the terminal rate is a hybrid rate determined by elimination and exchange; in general, \(\beta\ne CL/V_1\). Use the two-compartment relationships, and use the terminal slope as one constraint rather than as a substitute for the full model.

Key takeaway: Build two-compartment starting values from complementary information: concentration scale for \(V_1\), AUC for \(CL\), profile shape or terminal slope for hybrid rates, and distribution or moment information for \(V_2\) and \(Q\). Check that the resulting parameters reproduce the supplied summaries before using them as initial estimates.

Check Your Understanding

For each question, work from the stated assumptions. Brief solutions follow the questions.

For an IV bolus profile with \(A=6.0\) mg/L, \(B=2.0\) mg/L, \(\alpha=0.90\) h\(^{-1}\), and \(\beta=0.10\) h\(^{-1}\), what is \(V_1\) for a 160 mg dose?
Concentrations in a terminal phase are 0.80 mg/L at 10 h and 0.20 mg/L at 22 h. Estimate \(\beta\) and the terminal half-life.
For a linear IV bolus study, dose is 90 mg and AUC is 15 mg·h/L. What is the clearance estimate?
Why are terminal concentrations alone insufficient to determine \(Q\), \(V_1\), and \(V_2\)? Name one additional source of information that would help.

Solutions. (1) \(V_1=D/(A+B)=160\text{ mg}/(8.0\text{ mg/L})=20.0\) L. (2) \(\beta=\ln(0.80/0.20)/(22-10)=\ln(4)/12=0.1155\) h\(^{-1}\), so the terminal half-life is \(0.6931/0.1155=6.00\) h. (3) \(CL=D/\mathrm{AUC}=90/15=6.00\) L/h. (4) A terminal slope constrains the slow hybrid rate but does not uniquely specify the volumes and exchange clearance. Early distribution-phase concentrations, an initial concentration estimate, or suitable AUC and AUMC summaries add information needed to constrain those parameters.