Intravenous Infusion Pharmacokinetics

Why Infusion Records Matter

An intravenous infusion supplies drug over a recorded interval rather than instantaneously. The concentration profile therefore depends on both the drug’s disposition and the time-varying input. A model that treats a two-hour infusion as a bolus can predict the wrong peak and the wrong early decline, even if its elimination parameters are otherwise reasonable. Fitting concentration data together with accurate infusion start times, stop times, and rates is essential to interpret the profile.

This tutorial uses linear one- and two-compartment models to show how infusion input enters a prediction, how distribution differs from the terminal phase, and how clearance, volume, and exposure can be compared across dose cohorts. All data and parameter values below are hypothetical. The main assumptions are linear, time-invariant disposition over the observation period; known delivered dose and infusion timing; and concentration measurements that are appropriately timed and quantified. Model parameters are interpretable only to the extent that the sampling and model structure support them.

Concentration During and After an Infusion

For a one-compartment model with constant clearance and volume, define \(k=CL/V\). If infusion begins at time zero at constant rate \(R_0\) and stops at duration \(T\), then the concentration during the infusion and after it stops is:

The concentration approaches \(R_0/CL\) if an infusion continues long enough, but a finite infusion generally ends before that limiting concentration is reached. After the infusion, the one-compartment prediction declines with rate constant \(k\). This model is useful when a single exponential adequately describes the observed disposition; it cannot represent a distinct distribution phase.

In a two-compartment model, a short input is followed by two exponential components. Let \(\alpha\) denote the faster hybrid rate constant and \(\beta\) the slower terminal rate constant, with \(\alpha>\beta>0\). For constant infusion from zero to \(T\), the post-infusion concentration at \(t>T\) can be written:

The coefficients arise from the central-compartment response to an input accumulated over the infusion interval. During an infusion, the corresponding response must account for the continuing input; the post-infusion expression should not be applied before the infusion stops. In practice, software can predict concentration for each recorded infusion interval by integrating the compartment equations. The infusion record is part of the model input, not merely a note attached to the concentration data.

Distinguishing Distribution from Terminal Elimination

After a two-compartment dose or infusion, the early concentration decline can be steep because drug is moving from the central compartment into the peripheral compartment as well as being eliminated. This early, faster component is called the distribution phase. At later times, the faster component has diminished and the slower component dominates; the slope of that terminal log-concentration segment is associated with \(\beta\). The terminal phase is not a direct measurement of elimination from the body alone: its rate reflects the combined compartment system. In particular, \(\beta\) is not generally equal to \(CL/V_c\).

For post-infusion time \(\tau=t-T\), a fitted biexponential form is \(C(\tau)=A e^{-\alpha\tau}+B e^{-\beta\tau}\), where \(A\) and \(B\) are post-infusion coefficients. Plotting concentration against time on a logarithmic vertical scale can help reveal curvature: a visibly steeper early segment followed by a shallower late segment is consistent with two phases. A straight terminal segment alone does not establish that the full profile is one-compartment.

Clearance, Volume, and Exposure Across Cohorts

For a linear intravenous model with complete exposure measurement to the end of the disposition profile, the total area under the concentration-time curve satisfies \(AUC_{0-\infty}=D/CL\), where \(D\) is the delivered dose. Thus \(CL=D/AUC_{0-\infty}\). The relationship follows because total systemic elimination over time equals the delivered dose, and elimination rate is \(CL\cdot C(t)\); integrating gives \(D=CL\int_0^\infty C(t)\,dt\). This requires intravenous systemic delivery, linear clearance, and an AUC that adequately captures the full profile.

Volume has more than one meaning in a multicompartment model. The central volume \(V_c\) relates the central amount to central concentration. The steady-state volume \(V_{ss}\) describes the total distribution extent at equilibrium; in the standard two-compartment model, \(V_{ss}=V_c+V_p\). A terminal volume \(V_z=CL/\beta\) is derived from terminal slope and clearance. These quantities need not be equal. They should be reported with their definitions rather than generically called “the volume of distribution.”

Fitting and Checking the Infusion Model

A defensible fit starts with verified input records. Confirm the delivered amount, infusion start and stop times, any interruptions or rate changes, sample times, concentration units, and whether samples were collected during or after infusion. Fit predictions should use the actual input history. A timing error can resemble a disposition error: for instance, treating a post-infusion sample as if it were collected during infusion can distort both early concentrations and estimated distribution parameters.

For a simple individual fit, estimate positive disposition parameters using the concentration observations and the recorded inputs, then inspect residuals against time and predicted concentration. Residuals should not show a systematic time pattern if the model captures the profile adequately. In a two-compartment fit, sparse early samples may leave the fast phase poorly identified, while short follow-up may leave the terminal slope and extrapolated AUC uncertain. Parameter uncertainty and correlation matter: a numerically converged fit does not guarantee well-determined parameters.

Alternatives, Failure Modes, and Limits

A one-compartment fit is more parsimonious and can be appropriate when observed data do not support a distinct distribution phase. A two-compartment fit can represent biphasic decline but requires enough informative observations to estimate both rates and their coefficients. Adding a compartment solely because it lowers a fit statistic can produce unstable or non-identifiable parameters. Noncompartmental analysis provides an alternative summary of exposure and terminal slope without specifying a compartment structure, but it still depends on adequate sampling, correct dose records, and suitable terminal-phase selection. These approaches answer related but not identical questions: compartment models describe a specified disposition structure, while noncompartmental summaries characterize observed exposure and terminal behavior.

Common warning signs include implausible parameter estimates, strong residual trends, an estimated fast phase unsupported by early samples, and large changes in clearance or terminal slope when one late observation is removed. An AUC extrapolation based on a poorly estimated terminal slope is especially fragile. If cohort dose-normalized AUC differs, investigate uncertainty, sampling, assay and input records, and possible dose dependence before attributing the difference to a biological mechanism.

Common mistake: “The terminal slope is \(\beta\), so elimination clearance must equal \(\beta V_c\).” This is generally wrong for a two-compartment model because the terminal slope reflects the coupled central and peripheral system, and \(V_c\) is not the terminal volume. Estimate clearance using the appropriately defined intravenous dose-to-AUC relationship, and distinguish \(V_c\), \(V_{ss}\), and \(V_z=CL/\beta\).