Wagner–Nelson and Loo–Riegelman Methods

Why Compartment Structure Matters

An oral concentration–time profile reflects both drug entering the systemic circulation and drug being distributed and eliminated after it enters. A method that attributes all concentration change to absorption can therefore misstate how much drug has entered the body. Wagner–Nelson and Loo–Riegelman methods address this by combining oral concentrations with a specified systemic disposition model: Wagner–Nelson uses a one-compartment model, while Loo–Riegelman accounts for distribution into and out of a peripheral compartment.

Both methods estimate the cumulative fraction of the eventual systemically available amount that has entered the body by time \(t\). They do not, by themselves, determine absolute bioavailability: an estimate of fraction absorbed is not the fraction of the administered dose that reaches systemic circulation. The methods use the concentration history and disposition parameters to reconstruct input; they do not establish a relationship between in vitro dissolution and in vivo absorption.

Definition: The fraction absorbed by time \(t\), denoted here by \(F_a(t)\), is the cumulative amount of drug that has entered the systemic circulation by \(t\), divided by the total amount that will eventually enter the systemic circulation. It ranges from 0 to 1 under the model assumptions and complete observation of absorption.

Wagner–Nelson: One-Compartment Disposition

Assume linear, time-invariant, one-compartment disposition after systemic entry. Let \(C(t)\) be the oral concentration, \(k_e\) the first-order elimination rate constant, and \(V\) the apparent volume of distribution. Drug enters the central compartment as systemic input and leaves it by first-order elimination. The amount in the compartment is \(V C(t)\), and the cumulative amount eliminated by time \(t\) is \(k_e V\operatorname{AUC}_{0-t}\), where \(\operatorname{AUC}_{0-t}=\int_0^t C(s)\,ds\).

Mass balance says cumulative systemic input by time \(t\) equals the amount still in the compartment plus the amount already eliminated. At infinite time, the amount remaining is zero, so total systemic input is \(k_e V\operatorname{AUC}_{0-\infty}\). Dividing the time-\(t\) input by total input cancels \(V\), giving the Wagner–Nelson estimate:

F_a(t)=\frac{C(t)+k_e\operatorname{AUC}_{0-t}} {k_e\operatorname{AUC}_{0-\infty}}

This expression requires an estimate of the total AUC, including the terminal area after the last measured concentration. It also requires a credible \(k_e\) for the one-compartment disposition model. A terminal slope from an oral profile is not automatically a valid estimate of \(k_e\): slow absorption can make the terminal phase absorption-limited rather than elimination-limited.

Worked Example: Wagner–Nelson With a Terminal AUC Estimate

Suppose hypothetical oral data are measured in mg/L at the times below. Assume a one-compartment elimination rate constant \(k_e=0.600\ \mathrm{h}^{-1}\), and use linear trapezoids between observations.

Time (h)	0	1	2	4	6
Concentration (mg/L)	0	4	6	4	2

The interval areas are \(2\), \(5\), \(10\), and \(6\ \mathrm{mg\,h/L}\), respectively. Thus \(\operatorname{AUC}_{0-2}=2+5=7\ \mathrm{mg\,h/L}\), and the observed area through 6 h is \(23\ \mathrm{mg\,h/L}\). Assuming the terminal decline follows \(k_e\), the extrapolated area after 6 h is \(C(6)/k_e=2/0.600=3.333\ \mathrm{mg\,h/L}\). Therefore:

\operatorname{AUC}_{0-\infty}=23+3.333=26.333\ \mathrm{mg\,h/L}

At 2 h, substitute \(C(2)=6\ \mathrm{mg/L}\) and \(\operatorname{AUC}_{0-2}=7\ \mathrm{mg\,h/L}\):

F_a(2)=\frac{6+(0.600)(7)}{(0.600)(26.333)} =\frac{10.2}{15.8}=0.6456

The estimated cumulative fraction absorbed by 2 h is 0.646, or 64.6% of the eventual systemically available amount. As a check, at 4 h the estimate is \([4+(0.600)(17)]/15.8=0.8987\). At 6 h it is \([2+(0.600)(23)]/15.8=1.000\), because the assumed terminal extrapolation makes the remaining concentration vanish under the same one-compartment model.

Loo–Riegelman: Two-Compartment Disposition

If systemic disposition includes distribution to a peripheral compartment, central concentration alone does not describe all drug remaining in the body. Loo–Riegelman accounts for that peripheral amount. Consider a linear, time-invariant, two-compartment model with absorption entering the central compartment. Let \(V_1\) be the central volume; \(k_{10}\) the elimination rate constant from the central compartment; \(k_{12}\) the transfer rate constant from central to peripheral; and \(k_{21}\) the transfer rate constant from peripheral to central. The disposition parameters are determined independently, typically from an intravenous disposition reference, and are assumed to apply to the oral data.

If \(X_1(t)=V_1C(t)\) is the central amount and \(X_2(t)\) is the peripheral amount, the peripheral-compartment balance is \(dX_2/dt=k_{12}X_1-k_{21}X_2\). With no drug in either compartment before dosing, solving this linear equation gives \(X_2(t)=k_{12}V_1\int_0^t C(s)e^{-k_{21}(t-s)}\,ds\). Cumulative input is the amount in both compartments plus the amount eliminated from the central compartment. Since elimination by \(t\) is \(k_{10}V_1\operatorname{AUC}_{0-t}\), and total input is \(k_{10}V_1\operatorname{AUC}_{0-\infty}\), the Loo–Riegelman expression is:

F_a(t)= \frac{ C(t)+k_{10}\operatorname{AUC}_{0-t} +k_{12}\int_0^t C(s)e^{-k_{21}(t-s)}\,ds }{ k_{10}\operatorname{AUC}_{0-\infty} }

The integral is the peripheral amount expressed in central-concentration units: multiplying it by \(V_1\) gives \(X_2(t)\). In practice, the integral is evaluated numerically from the observed concentration history, often by a quadrature rule using the measured time points. The calculation depends on the history, not just the concentration at \(t\). It also depends on appropriate two-compartment parameters and a sufficiently reliable estimate of total AUC.

Worked Example: Loo–Riegelman With a Numerical Peripheral Correction

Consider hypothetical concentrations \(0\), \(2\), and \(4\ \mathrm{mg/L}\) at times \(0\), \(1\), and \(2\ \mathrm{h}\). Let \(V_1=1.00\ \mathrm{L}\), \(k_{10}=0.500\ \mathrm{h}^{-1}\), \(k_{12}=0.300\ \mathrm{h}^{-1}\), and \(k_{21}=0.200\ \mathrm{h}^{-1}\). Assume \(\operatorname{AUC}_{0-2}=4.00\ \mathrm{mg\,h/L}\), and for this illustration let the estimated total AUC be \(20.0\ \mathrm{mg\,h/L}\).

At 2 h, approximate the weighted integral using trapezoids at 0, 1, and 2 h. The weighted concentrations \(C(s)e^{-0.2(2-s)}\) are \(0\), \(2e^{-0.2}=1.6375\), and \(4\ \mathrm{mg/L}\). Hence:

\int_0^2 C(s)e^{-0.2(2-s)}\,ds \approx \frac{0+1.6375}{2}+\frac{1.6375+4}{2} =3.6375\ \mathrm{mg\,h/L}

The peripheral correction in concentration units is \((0.300)(3.6375)=1.0913\ \mathrm{mg/L}\). The central concentration and elimination terms are \(C(2)=4.00\ \mathrm{mg/L}\) and \((0.500)(4.00)=2.00\ \mathrm{mg/L}\). The denominator is \((0.500)(20.0)=10.0\ \mathrm{mg/L}\), so:

F_a(2)=\frac{4.00+2.00+1.0913}{10.0}=0.7091

The estimated fraction absorbed by 2 h is 70.9% of the eventual systemically available amount. The same result in amount units is \(7.091\ \mathrm{mg}\) absorbed by 2 h divided by \(10.0\ \mathrm{mg}\) total systemic input; the factor \(V_1\) cancels in the fraction.

Comparing the Estimates and Checking Their Behavior

The two methods share a mass-balance basis but differ in what they count as drug still in the body. Wagner–Nelson counts central drug only, which is appropriate under its one-compartment assumption. Loo–Riegelman adds drug in the peripheral compartment, reconstructed from concentration history and transfer parameters. Applying Wagner–Nelson when distribution is material can misattribute redistribution to absorption or elimination.

Worked Example: Why the Compartment Assumption Changes the Estimate

Use the hypothetical data and disposition parameters in the Loo–Riegelman example at 2 h. If one instead applies Wagner–Nelson with \(k_e=0.500\ \mathrm{h}^{-1}\) to the same concentration data and total AUC, the calculation is:

F_a(2)=\frac{4.00+(0.500)(4.00)} {(0.500)(20.0)} =\frac{6.00}{10.0}=0.600

This is 60.0%, compared with the Loo–Riegelman estimate of 70.9%. The difference of 10.9 percentage points comes from the Loo–Riegelman peripheral correction in this calculation. It does not show that one estimate is generally preferable: if the true disposition is two-compartment, the Wagner–Nelson assumptions are not met; if the true disposition is one-compartment, the extra two-compartment structure and its parameters are unnecessary. The comparison is a model-sensitivity illustration, not evidence that either model fits these hypothetical data.

Useful diagnostics follow directly from the mass-balance interpretation. With compatible parameters, adequate observations, and nonnegative systemic input, estimated fractions should generally be nondecreasing and should remain between 0 and 1. Decreases, values above 1, or abrupt jumps can indicate measurement noise, a poor terminal AUC estimate, numerical integration error, a wrong disposition model, or a mismatch between the disposition parameters and the oral data. These signs are prompts to investigate, not a license to truncate estimates to 0 or 1.

Other failure modes deserve attention. Sparse sampling around the peak or during distribution can make the Loo–Riegelman convolution inaccurate. An incorrect \(k_{21}\) changes how much prior concentration is attributed to drug remaining in the peripheral compartment. If absorption is so slow that the terminal oral slope is absorption-limited, using that slope as \(k_e\) can distort the Wagner–Nelson estimate and its extrapolated total AUC. Both methods rely on linear, time-invariant disposition and systemic input into the modeled compartment; nonlinear elimination, time-varying clearance, presystemic loss, or a different input site can violate those assumptions.

When a suitable intravenous disposition reference is available, direct deconvolution can estimate the input rate without imposing the specific one- or two-compartment absorption reconstruction used here. Wagner–Nelson and Loo–Riegelman are useful alternatives when their compartment models are adequate and their calculations are transparent. Neither removes uncertainty in the concentration observations, disposition parameters, interpolation, or terminal extrapolation. Report the model, parameter source, AUC method, and any assumptions about the unobserved tail alongside the estimated fraction-absorbed profile.

Common Mistake

Common pitfall: “The concentration is falling, so absorption must have stopped; the fall directly measures the fraction absorbed.” This is incorrect because concentration can fall while absorption continues, as elimination and distribution occur at the same time. The correction is to use a stated disposition model and mass balance to account for drug remaining in the body and drug already eliminated. Also, \(F_a(t)\) is normalized to the eventual systemic input; it is not automatically an estimate of the fraction of the swallowed dose absorbed.

Key takeaway: Wagner–Nelson reconstructs cumulative input under one-compartment disposition; Loo–Riegelman also accounts for peripheral drug under a two-compartment model. Credible estimates require the matching disposition assumptions, adequate concentration sampling, and a defensible total-AUC estimate.

Check Your Understanding

Answer each question, then compare your reasoning with the brief solutions.

What is the additional quantity counted by Loo–Riegelman that is absent from Wagner–Nelson?
For Wagner–Nelson, \(C(t)=3.0\ \mathrm{mg/L}\), \(k_e=0.40\ \mathrm{h}^{-1}\), \(\operatorname{AUC}_{0-t}=5.0\ \mathrm{mg\,h/L}\), and \(\operatorname{AUC}_{0-\infty}=15.0\ \mathrm{mg\,h/L}\). Calculate \(F_a(t)\).
Why can a terminal slope estimated from an oral profile be unsuitable for \(k_e\)?
What does an estimated fraction absorbed above 1 suggest, and what should an analyst do next?

Solutions. (1) Loo–Riegelman includes the amount in the peripheral compartment, estimated from the concentration history and \(k_{12}\) and \(k_{21}\). (2) \(F_a(t)=[3.0+(0.40)(5.0)]/[(0.40)(15.0)]=5.0/6.0=0.833\), or 83.3% of eventual systemic input. (3) Slow absorption can control the terminal decline, making that slope absorption-limited rather than a measure of elimination. (4) It can indicate incompatible assumptions, an underestimated total AUC, incorrect parameters, or numerical or data problems; investigate those sources rather than automatically truncating the estimate.