What a Single PK Sample Can Inform

Why One Sample Can Be Useful but Incomplete

A concentration measured at one time after a dose is not an exposure summary by itself. Its meaning depends on how much drug was given, the time since that dose, the route and regimen, and the pharmacokinetic model used to interpret it. A sample can nevertheless be informative: if enough of the model is known, it may estimate a particular parameter or help constrain exposure. The important distinction is between informing a quantity under stated assumptions and identifying an individual concentration-time profile.

The earlier material on information density versus subject diversity motivates asking what information each observation contributes. Here the question is narrower: what can one concentration contribute for one participant? The answer is not fixed by the sample count alone. It depends on the sample’s dose-relative time and on which model quantities are known independently.

Definition: A single PK sample is one measured drug concentration, \(C(t)\), at a recorded time \(t\) relative to a specified dose or dosing history. To interpret it, one must also specify the dose, route, relevant regimen, concentration units, and the model assumptions. A quantity is identified by the observation only if the observation and those assumptions determine it uniquely; if different parameter values can produce the same observation, it is not identified.

A Single Observation in a Simple Model

Consider a hypothetical one-compartment model after a single intravenous bolus dose \(D\), with linear elimination and no distribution phase. Let \(V\) be the volume of distribution, \(CL\) the clearance, and \(k=CL/V\) the elimination rate constant. Assume \(D\) is fully delivered into the compartment, \(V\) and \(CL\) are constant, and the concentration is measured without error for the moment. For \(t\geq 0\), the model is

C(t)=\frac{D}{V}e^{-kt}, \qquad k=\frac{CL}{V}.

This model makes the role of dose-relative time explicit. The initial concentration is \(D/V\), while the exponential term describes the decline since the dose. The total area under the concentration-time curve after the dose is \(AUC=D/CL\), under the same linear, one-compartment assumptions. A single sample supplies one concentration equation. If both \(V\) and \(k\) are unknown, that equation generally cannot determine both.

Key takeaway: One concentration may identify one unknown parameter if the other required quantities are known, or may constrain several unknowns to a set of possible values. It does not generally identify a unique individual profile when multiple model parameters are unknown.

Worked Example: One Sample, Several Compatible Profiles

Suppose a hypothetical participant receives a 500 mg intravenous bolus. A concentration of 5.00 mg/L is measured 2.00 hours after the dose. Assume the model above, but suppose neither \(V\) nor \(k\) is known. Show that the sample does not determine a unique profile.

Rearranging the concentration equation for \(V\) gives

V=\frac{D e^{-kt}}{C(t)}.

For one possible elimination rate, \(k=0.100\ \mathrm{h}^{-1}\),

V=\frac{500\ \mathrm{mg}\times e^{-(0.100\ \mathrm{h}^{-1})(2.00\ \mathrm{h})}}{5.00\ \mathrm{mg/L}} =81.87\ \mathrm{L}.

This gives \(CL=kV=(0.100)(81.87)=8.187\ \mathrm{L/h}\). For a different possible rate, \(k=0.300\ \mathrm{h}^{-1}\),

V=\frac{500\ \mathrm{mg}\times e^{-(0.300\ \mathrm{h}^{-1})(2.00\ \mathrm{h})}}{5.00\ \mathrm{mg/L}} =54.88\ \mathrm{L},

and \(CL=(0.300)(54.88)=16.46\ \mathrm{L/h}\). Each parameter set reproduces the observed 5.00 mg/L at 2.00 hours, but the predicted profiles and clearances differ. Their AUC values also differ: \(500/8.187=61.07\ \mathrm{mg\cdot h/L}\) for the first set and \(500/16.46=30.37\ \mathrm{mg\cdot h/L}\) for the second, using unrounded values in the calculations. Thus this sample alone does not establish a unique clearance, AUC, or profile.

When a Sample Can Estimate a Specific Quantity

The same observation can become informative about particular parameters when other information is available. This information might be a parameter fixed by a well-supported model or an independently established value. The resulting estimate is conditional on that information: it is not evidence that the assumed value is correct.

Worked Example: Estimating Volume, Clearance, and AUC When the Rate Is Known

Suppose a hypothetical 500 mg intravenous bolus is followed by a measured concentration of 4.00 mg/L at 8.00 hours. Assume a one-compartment model with \(k=0.200\ \mathrm{h}^{-1}\) known independently, linear elimination, and complete dose delivery. Estimate \(V\), \(CL\), and the single-dose \(AUC\).

Solving \(C(t)=(D/V)e^{-kt}\) for \(V\) gives \(V=De^{-kt}/C(t)\). Substituting the values,

V=\frac{500\ \mathrm{mg}\times e^{-(0.200\ \mathrm{h}^{-1})(8.00\ \mathrm{h})}}{4.00\ \mathrm{mg/L}} =\frac{500}{4.00e^{1.6}}\ \mathrm{L} \approx 25.237\ \mathrm{L}.

Since \(CL=kV\), using the unrounded volume gives \(CL=(0.200)(25.237\ldots)\approx 5.0474\ \mathrm{L/h}\). Then

AUC=\frac{D}{CL} =\frac{500\ \mathrm{mg}}{5.0474\ldots\ \mathrm{L/h}} \approx 99.06\ \mathrm{mg\cdot h/L}.

The sample and known \(k\) identify these quantities within the assumed model. The conclusion is conditional: if the participant’s actual elimination rate differs from \(0.200\ \mathrm{h}^{-1}\), the derived volume, clearance, and AUC will generally differ too. This calculation does not show that the one sample independently verified the rate or the model.

Worked Example: Estimating Elimination Rate When Volume Is Known

In a separate hypothetical case, a participant receives a 600 mg intravenous bolus. The model volume is known to be 30.0 L, and the measured concentration is 2.00 mg/L at 6.00 hours. Under the same one-compartment, linear-elimination assumptions, estimate \(k\), \(CL\), and single-dose \(AUC\).

The initial concentration implied by the dose and volume is \(D/V=600/30.0=20.0\ \mathrm{mg/L}\). Taking logarithms of \(C(t)=(D/V)e^{-kt}\) and solving for \(k\),

k=\frac{\ln\{(D/V)/C(t)\}}{t} =\frac{\ln(20.0/2.00)}{6.00\ \mathrm{h}} =\frac{\ln(10)}{6.00\ \mathrm{h}} \approx 0.3838\ \mathrm{h}^{-1}.

Clearance is \(CL=kV\). Using the unrounded value of \(k\),

CL=\frac{30.0\ \mathrm{L}\times\ln(10)}{6.00\ \mathrm{h}} \approx 11.51\ \mathrm{L/h}.

Therefore, \(AUC=D/CL=600/11.5129\ldots\approx 52.12\ \mathrm{mg\cdot h/L}\). The estimate is valid only if the known volume and the assumed model are appropriate. If the volume is not known, the sample does not separately determine \(k\) and \(V\).

Why One Sample Does Not Reveal the Whole Profile

The limitation can be shown directly. For any selected positive value of \(k\), the volume

V(k)=\frac{D e^{-kt}}{C(t)}

makes the model match the observed concentration. Thus one exact observation is compatible with a continuum of \(k,V\) pairs. Because \(CL=kV\), these pairs generally imply different clearances and different future concentrations. The observation anchors the model at one time only after its parameters have been specified; it does not reveal the curve’s slope or its value at other times by itself.

The argument is an identifiability result for this model, not a claim that the observation contains no information. If independent data, prior population information, or a fixed parameter narrows the possible values, the concentration can help distinguish among those possibilities. An individual estimate from a population model is therefore conditional on both the observed concentration and the model assumptions or population information used. It should not be described as a profile measured directly by one sample.

The same caution applies outside the simple intravenous example. With an extravascular dose, a concentration can depend on absorption, bioavailability, distribution, and elimination as well as on sampling time. If several of these are unknown, one concentration generally cannot separate their contributions. The detailed choice of times for characterizing absorption or other phases requires additional design reasoning; the general point here is that the sample must be interpreted relative to its dose history and the model.

Assumptions, Diagnostics, and Alternatives

Before using a single concentration to estimate a parameter or exposure, check the assumptions that make the calculation meaningful. The dose and route must be known, the sample time must be measured relative to the relevant dose, and the concentration and dose units must be compatible. The proposed structural model must describe the relevant concentration-time behavior. For the displayed equations, that means a one-compartment intravenous bolus model with linear, first-order elimination and constant parameters. A sample affected by an unrecorded dose, an incorrect collection time, a sample-handling problem, or a model misspecification cannot be repaired merely by algebra.

A single observation offers limited diagnostic leverage: there is no second concentration from the participant with which to assess the shape of the predicted decline. Potential warning signs include a concentration incompatible with plausible parameter ranges, a dose or time record that is uncertain, or a sample taken in a phase not represented by the assumed model. These signs call for checking records and model assumptions, not for treating a conditional estimate as certain. Measurement error also matters. With assay error, the measured concentration is a noisy observation of the model-predicted concentration, so estimates based on it carry uncertainty even when the model structure is fixed.

The main alternatives provide different kinds of information. Their strengths are not interchangeable:

Approach	What it can contribute	Important limitation
One sample with a fixed parameter or strong external information	A conditional estimate of a parameter or exposure quantity	The result depends strongly on the fixed value and model assumptions
Multiple samples from one participant	More direct information about that participant’s concentration-time pattern	May increase participant burden and still may not identify every parameter
Sparse samples across many participants analyzed with a population model	Population-level information and model-based individual estimates	Individual estimates depend on population assumptions and the information available per participant

A rich profile is not automatically ideal for every study, and a single sample is not automatically uninformative. The choice depends on the study objective. If the goal is a conditional estimate of a specified quantity and the needed supporting information is credible, one sample may be useful. If the goal is to learn an individual profile or separate several unknown parameters, one point is usually insufficient.

Common Pitfall

Common mistake: “The concentration at 8 hours is 4 mg/L, so the participant’s clearance and AUC are determined.” This reasoning treats one observed point as if it specified the whole curve. Correction: a single concentration determines clearance and AUC only under additional information and model assumptions, such as a known elimination rate in the worked example. State those conditions explicitly, and distinguish the conditional estimate from what the sample alone identifies.

Check Your Understanding

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

For a one-compartment intravenous bolus model, what does one concentration at a known time identify if both \(V\) and \(k\) are unknown?
A 400 mg bolus produces a concentration of 2.00 mg/L at 5.00 hours. If \(k=0.100\ \mathrm{h}^{-1}\) is known, calculate \(V\).
In the same model, if \(D\), \(V\), \(C(t)\), and \(t\) are known, write the equation for estimating \(k\).
Why can a concentration from an oral dose be harder to interpret from one sample than the simple intravenous example?

Solutions. (1) It gives one equation involving \(V\) and \(k\), so generally it does not uniquely identify either parameter or the full profile; multiple parameter pairs can fit the point. (2) \(V=De^{-kt}/C(t)=400e^{-(0.100)(5.00)}/2.00=200e^{-0.5}\approx 121.3\ \mathrm{L}\). (3) \(k=\ln\{(D/V)/C(t)\}/t\), provided the assumed model applies and the logarithm’s ratio is positive. (4) After an oral dose, the observed concentration may depend on absorption and bioavailability as well as distribution and elimination. One value generally cannot separate these contributions without additional information or assumptions.