1. What Is PK Sampling Design?
PK sampling design is the process of deciding when biological samples—typically plasma, serum, blood, or another matrix—should be collected after drug administration.
The objective is not simply to collect as many samples as possible. The sampling schedule should provide sufficient information about the concentration-time profile to answer the scientific question and estimate the PK quantities or model parameters of interest.
For example, a study intended to estimate exposure may emphasize broad temporal coverage, whereas a study intended to characterize absorption rate or early distribution may require samples concentrated in specific portions of the profile.
2. Why Do Sampling Times Matter?
PK data are fundamentally longitudinal. A concentration measured at one time point provides limited information by itself; the pattern of concentrations across time reveals the dynamics of drug disposition.
Different portions of a concentration-time profile provide information about different processes. Early observations may be important for absorption and distribution, while later observations are often important for characterizing elimination and terminal behavior.
A useful PK schedule distributes observations across the portions of the concentration-time profile that contain information about the processes and parameters of interest.
A schedule that samples only near the peak may characterize Cmax and approximate Tmax, but it may provide little information about the terminal elimination phase. Conversely, a schedule dominated by late samples may characterize terminal decline while missing the absorption phase.
3. Start With the Scientific Question
The first step in PK sampling design is to determine what the study needs to learn. Sampling times should follow from the scientific objectives rather than being selected independently of them.
| Scientific objective | Sampling information that may be important |
|---|---|
| Characterize absorption | Early samples spanning the expected rise in concentration and the vicinity of the expected peak |
| Estimate Cmax and Tmax | Sufficient observations around the expected peak rather than a single assumed peak time |
| Characterize distribution | Early post-dose observations that capture the initial decline or distribution phase |
| Estimate AUC | Broad coverage across the concentration-time profile, including sufficiently late observations to characterize the relevant exposure interval |
| Estimate terminal half-life | Multiple sufficiently separated observations during the terminal elimination phase |
| Develop a compartmental model | Sampling distributed across phases that distinguish competing structural models and identify their parameters |
| Population PK modeling | Samples that provide information on relevant PK parameters while balancing participant burden and the intended analysis design |
4. Think in Terms of PK Phases
A concentration-time profile can often be divided conceptually into regions associated with different kinetic processes. The exact interpretation depends on the route of administration and the underlying PK model.
Absorption phase
Following an extravascular dose, concentration generally rises as drug enters the systemic circulation. Sampling during this interval can provide information about the rate and extent of absorption.
Peak region
The neighborhood around the maximum observed concentration is important for estimating Cmax and Tmax. If the sampling grid is too coarse around the expected peak, the observed maximum may substantially depend on which scheduled time happened to fall closest to the true peak.
Distribution phase
After an IV bolus or other rapidly available administration, an early decline may reflect distribution as well as elimination. Early samples can therefore be particularly informative for distinguishing simple and multi-compartment behavior.
Terminal phase
Later observations can provide information about the terminal decline. These samples are especially important when estimating terminal half-life or extrapolating exposure beyond the final observed concentration.
5. Why Early Sampling Can Be Critical
Early sampling is often one of the most difficult parts of PK study design. Concentrations may change rapidly during the first few hours—or even the first few minutes—after administration.
If the first sample is collected too late, the study may miss an important portion of the concentration-time trajectory.
This can affect estimation of:
- the early distribution phase,
- the absorption rate,
- the observed maximum concentration,
- the timing of the maximum concentration, and
- structural parameters associated with rapid kinetic processes.
6. Sampling Around Cmax
The peak region deserves special attention in studies where Cmax or Tmax is an important endpoint.
Suppose the expected peak occurs at approximately 2 hours. Collecting only a sample at exactly 2 hours assumes that the expected peak is known accurately and that between-subject variation will not shift the actual peak.
A more informative schedule may place observations before and after the anticipated peak.
| Sampling pattern | Potential information |
|---|---|
| One sample near expected peak | Provides one observation near the anticipated maximum but gives limited information about the local shape of the curve. |
| Samples before and after expected peak | Provides information about whether concentration is still rising or has begun to decline. |
| Several closely spaced samples around expected peak | Can better characterize the local shape when peak timing is scientifically important. |
The appropriate spacing depends on the expected absorption or disposition timescale, route of administration, dose, formulation, and study objective.
7. Why Late Sampling Matters
Late samples provide information about the decline in concentration after the more rapid portions of the profile have occurred.
For a simple first-order terminal process:
Taking logarithms gives:
Thus, the terminal rate constant \(\lambda_z\) is estimated from the slope of the terminal log-concentration relationship, and the corresponding half-life is:
If late sampling is insufficient, the terminal slope may be poorly estimated. That can affect terminal half-life estimates and, in noncompartmental analysis, the extrapolated portion of AUC.
8. Sampling Design for AUC
AUC represents exposure over time. For observed concentrations, the trapezoidal rule provides a common approximation:
The quality of the numerical approximation depends partly on how well the sampling schedule represents changes in concentration between observations.
If concentrations change rapidly between two widely separated samples, a straight-line approximation may poorly represent the actual curve.
9. Sampling for Structural Model Identification
When the objective is compartmental PK modeling, sampling design becomes more than an exercise in estimating summary endpoints. The schedule must contain information capable of distinguishing different kinetic structures.
For example, consider two candidate models:
- a one-compartment model with a single disposition rate, and
- a two-compartment model with rapid distribution followed by slower elimination.
If observations are collected only after the rapid distribution phase has largely disappeared, both models may produce very similar predictions over the observed time range.
This is why sampling design and model development are closely connected. Sampling times should be chosen with the intended structural model in mind.
10. Which Sampling Times Inform Which Parameters?
Different sampling regions can contribute different amounts of information about model parameters.
| Parameter or quantity | Potentially informative sampling region | Why |
|---|---|---|
| ka | Early rising portion | Absorption rate influences how quickly concentrations increase after an extravascular dose. |
| Cmax | Peak region | The observed maximum depends on observations surrounding the actual peak. |
| Tmax | Peak region | Sampling resolution influences the observed timing of the maximum. |
| V | Early concentration data and model-specific regions | Volume parameters affect the concentration scale and distribution behavior. |
| CL | Broad profile / exposure information | Clearance is related to systemic exposure under linear conditions. |
| λz | Terminal phase | The terminal slope is determined primarily by appropriately selected late observations. |
| Terminal half-life | Terminal phase | Half-life is calculated from the terminal rate constant. |
These relationships are conceptual rather than absolute. In a fitted PK model, information about a parameter can come from multiple regions of the profile, and parameters can be correlated with one another.
11. Should Sampling Times Be Evenly Spaced?
Not necessarily.
Equal spacing is convenient operationally, but PK processes are rarely equally informative across time. A concentration-time profile may change very rapidly early and much more slowly later.
Consequently, a useful schedule may have:
- relatively dense early sampling,
- several observations around the expected peak,
- appropriately spaced observations during the distribution phase, and
- later observations extending sufficiently into the terminal phase.
Uniform spacing is operationally simple, but phase-focused sampling can place more observations where the concentration changes rapidly or where the scientific question requires greater resolution.
The optimal schedule is therefore usually a compromise among information, participant burden, assay feasibility, operational constraints, and the intended analysis.
12. What About Sparse PK Sampling?
Not every PK study requires intensive sampling in every participant. Sparse sampling can be particularly useful when intensive sampling would be impractical or when the primary analysis uses a population PK framework.
In a population PK study, different participants may contribute samples at different times. Collectively, those observations can cover a broader portion of the concentration-time profile than would be practical for each individual.
The design must nevertheless provide adequate information for estimating the parameters and variability of interest.
13. Sampling and Population PK
Population PK analysis estimates typical PK behavior while modeling variability between individuals and, where appropriate, relationships between PK parameters and covariates.
Sampling design can influence how well these components are separated. Concentration measurements collected at informative times contribute to the estimation of the structural model, while observations across participants help characterize between-subject variability.
A population PK design therefore needs to consider at least three dimensions:
- Within-profile information: are the available times informative about the PK trajectory?
- Between-subject coverage: do participants contribute information across the relevant time and covariate ranges?
- Operational feasibility: can the intended samples actually be collected and quantified reliably?
14. The Sampling Design Trade-Off
Real studies rarely permit unlimited sampling. Each additional sample can increase participant burden, staff workload, processing requirements, assay cost, and the likelihood of missed or mistimed collections.
The design problem can therefore be viewed as balancing information against burden.
| Consideration | Design implication |
|---|---|
| Expected rapid absorption | Consider more frequent early observations. |
| Expected short half-life | Ensure the schedule captures the relevant decline before concentrations approach the assay's quantification limits. |
| Expected long half-life | Consider whether follow-up extends sufficiently to characterize terminal behavior. |
| High participant burden | Prioritize times that are most informative for the primary objective. |
| Population variability | Consider whether fixed times are likely to capture important differences in peak timing or elimination. |
| Assay limitations | Ensure concentrations at planned late times remain measurable with adequate reliability. |
| Complex structural model | Design sampling to distinguish the kinetic features required by the model. |
15. Why Simulate a Sampling Design?
Simulation can be used before a study begins to investigate whether a proposed sampling schedule is likely to provide useful information.
For example, suppose a candidate model predicts concentrations according to:
A designer can simulate concentration-time profiles using plausible values of \(C_0\), \(k\), and residual variability, then compare alternative sampling schedules.
For more complicated models, simulation can investigate whether the planned observations distinguish:
- different absorption rates,
- one- versus two-compartment disposition,
- different clearance values,
- different terminal half-lives, or
- important covariate effects.
16. Worked Example: Choosing Informative Sampling Times
Suppose a hypothetical oral drug is expected to have:
- an absorption phase during the first several hours,
- a peak concentration near 2 hours, and
- a terminal half-life of approximately 8 hours.
Consider the following candidate schedule:
| Time after dose | Purpose |
|---|---|
| 0.25 h | Early absorption |
| 0.5 h | Early absorption |
| 1 h | Rising concentration |
| 2 h | Expected peak region |
| 3 h | Post-peak behavior |
| 6 h | Disposition |
| 12 h | Elimination |
| 24 h | Later elimination |
| 36 h | Terminal phase |
| 48 h | Terminal phase |
Step 1: Examine the early phase
The 0.25-, 0.5-, and 1-hour samples provide several observations while concentration is expected to be changing rapidly. These samples can help characterize the rising portion of the profile.
Step 2: Examine the peak
The 2-hour sample targets the expected peak, while the 1- and 3-hour samples provide observations on either side of it. This is more informative than relying on a single assumed peak time.
Step 3: Examine the later phase
The 12-, 24-, 36-, and 48-hour observations extend the profile into the expected terminal phase.
Step 4: Consider the half-life
With an approximate half-life of 8 hours, 48 hours corresponds to:
The expected concentration fraction after six half-lives under a simple first-order model is:
This illustrates why late samples can extend the observable profile far beyond the peak and provide information about terminal decline.
17. Common PK Sampling Design Problems
1. Sampling starts too late
The study may miss rapid absorption or early distribution behavior.
2. Too few samples around the peak
Cmax and Tmax can become highly dependent on the particular scheduled times.
3. No adequate terminal samples
Terminal slope and half-life may be poorly characterized, and AUC extrapolation may become more substantial or uncertain.
4. Samples are concentrated in one region
Many observations clustered around the same phase may provide less information than a smaller number of observations distributed across informative regions.
5. Sampling times assume the PK is already known
A design based too heavily on a single expected parameter set may perform poorly when actual PK varies substantially among individuals.
6. Operationally unrealistic schedules
A theoretically attractive design is not useful if samples cannot be collected within the required windows or processed reliably.
18. Planned Time vs. Actual Time
In a clinical study, the intended sampling time and the actual sampling time are not always identical.
For example, a sample planned for 2.0 hours might actually be collected at 1.92 hours or 2.14 hours.
For rapidly changing concentrations, this difference can matter more than it would later in the profile.
Therefore, PK datasets should generally preserve accurate actual collection times whenever possible. The analysis can then use the observed time rather than assuming that every sample occurred exactly at the nominal schedule.
19. Sampling Relative to Dose Time
PK sampling times are interpreted relative to a meaningful reference point, usually the time of drug administration.
If the dose time is recorded incorrectly, all subsequent concentration-time observations can effectively be shifted along the time axis.
For a sample collected at clock time \(t_{\mathrm{sample}}\) after a dose given at \(t_{\mathrm{dose}}\), the elapsed time is:
Accurate dose timing is therefore essential when the scientific interpretation depends on short intervals between dose and sample.
20. Can Sampling Designs Be Adaptive?
Some clinical and pharmacometric settings permit flexible sampling windows or adaptive approaches, although the feasibility depends on the study design, analysis plan, and operational environment.
An adaptive strategy might respond to information accumulated during a study. For example, early data could alter expectations about the location of a peak or the duration of the terminal phase.
Any such approach should be prespecified appropriately and evaluated with respect to its statistical and operational consequences.
21. Sampling Design Depends on the Planned Analysis
Sampling requirements differ depending on whether the primary analysis is noncompartmental, compartmental, population PK, or another modeling approach.
| Analysis approach | Sampling emphasis |
|---|---|
| Noncompartmental analysis | Adequate coverage of the concentration-time profile for Cmax, Tmax, AUC, and terminal characterization when required. |
| Compartmental PK | Sampling that captures the kinetic features needed to estimate and distinguish the structural model parameters. |
| Population PK | Informative sampling across participants and time, often with sparse sampling combined across individuals. |
| PK/PD modeling | Sampling that supports both the PK trajectory and the exposure-effect relationship of interest. |
A schedule that is adequate for one analysis may be insufficient for another. The sampling plan should therefore be developed alongside the analysis plan.
22. A Practical PK Sampling Design Workflow
- Define the scientific objective. Determine which PK quantities, model parameters, or exposure measures need to be estimated.
- Specify the intended analysis. Decide whether the study will rely primarily on NCA, compartmental PK, population PK, or another modeling framework.
- Establish plausible PK behavior. Use prior knowledge, previous studies, literature, or pilot data to define reasonable ranges for absorption, distribution, clearance, and half-life.
- Map the expected concentration-time profile. Identify the regions where concentration is expected to change rapidly and where important parameters are informed.
- Place early samples deliberately. Ensure that important early processes are not missed.
- Cover the peak region. Use sufficient resolution around the anticipated maximum when Cmax or Tmax matters.
- Extend into the terminal phase. Collect enough late observations to characterize terminal behavior when required.
- Evaluate alternative schedules. Simulation can be used to compare candidate sampling designs before study implementation.
- Check operational feasibility. Confirm that the schedule can be implemented accurately and consistently.
- Preserve actual times. Record actual dose and sample times so the analysis reflects the observations that were truly obtained.
23. Key Takeaways
- PK sampling design is the deliberate selection of biological sampling times to obtain useful information about the concentration-time profile.
- The scientific question should determine the sampling schedule; there is no single universally optimal set of PK sampling times.
- Early samples can be critical for characterizing absorption and rapid distribution processes.
- Samples around the expected peak help characterize Cmax and Tmax and reduce dependence on a single nominal peak time.
- Late samples provide information about terminal elimination and can be important for estimating terminal half-life and extrapolated exposure.
- A large number of poorly positioned samples may provide less information than a smaller number of strategically positioned samples.
- Compartmental and population PK analyses require sampling designs that contain information about the parameters and structural features of interest.
- Sparse sampling can be useful, particularly in population PK, but it should still be designed deliberately.
- Simulation can be used before a study to investigate whether alternative sampling schedules provide adequate information.
- Actual dose and sample times should be recorded accurately because PK interpretation depends on elapsed time.
- The best sampling schedule is the one that provides adequate information for the scientific objective while remaining feasible for participants, investigators, and laboratory operations.
Where to Go Next
A natural progression from sampling design is to study Noncompartmental Analysis: Core Concepts, followed by AUC and AUMC, terminal half-life, partial AUC, and regulatory exposure comparisons.
For model-based pharmacometrics, the next step is to examine how informative sampling times support one- and two-compartment PK models, population PK, parameter identifiability, and simulation-based study design.