Reducing burden is a design trade-off
A pharmacokinetic (PK) schedule asks participants to provide samples at specified times, often around dosing and across one or more visits. Each draw, trip, waiting period, and timing restriction can add burden. Reducing that burden can make a study more feasible and may help participants complete the planned procedures. But removing a draw or visit also removes an observation opportunity. The design question is therefore not simply “How can the schedule be shorter?” It is “What can be simplified while keeping the concentration-time information the study needs?”
A useful first step is to describe burden in separate, countable ways. The number of blood draws measures venipuncture burden; the number of visits captures travel and appointment burden; and the time spent at a visit captures time away from other activities. Total blood volume may also matter. These quantities are related but not interchangeable: moving four draws into two visits may leave the draw count unchanged while reducing travel, whereas removing two draws may reduce venipunctures without changing the number of visits.
Windows, visits, and draws are different design levers
A sampling window specifies when a sample may be collected relative to a nominal time. Narrowing that window changes the permitted timing range; it does not, by itself, reduce the number of samples. It may shorten a visit if it allows procedures to be scheduled more compactly, but a narrower window can also make attendance harder, increase rescheduling, or result in more missed samples when travel or clinic delays occur.
Fewer visits can reduce travel and the disruption of repeated clinic attendance. This may be achieved by combining procedures that were already planned at separate visits. It need not change the total number of draws, and it can increase the length of a particular visit if several samples are clustered there. Removing a visit that contains a unique phase-specific sample is a different change: it reduces burden but also removes that planned observation.
Likewise, reducing the number of draws can lower venipuncture burden, but the effect on information depends on which times are removed. The earlier material on sampling windows for absorption parameters, distribution and clearance, terminal-phase samples, and peak and trough samples establishes why sample timing matters: the same number of observations can be more or less informative depending on where they fall in the concentration-time profile. A phase checklist is a practical first screen, not a substitute for comparing parameter precision when that comparison is needed.
A simple information justification
The direction of the information trade-off can be shown under a deliberately simple model. Suppose a concentration measurement at time \(t_j\) has a model-predicted mean \(f(t_j,\boldsymbol{\theta})\), where \(\boldsymbol{\theta}\) is the vector of PK parameters, and independent normally distributed measurement error with known variance \(\sigma^2\). Assume the model and variance are correctly specified, and consider local information near a specified parameter value. Define the sensitivity vector \(\mathbf{g}_j=\partial f(t_j,\boldsymbol{\theta})/\partial\boldsymbol{\theta}\). The expected Fisher information from the planned measurements is then:
Each term is positive semidefinite: for any vector \(\mathbf{v}\), \(\mathbf{v}^\mathsf{T}(\mathbf{g}_j\mathbf{g}_j^\mathsf{T}/\sigma^2)\mathbf{v}=(\mathbf{g}_j^\mathsf{T}\mathbf{v})^2/\sigma^2\geq 0\). Removing a planned sample subtracts one such term, so under these assumptions it cannot increase this model-based information matrix. This does not say that every removed sample is equally costly: a sample with little sensitivity to the parameter of interest may contribute less than one that distinguishes competing parameter effects. Nor does the equation account for real-world gains if a simpler schedule improves attendance or data completeness. If sampling times change, their sensitivity vectors change too, so the comparison must use the proposed times rather than merely the sample count.
Worked Example: Separating Draws, Visits, and Time
In a hypothetical study, a participant is scheduled for two visits. The first has three draws and takes about 3 hours onsite; the second has two draws and takes about 2 hours onsite. A proposed plan combines the procedures into one visit lasting about 4.5 hours, with all five draws retained. Compare the burden measures.
Draws: the original plan has \(3+2=5\) draws, and the proposed plan also has 5. The number of venipunctures has not been reduced.
Visits: the original plan has 2 visits and the proposed plan has 1, a reduction of \(2-1=1\) visit, or \(1/2=50\%\).
Onsite time: the original total is \(3+2=5\) hours; the proposed total is 4.5 hours. The change is \(5-4.5=0.5\) hour less onsite, a \(0.5/5=10\%\) reduction in total onsite time.
This change clearly reduces travel occasions, while retaining the planned draws. It also creates a longer single visit than either original visit, so it may suit participants who prefer fewer trips but be difficult for those who cannot commit to a 4.5-hour visit. The schedule still needs a phase check: combining visits is acceptable only if the new timing preserves the required observations relative to dose.
Use phase coverage as a screen, not a guarantee
Before removing a sample or visit, write down which concentration-time phase each planned observation is intended to represent and which study question depends on it. A schedule may need observations during absorption, distribution, near a peak or trough, or during terminal decline; the relevant set depends on the drug, regimen, and objectives. A phase can be represented in the overall study even when no single participant contributes every phase, for example when sampling is deliberately staggered across participants. But a phase represented only by a small or atypical subset may be fragile, particularly if those observations are missed.
A practical review can proceed in order: list the study questions, identify the phases needed for those questions, map planned samples to phases, and then mark what changes in the reduced plan. Check both the individual participant’s experience and the study-wide coverage. A schedule that is tolerable for one participant but leaves the study with no useful terminal observations has not preserved the required information. Conversely, retaining one sample in each broad phase does not establish that the reduced schedule estimates every parameter adequately; parameter precision can depend on exact times, variability, and the model.
Worked Example: Checking a Reduced Schedule Against Required Phases
Suppose a hypothetical single-dose study identifies four required phase categories for its primary questions: early absorption, distribution, peak region, and terminal decline. The original plan has seven samples, assigned by the design team as follows: one early absorption sample, two distribution samples, one peak-region sample, and three terminal-decline samples. A proposed reduced plan has five samples: one early absorption, one distribution, one peak-region, and two terminal-decline samples.
Count the reduction. The plan falls from 7 to 5 samples, so it removes \(7-5=2\) draws per participant, a reduction of \(2/7\approx0.286\), or about 28.6%.
Check broad phase representation. Both plans retain at least one planned observation in each of the four required categories. The reduced plan therefore passes this initial coverage screen.
State what the screen does not establish. The reduced plan has fewer distribution observations and fewer terminal observations. The phase labels alone cannot show whether this loss materially increases uncertainty in the parameters of interest. The exact times, expected concentration profile, measurement variability, and model matter. A next step is to compare candidate schedules’ parameter precision using the study’s model and assumptions, rather than concluding that the two plans are equally informative because both list four phases.
When narrowing windows helps—and when it backfires
Consider a hypothetical two-sample visit with nominal sample times of 2 and 4 hours after dose. Suppose the original permitted windows are 1.5–2.5 hours and 3–5 hours. The earliest permitted first draw is at 1.5 hours and the latest permitted second draw is at 5 hours, giving a possible span of \(5-1.5=3.5\) hours between those extremes. If the windows are narrowed to 1.75–2.25 and 3.75–4.25 hours, the corresponding span is \(4.25-1.75=2.5\) hours. That may allow a shorter appointment under an operational plan that schedules participants within the narrowed ranges.
The arithmetic describes possible timing bounds, not a guaranteed reduction in actual time at the clinic. A participant could still arrive early, experience delays, or wait for other procedures. More importantly, narrowing the windows may make it harder to attend at the required times. If actual times fall outside the permitted ranges, the data may be less useful for the intended analysis or may require careful review. The design should therefore consider whether the narrowed windows are feasible at the study sites and for the participant population.
A further distinction is important: a narrow allowable window around a planned time is not the same as choosing a narrower range of nominal sampling times across the concentration-time curve. The first restricts when a particular draw can occur; the second can concentrate observations into fewer parts of the curve and leave other phases unrepresented. Both changes may be described informally as “narrowing the window,” but their design consequences differ.
Worked Example: Comparing Window Width and Visit Span
A hypothetical participant has two scheduled draws, nominally at 2 and 4 hours after dosing. Under Plan A, the windows are 1.5–2.5 hours and 3–5 hours. Under Plan B, they are 1.75–2.25 hours and 3.75–4.25 hours. Calculate the timing span from the earliest permitted first draw to the latest permitted second draw for each plan.
Plan A: latest second time minus earliest first time is \(5-1.5=3.5\) hours.
Plan B: the span is \(4.25-1.75=2.5\) hours.
The bounds are 1 hour closer together in Plan B, while both plans still schedule one sample near 2 hours and one near 4 hours. This suggests that Plan B could support a shorter visit if the clinic can coordinate collection accordingly. It does not prove that actual visits will be 1 hour shorter, that the windows are operationally feasible, or that the information is unchanged. Those are separate checks: actual onsite duration should be monitored, feasibility should be assessed, and the importance of the allowed timing deviations should be considered for the study’s PK questions.
Practical diagnostics and common failure modes
A reduced schedule should be reviewed against both its intended benefits and its observable consequences. Useful checks include the planned and actual number of draws per participant, completed visits, missed or out-of-window samples, time onsite, and whether the planned phase coverage is present in the data. If a proposed change aims to reduce travel, count visits; if it aims to reduce venipunctures, count draws. A change can succeed on one measure and fail on another.
- Phase gaps: a reduced plan has no observations in a phase required by the study question, or coverage depends on very few participants.
- Timing feasibility: narrowed windows are associated with frequent late, early, or missed collections, suggesting the constraints may be unrealistic.
- Burden not actually reduced: visits decrease, but the combined appointment becomes impractically long; or fewer samples are planned but rescheduled procedures restore the burden.
- Information assumed from counts: a schedule is judged adequate because it retains a certain number of samples, without checking their timing or the parameters of interest.
- Overly simple phase labels: a sample is counted as “terminal” or “absorption” without confirming that its actual time and concentration profile support that role.
A common mistake is to say, “The reduced plan has at least one sample in every phase, so no important information was lost.” The correction is that phase coverage is a necessary design screen only when those phases are required; it is not sufficient evidence of unchanged precision. Review exact sampling times and, when the study objectives require it, compare candidate designs using model-based precision or another prespecified information criterion. The opposite mistake is to assume that every additional draw is valuable. A draw that adds little information for the target question may be a reasonable candidate for removal, but that judgement depends on the model and objectives rather than the draw count alone.
Choosing among practical alternatives
| Design change | Burden it may reduce | Main risk to check |
|---|---|---|
| Combine procedures into fewer visits | Travel and appointment count | Longer or less convenient individual visits; altered timing relative to dose |
| Remove selected draws | Venipunctures and possibly visit duration | Loss of observations needed to represent a key phase or estimate a target parameter |
| Narrow allowable collection windows | Potentially shorter or more predictable visit span | More difficult attendance, out-of-window samples, or missed collections |
| Stagger sampling across participants | May allow fewer draws per participant while distributing sampling effort | Some phases may be weakly represented or vulnerable to missed data across the group |
These approaches are alternatives, not interchangeable guarantees. Combining visits changes the pattern of attendance; removing draws changes the number and location of observations; narrowing windows changes permitted timing; and staggering changes which participants contribute at which times. Each should be judged against the study’s objectives and the participants’ likely experience. A simple coverage table is useful early in the decision, while formal comparisons of parameter precision are more appropriate when the design decision depends on how much uncertainty the candidate schedules produce.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions.
- A plan changes from 6 draws over 3 visits to 6 draws over 2 visits. Which burden measure definitely decreases, and which does not change?
- A proposed schedule removes 2 of 8 draws per participant. Calculate the percentage reduction in draws.
- Two samples have permitted windows of 2.0–2.5 hours and 4.0–5.0 hours. What is the span from the earliest possible first draw to the latest possible second draw?
- Why does retaining one sample in each required phase not prove that parameter precision is unchanged?
- A narrowed window is followed by more missed samples. Name one reason this could happen and one design measure that should be checked.
Solutions. 1. The visit count decreases from 3 to 2, so travel occasions may decrease; the draw count remains 6. 2. The reduction is \(2/8=0.25\), or 25%. 3. The span is \(5.0-2.0=3.0\) hours. 4. Phase labels do not capture exact sampling times, model sensitivities, or measurement variability, all of which can affect precision. 5. The narrower timing requirement may be harder to meet because of travel, scheduling, or clinic delays; check the rate of missed or out-of-window samples, alongside actual visit duration.