Defining Partial AUC Window Boundaries

Why Window Boundaries Matter

A total AUC summarizes exposure across a broad time range. A partial AUC instead summarizes exposure within a specified part of that range. Its meaning therefore depends on more than the concentration measurements: it depends on which start and end times define the window. A window chosen to represent early exposure is not interchangeable with one chosen to represent later exposure, even if both have the same duration.

The boundary rule must be clear before profiles are summarized or compared. For example, “early exposure” is not an operational definition unless the elapsed-time interval, such as 0 to 2 hours after dosing, is stated. The time origin and the handling of observations at a boundary should also be specified. Without these choices, two analysts can use the same concentration–time data and calculate quantities that answer different questions.

Definition: For a concentration–time profile $C(t)$, a partial AUC over the window from time $a$ to time $b$, where $a<b$, is the area under that profile between those times: $$ \operatorname{AUC}_{a-b}=\int_a^b C(t)\,dt. $$ The window is defined by its time origin, its start $a$, and its end $b$. If concentration is measured in mg/L and time in hours, the partial AUC has units of mg·h/L.

This definition describes the target quantity, not a particular calculation from sampled data. A measured profile usually supplies concentrations at discrete times rather than at every instant. The concentration values at, or around, a boundary and the calculation used to represent the profile between observations require additional decisions. The introductory material on AUC as a numerical integral establishes the area interpretation; here the focus is on defining exactly which time interval that area is meant to cover.

Specify the Time Coordinate and Both Boundaries

A boundary is meaningful only relative to a time coordinate. In a single-dose profile, a natural coordinate is elapsed time since the dose: time zero is the dose time, and a window such as 1 to 4 hours means the interval beginning one hour after dosing and ending four hours after dosing. In other settings, a protocol may define boundaries by clock time, by time since a meal or another event, or relative to a dosing interval. The coordinate must be named rather than left implicit.

Both endpoints must be reported. Naming only a start time, such as “from 2 hours onward,” leaves the end of the window unspecified unless the endpoint is explicitly defined by the study design. Likewise, “through the absorption phase” is not a reproducible boundary rule unless that phase is operationalized. For a between-profile comparison, the same prespecified rule should ordinarily be applied to every profile being compared. Otherwise, differences in the partial AUC may reflect different windows as well as different concentration profiles.

A useful definition includes the time origin, the start and end, the units, and any event or dosing-interval reference. For example: “from 0 to 2 hours after the recorded dose time, in elapsed hours.” If the study uses a window within each dosing interval, state which dose starts that interval and whether the time scale resets at the next dose.

Worked Example: Separating Adjacent Windows

Suppose a hypothetical single-dose profile is to be summarized in two windows: 0 to 2 hours after dosing and 2 to 8 hours after dosing. Measurements are available at elapsed times 0, 1, 2, 4, and 8 hours. Specify what profile portions each window includes.

The time origin is the dose, so $t=0$ is the dosing time. The first window is $[0,2]$ hours and has duration $2-0=2$ hours. It includes the profile from dosing through 2 hours. The second window is $[2,8]$ hours and has duration $8-2=6$ hours. It includes the profile from 2 through 8 hours.

The observation at 2 hours lies exactly at the shared boundary. In the continuous-area definition, counting the single instant $t=2$ in both intervals does not duplicate any area: a single point has zero width and contributes zero to the integral. The interval before 2 hours belongs to the first window; the interval after 2 hours belongs to the second. A discrete implementation still needs a consistent boundary-handling rule for its observations and intervals, but the mathematical windows are unambiguous once their endpoints are specified.

Endpoint Conventions and Their Consequences

Mathematical interval notation may describe a window as $[a,b]$, $(a,b)$, or with one endpoint included and the other excluded. For an integrable concentration profile, these choices produce the same area because they differ only at one or two individual time points. Thus, $ \int_a^b C(t)\,dt $ does not change merely because the endpoint itself is labeled included or excluded.

This mathematical fact does not mean that endpoint conventions can be ignored in practice. A sampled profile is represented by a finite set of observations and a specified calculation procedure. The exact observation at a boundary may be used, omitted, or assigned to one of two adjacent windows according to the protocol and analysis rule. More importantly, the rule for which intervals of the profile contribute must be consistent. A clear convention prevents an observation or segment from being inadvertently omitted or used in a way that changes the intended window.

The distinction is between the exact point and the neighborhood around it. Including or excluding the concentration at exactly 2 hours does not alter a continuous integral. Moving the boundary from 2 to 2.5 hours does alter the interval and can alter the area. The latter is a change in the estimand, not merely a notation choice.

Worked Example: A Shifted Start Time

Consider a hypothetical profile with concentration $C(t)=4$ mg/L from 1.0 to 1.5 hours after dosing. Compare the area included by windows starting at 1.0 and 1.5 hours, with both windows ending at 4 hours. Assume this constant concentration description applies throughout the interval from 1.0 to 1.5 hours.

The window starting at 1.0 hours includes the full 0.5-hour interval in question. Its contribution from that interval is $$ 4\ \text{mg/L}\times(1.5-1.0)\ \text{h}=2\ \text{mg·h/L}. $$

The window starting at 1.5 hours excludes that interval, so it excludes this 2 mg·h/L contribution. In general, shifting a start time later removes the area between the old and new starts; shifting an end time later adds the area between the old and new ends. This calculation concerns only the portion whose boundaries changed, not the entire partial AUC.

A Short Justification for Boundary Sensitivity

The effect of moving a boundary follows directly from the additivity of integrals. Let $A(a,b)$ denote the area from $a$ to $b$, and choose $\delta>0$ such that $a+\delta<b$. Splitting the interval at $a+\delta$ gives

A(a,b)=\int_a^{a+\delta}C(t)\,dt+\int_{a+\delta}^b C(t)\,dt.

Therefore,

A(a,b)-A(a+\delta,b)=\int_a^{a+\delta}C(t)\,dt.

Similarly, if $b+\delta$ is within the profile,

A(a,b+\delta)-A(a,b)=\int_b^{b+\delta}C(t)\,dt.

For a nonnegative concentration profile, both added or removed areas are nonnegative. A later start cannot add exposure from the removed interval, and a later end cannot remove exposure from the added interval. Their numerical importance depends on the concentrations and durations in the affected portions. This result assumes the profile is integrable over the stated range; it does not require a particular shape or a specific sampling schedule.

Choose a Boundary Rule That Matches the Question

Different boundary choices answer different questions. Fixed elapsed-time windows compare exposure during the same post-dose times, such as 0 to 2 hours after each dose. These are suitable when the scientific question concerns exposure within that prespecified period. An event-relative window, such as the period beginning at a specified clinical event, instead follows that event. Because the time origin changes, it may include different elapsed post-dose portions in different profiles.

Worked Example: Fixed-Time and Event-Relative Windows

Suppose two hypothetical profiles have peak times of 2 and 3 hours after dosing. Compare a fixed window of 1 to 4 hours post-dose with a 3-hour window beginning at each profile’s peak time. Determine the time ranges and whether they represent the same part of the post-dose profile.

The fixed window is 1 to 4 hours after dosing for both profiles. Its duration is $4-1=3$ hours in each case. It summarizes the same elapsed post-dose period for both profiles.

For the event-relative rule, the first profile’s window is 2 to 5 hours after dosing, because its peak is at 2 hours and $2+3=5$. The second profile’s window is 3 to 6 hours after dosing, because its peak is at 3 hours and $3+3=6$. Both windows last 3 hours, but they do not cover the same post-dose times. The event-relative rule follows the peak; the fixed rule does not. Neither definition is automatically preferable: the choice depends on the question, and the resulting partial AUCs must be interpreted according to the rule used.

Another possible source of inconsistency is using nominal instead of recorded dose times. If the intended window is defined in elapsed time after the actual dose, a dose administered later than scheduled shifts the corresponding clock-time boundaries. Treating the scheduled time as the origin when the dose was given at a different time changes which observations fall within the elapsed-time window.

Worked Example: Recorded Dose Time Versus Scheduled Time

In a hypothetical study, a dose scheduled for 08:00 is actually given at 08:12. A sample is collected at 10:12, and the intended window is 1 to 4 hours after dosing. Locate the sample relative to the dose and state the window on the clock.

Relative to the recorded dose time, the sample is $10{:}12-08{:}12=2$ hours after dosing. It lies within the intended 1-to-4-hour window. The clock-time boundaries corresponding to that rule are 09:12 and 12:12.

If the scheduled time were incorrectly used as the origin, the same sample would be $10{:}12-08{:}00=2.2$ hours after dosing, and the clock-time window would be 09:00 to 12:00. That is a different mapping of clock times to elapsed times. The analysis should use the origin specified by the protocol and recorded dosing information, not switch between scheduled and actual times without justification.

Diagnostics, Alternatives, and Limitations

Before summarizing a partial AUC, check that each profile has a usable time origin and that both boundaries are explicit, ordered, and expressed in the same units. Confirm that the selected times lie within the intended profile or explain how the analysis handles a window extending beyond observed data. Check whether boundaries refer to elapsed time, clock time, or an event. Finally, verify that the same rule has been applied across profiles when comparability is the objective.

Several warning signs point to a boundary-definition problem: a window is described only as “early” or “late”; different analyses use different endpoints without explanation; nominal and actual dose times are mixed; or a boundary is changed after seeing the concentration profiles. A window that depends on a profile-specific feature may be scientifically defensible, but it should not be presented as if it were the same fixed post-dose interval for every profile.

The main alternatives are not competing numerical formulas for the same quantity; they are different definitions of the time range. A fixed post-dose window emphasizes exposure at common elapsed times. A clock-time window emphasizes a shared calendar interval. An event-relative window follows a specified event. These can yield different partial AUCs, so the choice should be driven by the scientific question and stated before interpreting comparisons.

A boundary definition alone does not determine how concentrations between sampling times are handled, nor does it guarantee that the profile is adequately observed. Those issues affect estimation and are separate from specifying the target window. Also, a window can omit relevant exposure outside its bounds. A partial AUC is therefore a summary of exposure within its defined interval, not automatically a substitute for total exposure or a complete description of the concentration–time profile.

Common Mistakes

Giving a duration but not a time origin. “A 3-hour window” could begin at dose time, a later event, or a clock-time boundary. State the origin as well as the duration.
Assuming equal durations mean equal windows. A 1-to-4-hour window and a 3-to-6-hour window each last 3 hours but include different post-dose periods.
Treating endpoint inclusion as the main source of area differences. Including or excluding one exact point does not change a continuous integral. Moving a boundary changes a whole interval and can change the area.
Using different dose-time origins across profiles. Applying scheduled time to one profile and recorded time to another can produce inconsistent elapsed-time windows.
Choosing a window after inspecting the results without explanation. This can make comparisons dependent on observed profiles rather than on a clearly stated scientific question.

Key takeaway: Define a partial-AUC window by naming its time origin, start, and end. Apply that rule consistently, and distinguish the mathematical endpoints from the sampled-data rules used to represent the profile near them.

Check Your Understanding

Answer each question using the stated window definitions. Solutions are provided after the questions.

A window is specified as 0.5 to 3 hours after a recorded dose. What is its duration, and what additional detail would be needed if only clock-time boundaries were reported?
Why does including the exact concentration value at the endpoint $b$, rather than excluding it, not change the continuous partial AUC?
A start boundary moves from 2 to 2.5 hours while the end remains at 6 hours. Which portion is removed from the window?
Two profiles are summarized from 1 to 4 hours after dosing. One profile’s dose time is 07:50 and the other’s is 08:10. Are the clock-time windows identical? Are the elapsed-time windows identical?
A 2-hour event-relative window begins at 3 hours after dosing for one profile and 4 hours after dosing for another. What are the post-dose ranges, and do they represent the same fixed interval?

Solutions. (1) Its duration is $3-0.5=2.5$ hours. Clock-time boundaries alone require a stated origin or reference event to establish elapsed time after dosing. (2) A single endpoint has zero width, so changing whether that point is included does not add or remove area from the integral. (3) The portion from 2 to 2.5 hours is removed; its contribution is $\int_2^{2.5} C(t)\,dt$. (4) The clock-time windows are 08:50 to 11:50 and 09:10 to 12:10, so they are not identical. Both are 1 to 4 hours after their respective doses, so the elapsed-time windows follow the same rule. (5) The ranges are 3 to 5 hours and 4 to 6 hours after dosing. They have equal duration but are not the same fixed post-dose interval.