Why Equal Total AUC Does Not Mean the Same Profile
Total AUC summarizes concentration accumulated over a specified time interval. It is useful for comparing overall exposure, but it does not show when that exposure occurred. One profile can be higher early and lower later than another, with the gains and losses balancing to the same total area. When the ordering of two profiles exchanges over time, the profiles cross.
This distinction matters in pharmacokinetic interpretation. Two treatments may have the same estimated AUC over the interval being examined but differ in the timing of concentrations. Conversely, profiles may look quite different at some times even when their total AUCs are close. Equal total AUC alone does not establish that the profiles are identical or that a formal bioequivalence conclusion follows.
The earlier material on AUC as a Numerical Integral established that AUC can be estimated from concentration measurements using a specified numerical integration rule. Here, the comparison is between two such estimated profiles, evaluated on the same time interval with the same time points and integration rule. The examples use the linear trapezoidal rule.
Signed Areas Explain the Balance
For matched measurement times \(t_0,\ldots,t_n\), let \(D_i=C_A(t_i)-C_B(t_i)\) and let \(\Delta t_i=t_{i+1}-t_i\). With linear trapezoids, the difference between the two estimated total AUCs is the sum of the signed trapezoidal areas for the difference profile. A positive interval contribution means profile A contributes more area on that interval; a negative contribution means profile B contributes more.
This equality follows directly from the trapezoid formula. On interval \(i\), the estimated area for A is \(\Delta t_i(C_A(t_i)+C_A(t_{i+1}))/2\), and the corresponding area for B is \(\Delta t_i(C_B(t_i)+C_B(t_{i+1}))/2\). Subtracting the second expression from the first gives \(\Delta t_i(D_i+D_{i+1})/2\). Summing over intervals gives the displayed result.
Therefore, equal estimated total AUCs require the signed interval contributions to sum to zero. Positive contributions can be balanced by negative ones. This is why a profile can be above another during one part of the interval and below it during another without having a different total area.
With linear interpolation, a crossing can also be located from the endpoint differences. If \(D_i\) and \(D_{i+1}\) have opposite signs, then the zero occurs at fraction \(u=-D_i/(D_{i+1}-D_i)\) of the way across the interval, where \(0<u<1\). Its time is \(t_i+u\Delta t_i\). This calculation identifies a crossing of the interpolated segments, not necessarily the exact crossing of the underlying, unobserved concentration curves.
Worked Example: Equal AUC With One Exchange in Ordering
Suppose two hypothetical profiles have the following concentrations, in mg/L, at times 0, 1, 2, and 3 h. Use linear trapezoids over 0–3 h.
| Time (h) | Profile A (mg/L) | Profile B (mg/L) | Difference A − B (mg/L) |
|---|---|---|---|
| 0 | 0 | 0 | 0 |
| 1 | 4 | 2 | 2 |
| 2 | 2 | 4 | −2 |
| 3 | 0 | 0 | 0 |
For profile A, the three trapezoid areas are \((0+4)(1)/2=2\), \((4+2)(1)/2=3\), and \((2+0)(1)/2=1\). Thus \(\widehat{\mathrm{AUC}}_A=2+3+1=6\ \text{mg·h/L}\).
For profile B, the areas are \((0+2)(1)/2=1\), \((2+4)(1)/2=3\), and \((4+0)(1)/2=2\). Thus \(\widehat{\mathrm{AUC}}_B=1+3+2=6\ \text{mg·h/L}\). The estimated totals are equal.
The differences at 1 and 2 h change from \(2\) to \(-2\) mg/L, so the profiles exchange ordering between those times. Under linear interpolation, the difference falls from 2 to −2 mg/L across 1 h. Its zero is halfway across: \(u=-2/(-2-2)=0.5\), giving a crossing time of \(1+0.5(1)=1.5\) h.
The early interval from 0 to 1 h has area 2 mg·h/L for A and 1 mg·h/L for B. The later interval from 1 to 3 h has area 4 mg·h/L for A and 5 mg·h/L for B. The 1 mg·h/L early advantage for A is balanced by a 1 mg·h/L later advantage for B. Equal total AUC therefore coexists with a difference in when exposure occurs.
Multiple Crossings Can Also Balance to Zero
A profile need not cross only once. Several exchanges in ordering can produce positive and negative signed areas that cancel overall. Checking only the beginning and end of an interval, or looking at one selected time, can miss this pattern. The full sequence of matched measurements is needed to describe the ordering represented by the sampled profiles.
Worked Example: Two Ordering Changes With Equal Total AUC
Consider these hypothetical concentrations, measured at hourly intervals from 0 to 4 h. The data use mg/L, and AUC is estimated by the linear trapezoidal rule.
| Time (h) | Profile A (mg/L) | Profile B (mg/L) | Difference A − B (mg/L) |
|---|---|---|---|
| 0 | 0 | 0 | 0 |
| 1 | 6 | 2 | 4 |
| 2 | 2 | 8 | −6 |
| 3 | 4 | 2 | 2 |
| 4 | 0 | 0 | 0 |
Profile A has interval areas \((0+6)/2=3\), \((6+2)/2=4\), \((2+4)/2=3\), and \((4+0)/2=2\), each in mg·h/L. Its total is \(3+4+3+2=12\ \text{mg·h/L}\).
Profile B has interval areas \((0+2)/2=1\), \((2+8)/2=5\), \((8+2)/2=5\), and \((2+0)/2=1\). Its total is \(1+5+5+1=12\ \text{mg·h/L}\). The estimated total AUCs are equal.
The difference changes from \(4\) to \(-6\) mg/L between 1 and 2 h, then from \(-6\) to \(2\) mg/L between 2 and 3 h. Thus there are two changes in ordering. Under linear interpolation, the first zero occurs at fraction \(4/(4+6)=0.4\) of the 1–2 h interval, or 1.4 h. The second occurs at fraction \(6/(6+2)=0.75\) of the 2–3 h interval, or 2.75 h.
The signed interval areas for A minus B are \((0+4)/2=2\), \((4-6)/2=-1\), \((-6+2)/2=-2\), and \((2+0)/2=1\) mg·h/L. Their sum is \(2-1-2+1=0\), matching the difference between the total AUC estimates. The profiles exchange ordering twice, yet the total signed area difference is zero.
Similar Totals Are Not Necessarily Equal
In practice, profiles may have close but nonidentical AUC estimates. It is important to report the calculated difference rather than describe a small numerical gap as zero. Whether a difference is important or acceptable cannot be decided from visual similarity or the AUC arithmetic alone; it depends on the comparison and its relevant evaluation criteria.
Worked Example: Crossing Profiles With Close, Unequal AUCs
Suppose hypothetical profiles A and B are measured at 0, 1, 2, and 3 h. Their concentrations are \(0,4,2,0\) mg/L for A and \(0,2,4.2,0\) mg/L for B. Use linear trapezoids.
For A, the interval areas are \((0+4)/2=2\), \((4+2)/2=3\), and \((2+0)/2=1\) mg·h/L. Therefore, \(\widehat{\mathrm{AUC}}_A=6\ \text{mg·h/L}\).
For B, the interval areas are \((0+2)/2=1\), \((2+4.2)/2=3.1\), and \((4.2+0)/2=2.1\) mg·h/L. Therefore, \(\widehat{\mathrm{AUC}}_B=6.2\ \text{mg·h/L}\). The absolute difference, A minus B, is \(-0.2\ \text{mg·h/L}\); B’s estimate is higher.
The difference values are \(0,2,-2.2,0\) mg/L, so the profiles exchange ordering between 1 and 2 h. The crossing occurs at fraction \(2/(2+2.2)=0.4762\) of that interval, approximately 1.476 h under linear interpolation. The total AUC estimates are close but not equal: the difference is not erased by the crossing.
This example separates two questions: whether the profiles exchange ordering, and whether their integrated areas match. A crossing answers the first; summing interval areas answers the second. Neither fact alone supplies a full assessment of the profiles.
Assumptions, Diagnostics, and Common Pitfalls
The numerical comparisons above rely on matched quantities: the same analyte and concentration units, the same time interval, and the same integration rule. The examples also use the same observation times for both profiles, so each trapezoid spans the same interval. If the time grids differ, first align the profiles on a common set of times using a stated interpolation approach; otherwise, apparent area differences may reflect inconsistent calculations rather than the profiles themselves.
Linear interpolation is an explicit modeling choice between observations. A calculated crossing between sample times is a crossing of the interpolated segments. Sparse sampling cannot establish that the underlying biological curves crossed at precisely that time, or rule out additional crossings between measurements. Similarly, equal trapezoidal AUCs are equality of estimates under the chosen rule and observed interval, not proof that the unknown continuous-time curves have identical integrals.
Useful checks include recalculating interval areas individually, confirming the time increments and units, and verifying that the signed interval differences sum to the reported total difference. Plotting both profiles on the same axes can help reveal ordering changes, but a plot is not a substitute for checking the arithmetic. If the analyzed intervals, time points, or treatment of unobserved portions differ, a direct interpretation of the totals requires caution.
- Mistake: “The AUCs are equal, so the profiles are the same.” Equal areas constrain only the total integral. As the examples show, concentrations can differ and the profiles can cross while the signed differences cancel.
- Mistake: “A crossing guarantees equal AUC.” A crossing identifies a change in ordering, not the size of the positive and negative areas. The third example crosses but has a 0.2 mg·h/L AUC difference.
- Mistake: “A small difference is automatically negligible.” Report the difference and units, then interpret it against the relevant scientific or regulatory criteria. A close numerical total is not, by itself, a bioequivalence conclusion.
Total AUC and window-specific AUC answer different questions: the total summarizes area across the full stated interval, while a window-specific value describes area over a selected part of it. The examples use early and later areas only to show how a total can balance; they do not imply that an overall total captures every feature of exposure timing.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions.
- Two profiles have signed trapezoidal interval differences of \(1.5\), \(-2.0\), and \(0.5\) mg·h/L. What is the difference between their total AUC estimates, A minus B?
- At the endpoints of a 2 h interval, the concentration differences A minus B are \(3\) and \(-1\) mg/L. Under linear interpolation, when do the profiles cross?
- Can profiles with equal total AUC differ in concentration at sampled times? Explain what must balance for their estimated totals to match.
- Profiles have AUC estimates of 10.0 and 10.4 mg·h/L and cross once. Is it correct to report equal AUC? What can be concluded from these facts alone?
Solutions. 1. The total difference is \(1.5-2.0+0.5=0\ \text{mg·h/L}\), so the estimates are equal under the stated calculation. 2. The difference declines by \(3-(-1)=4\) mg/L; the zero is \(3/4=0.75\) of the interval. The crossing time is \(0+0.75(2)=1.5\) h. 3. Yes. Different concentrations are compatible with equal totals when the positive and negative signed interval areas sum to zero. 4. No. The estimates differ by \(10.0-10.4=-0.4\ \text{mg·h/L}\). The crossing shows an exchange in ordering, but these facts alone do not establish whether the difference is important or whether the profiles meet bioequivalence criteria.