Why a Regulatory Criterion May Focus on Part of the Profile
Total AUC summarizes exposure over an interval, while Cmax summarizes the observed peak concentration. Neither necessarily answers every product-specific question about the timing of exposure. For some products, regulatory guidance may therefore identify a partial AUC (pAUC): exposure accumulated over a specified segment of the concentration-time profile. The segment might be early after dosing, or it might be another interval identified as relevant for that product.
A partial AUC is not a generic substitute for total AUC or Cmax. Its regulatory meaning depends on the product-specific guidance and the study protocol: these specify the endpoint, the time boundaries, and the comparison or acceptance criterion. The boundaries are part of the definition of the endpoint, not a choice to make after looking at which interval produces a favorable result.
The integral describes the underlying curve; in a study, it is estimated from sampled concentrations using the prespecified calculation method. If a boundary falls between sampling times, the method must specify how the boundary value is obtained. Sampling should be adequate to characterize the segment, especially when concentrations change rapidly there. The earlier material on AUC calculation and sampling schedules provides the background for numerical integration; here the key point is that the selected interval determines what is integrated.
From a Specified Window to a Regulatory Endpoint
A regulator may request a partial AUC when exposure in a particular time window is considered informative for the product's performance. For example, a guidance document may identify early exposure as relevant and specify a cutoff time. The choice of cutoff should come from the applicable guidance or a justified, prospectively specified protocol—not from an exploratory search across many possible cutoffs.
The endpoint must be stated precisely. “Early exposure” alone is not a reproducible definition. A usable definition gives the interval, such as \(\operatorname{pAUC}_{0,2\text{ h}}\), and the method for calculating it. That definition also requires compatible time origins, units, sampling schedules, and analyte measurements for the products being compared. Otherwise, an apparent difference could reflect inconsistent endpoint construction rather than a product difference.
In a typical ratio-based comparison, the partial AUC is calculated for each participant and treatment, and the prespecified statistical analysis estimates the test-to-reference ratio. Positive pAUC values are commonly analyzed on the logarithmic scale; the estimated log ratio and its confidence interval are then exponentiated to express the result as a ratio. The applicable regulatory document determines the required analysis and decision limits. Conventional limits must not be assumed to apply to every pAUC.
The central decision question is whether the required interval for the pAUC ratio satisfies the product-specific acceptance criterion. The pAUC may be required alongside AUC and Cmax, or a guidance document may specify another role for it. A favorable pAUC result does not override a failed required endpoint, and a pAUC should not be added or omitted based on observed results.
Worked Example: Calculating a Partial AUC Over a Specified Window
Suppose a hypothetical product-specific protocol specifies \(\operatorname{pAUC}_{0,2\text{ h}}\). The following are hypothetical concentration profiles, in mg/L, for the reference and test products. Use the linear trapezoidal rule for this illustration.
| Time (h) | Reference concentration (mg/L) | Test concentration (mg/L) |
|---|---|---|
| 0 | 0 | 0 |
| 1 | 8 | 10 |
| 2 | 12 | 14 |
| 4 | 6 | 7 |
| 6 | 2 | 2 |
Reference partial AUC. Only the intervals from 0 to 1 h and from 1 to 2 h contribute:
Test partial AUC. Applying the same calculation to the test concentrations:
The ratio of these two illustrative areas is \(17/14=1.214\), or about 121.4%. This is a descriptive comparison of the supplied profiles, not a bioequivalence decision: it has no participant-level variability or confidence interval. The calculation also shows why the protocol’s 2 h boundary matters. Concentrations measured after 2 h do not enter this endpoint.
How the Partial-AUC Comparison Enters the Decision
A regulatory decision is based on the prespecified statistical analysis, not simply on whether two point estimates look close. For a ratio-based criterion, a confidence interval for the test-to-reference pAUC ratio is compared with the required limits. Both the confidence level and the limits must come from the applicable protocol or guidance. A 90% confidence interval is used in many average bioequivalence settings, but the required approach for a particular pAUC must be checked rather than presumed.
The log-scale calculation has a direct interpretation. If the estimated log ratio is \(\hat{\theta}\), its standard error is \(SE\), and the relevant \(t\) critical value is \(t_{1-\alpha/2,df}\), a two-sided interval on the log scale has endpoints \(\hat{\theta}\pm t_{1-\alpha/2,df}SE\). Exponentiating both endpoints gives the confidence interval for the ratio. The interval, not just the point estimate, is compared with the prespecified ratio-scale limits.
Worked Example: Evaluating a Hypothetical pAUC Criterion
Assume, solely for illustration, that a hypothetical product-specific protocol requires a two-sided 90% confidence interval for a pAUC ratio to lie within 0.80 to 1.25. Suppose the estimated log test-to-reference ratio is \(0.03922\), its standard error is \(0.04500\), and \(df=23\). The two-sided 90% critical value is \(t_{0.95,23}\approx1.714\).
Check the interval calculation. The margin of error on the log scale is:
Therefore, the log-scale interval is approximately:
Exponentiating gives the ratio-scale interval:
The estimated ratio is \(\exp(0.03922)\approx1.040\), or 104.0%. The entire 90% interval, from about 96.28% to 112.34%, lies within the hypothetical limits of 80% to 125%. Thus, the pAUC criterion is met under this illustrative rule. This conclusion is limited to that endpoint and rule; other required criteria may also matter. The limits in this example are not asserted to be universal pAUC limits.
The critical value must match both the confidence level and degrees of freedom. For a two-sided 90% interval with 23 degrees of freedom, the relevant quantile is \(t_{0.95,23}\approx1.714\), not the larger critical value used for a 95% interval. Using the wrong quantile changes the interval and can change the decision.
Why the Window Can Change the Answer
A partial AUC weights concentrations only within its selected time window. Total AUC can conceal a difference concentrated in one segment because exposure elsewhere contributes to the total. The following arithmetic illustrates this point, but it is not a statistical bioequivalence analysis.
Worked Example: A Similar Total AUC Can Mask an Early Difference
Consider hypothetical exposure summaries in which the reference product has an early pAUC of 20 mg·h/L from 0 to 2 h and a later area of 80 mg·h/L from 2 to 6 h. Suppose the test product has corresponding areas of 26 and 72 mg·h/L.
The early partial-AUC ratio is:
or 130%. The later-area ratio is \(72/80=0.900\), or 90%. The total areas are 100 mg·h/L for the reference and 98 mg·h/L for the test, so the total-AUC ratio is:
or 98%. The near-unity total ratio does not remove the early-window difference. If the applicable guidance requires an early pAUC comparison, the early endpoint must be evaluated using the prescribed statistical criterion; the total-area ratio alone cannot answer that requirement. Conversely, the early point estimate alone does not determine whether a statistical criterion is met.
This example also shows why the interval cannot be selected after inspecting results. Choosing 0–2 h because it appears different, or choosing 0–6 h because the ratio appears closer to one, would make the endpoint data-dependent. The regulated endpoint and analysis must instead be prespecified.
Diagnostics, Alternatives, and Limitations
Several checks help establish whether a pAUC comparison is interpretable. First, verify that the endpoint boundaries match the relevant guidance and are applied consistently to both products. Confirm that sampling adequately covers the chosen interval and that any interpolation at a boundary follows the prespecified method. Check units, time origin, concentration handling, and missing or nonquantifiable observations. These issues can matter especially when the selected segment contains rapid concentration changes.
A pAUC is also sensitive to the chosen window. Moving a cutoff can alter the endpoint and its ratio, so a result should not be generalized to a different interval without a new definition and analysis. A pAUC may be estimated imprecisely if few samples inform the interval. Finally, the regulatory reason for selecting a window is product-specific: a measured difference in partial exposure is not, by itself, proof of a clinical difference or its absence.
| Endpoint | What it summarizes | Key limitation for this question |
|---|---|---|
| Total AUC | Exposure over the specified overall interval | Can combine segments that differ in opposite directions |
| Partial AUC | Exposure within a specified time window | Depends on the prespecified boundaries and adequate sampling there |
| Cmax | Observed peak concentration | Does not summarize accumulated exposure over a window |
These measures are alternatives only in the limited sense that they summarize different features of a profile. They are not interchangeable. The applicable guidance determines whether a pAUC is required and how it contributes alongside other required endpoints. The all-criteria decision rule for multiple primary endpoints is a separate issue; here, the important point is that an endpoint specified as required must be assessed as directed, rather than replaced by a more favorable summary.
- Choosing the window after seeing results. This can bias the comparison. Use the boundaries specified before examining comparative results.
- Treating any pAUC ratio near 1 as a pass. A point estimate is not enough; apply the prescribed confidence-interval and limit rule.
- Assuming one set of limits applies to every pAUC. Confirm the product-specific criterion instead of importing limits from another endpoint or product.
- Calling the endpoint “early AUC” without defining it. State both boundaries, units, and calculation method so the measure is reproducible.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions.
- A protocol defines \(\operatorname{pAUC}_{0,3\text{ h}}\). Which concentration-time segment contributes to this endpoint?
- Why can a total-AUC ratio close to 1 fail to describe a difference in early exposure?
- A hypothetical pAUC ratio estimate is 1.02, but its required confidence interval extends beyond an acceptance limit. Does the point estimate alone establish that the criterion is met?
- Why should the time window not be selected after comparing test and reference results?
- For a two-sided 90% confidence interval with 23 degrees of freedom, should the critical value be approximately 1.714 or 2.069? Explain.
Solutions. (1) The curve from time zero through 3 h contributes; concentrations after 3 h do not. (2) Total AUC combines exposure across intervals, so a larger early area can be offset by a smaller later area. (3) No. The full required interval must satisfy the prespecified limits; a favorable point estimate is insufficient. (4) Selecting a window after seeing results makes the endpoint data-dependent and can favor a desired conclusion. (5) Approximately 1.714: a two-sided 90% interval uses \(t_{0.95,23}\). The value 2.069 is approximately the two-sided 95% critical value for 23 degrees of freedom.