Why the Exposure Metric Matters
A concentration-time profile contains more information than any single number can preserve. Two patients can have the same total exposure over a dosing interval but different peaks and troughs; conversely, similar peaks can accompany different total exposure. The exposure metric used in an exposure-response analysis therefore determines which aspect of the profile is being related to the clinical outcome.
The practical goal is not to select the metric with the smallest p-value among several candidates. Instead, define the outcome and its relevant observation window, then choose a concentration summary that has a defensible pharmacologic connection to the effect and can be estimated adequately from the available data. The introductory material on estimands for exposure-response analyses helps frame which outcome and time window are being targeted; here, the focus is how to summarize exposure for that target.
What the Main Metrics Summarize
For an interval of length \(\tau\), AUC is the integral of concentration \(C(t)\) over that interval. AUC has units of concentration multiplied by time, such as mg·h/L. Dividing by the interval length gives the time-averaged concentration over that interval, with units of concentration.
\(C_{\max}\) is the highest concentration in a specified window, and \(C_{\text{trough}}\) is the concentration immediately before a scheduled dose. Both have concentration units. The window and timing matter: a peak during the first dose is not necessarily the peak at steady state, and a trough from one dosing interval may not describe exposure during a different interval.
Cumulative exposure summarizes concentration over a longer window, often across multiple doses. For example, cumulative AUC over \(T\) hours is \(\int_0^T C(t)\,dt\). If the window comprises \(m\) dosing intervals, it can also be written as the sum of the AUC values for those intervals. Its units remain concentration multiplied by time. Cumulative exposure is not interchangeable with average exposure: over a fixed window, cumulative AUC is the average concentration multiplied by the window length.
Matching the Metric to the Pharmacology
The mechanism and timing of the effect help determine which profile feature is most relevant. A brief concentration peak may be pertinent when the effect is initiated by a peak-related process. AUC may be more informative when the effect reflects overall exposure across an interval. Trough concentration may be useful when maintaining concentration near the end of each interval is central to the intended effect. Cumulative exposure can be appropriate when the proposed effect depends on total exposure accumulated over a longer period.
These are rationales to evaluate, not universal rules. For instance, if efficacy depends on remaining above a threshold concentration, neither AUC nor trough alone necessarily records how long the threshold was exceeded. A time-above-threshold summary may better express that mechanism. If the effect builds up or dissipates slowly, the concentration during the outcome window may not be the whole story; a simple interval metric can then be inadequate.
The outcome window also constrains the choice. An exposure summary should refer to the time period relevant to the outcome, and should not include concentrations occurring after that outcome when the analysis is intended to relate prior exposure to the outcome. The dosing interval matters as well: with short intervals, peaks and troughs may differ little; with longer intervals, they may separate substantially. At steady state, interval-specific metrics describe a recurring profile; before steady state, exposure can change from one interval to the next.
Worked Example: Calculating Interval AUC and Comparing Peak-Driven Candidates
Suppose a hypothetical drug is given every 12 hours. A model predicts concentrations of 1, 7, 5, 2, and 1 mg/L at 0, 2, 4, 8, and 12 hours after a dose. Assume linear interpolation between these points is adequate for a trapezoidal AUC calculation, and that the peak is captured by the sampling or model prediction.
For consecutive observations, the trapezoidal contribution is the average of the two concentrations multiplied by the time step. Thus:
The interval average concentration is \(40/12=3.33\) mg/L. The predicted \(C_{\max}\) is 7 mg/L at 2 hours, and \(C_{\text{trough}}\) is 1 mg/L immediately before the next dose. If the proposed clinical mechanism is initiated by a brief peak, \(C_{\max}\) has a direct mechanistic rationale; AUC and trough answer different questions. This calculation does not establish that the peak causes an outcome. It shows how the candidate metrics summarize this hypothetical profile.
Equal AUC Does Not Mean Equal Profiles
A short derivation clarifies why the metrics cannot be substituted for one another. By definition, interval AUC is the area under the concentration curve. The average concentration is that area divided by the interval length. Therefore, two profiles with the same AUC over the same interval have the same average concentration. But equal areas do not require equal maxima or minima: a profile can concentrate more of its area near a peak, while another maintains a flatter concentration.
This distinction is especially important when comparing dosing regimens. AUC may be similar while peak-related effects or end-of-interval coverage differ. Metric choice should preserve the profile feature that the pharmacologic rationale says matters.
Worked Example: Same AUC, Different Peaks and Troughs
Consider two hypothetical 12-hour profiles, each represented by concentrations at 0, 2, 4, 8, and 12 hours. Regimen A has concentrations 1, 7, 5, 2, and 1 mg/L; Regimen B has concentrations 2, 5, 4, 3, and 2 mg/L. Assume the same trapezoidal interpolation used above.
For Regimen B, the interval AUC is:
Regimen A also has an AUC of 40 mg·h/L, so both have an average concentration of \(40/12=3.33\) mg/L. However, Regimen A has \(C_{\max}=7\) mg/L and \(C_{\text{trough}}=1\) mg/L, whereas Regimen B has \(C_{\max}=5\) mg/L and \(C_{\text{trough}}=2\) mg/L. If maintaining concentrations above 1.5 mg/L is the proposed concern, Regimen B's higher trough is more relevant than the equal AUCs. These point summaries still do not establish exactly how long either profile stays above that threshold; that question requires examining the profile or an appropriate time-above-threshold metric.
When Cumulative Exposure Is the Relevant Summary
A cumulative metric may be considered when the proposed effect depends on exposure accumulated over an outcome-relevant period rather than on one dosing interval alone. The period must be specified: cumulative exposure during the first four days and cumulative exposure over a full treatment course are different quantities. The longer the window, the more important it is to use concentration predictions that are supported over that full period.
Worked Example: Calculating Cumulative AUC Over a Four-Day Window
Suppose a hypothetical mechanism proposes that a clinical effect relates to total concentration exposure over four treatment days. For simplicity, model-predicted mean concentrations for successive 24-hour days are 2, 3, 4, and 3 mg/L. Assume each daily mean adequately represents its day's average concentration and that all four days precede the outcome being evaluated.
The daily AUC contributions are the daily mean concentrations multiplied by 24 hours:
The average concentration over the full 96-hour window is \(288/96=3.00\) mg/L. The cumulative AUC is 288 mg·h/L, not 3 mg/L; the two quantities have different units and meanings. If the proposed mechanism instead depends on the largest daily peak, this cumulative calculation would discard that feature and would not be the natural primary metric.
A Practical Selection and Diagnostic Process
Specify when the clinical outcome is assessed and what prior exposure period is pharmacologically relevant. State the window as well as the metric.
Explain why a peak, trough, interval-wide area, or accumulated exposure should correspond to the effect. If the mechanism is uncertain, identify plausible candidates before examining their apparent associations.
Ask whether the concentration data or PK model support the relevant peak, trough, and time window. Sparse sampling may miss a brief peak; predictions far outside observed follow-up may be weakly supported.
Check whether the chosen metric varies meaningfully across individuals or regimens and whether its values agree with the underlying concentration profiles. Compare alternatives for scientific robustness, not only for statistical significance.
A metric can fail even when its formula is calculated correctly. A reported \(C_{\max}\) based on sparse observations may be a sampled maximum rather than the actual profile maximum. An AUC estimated over a window that excludes a pharmacologically important period may not reflect the intended exposure. Cumulative exposure can be sensitive to how long patients are followed and to the handling of gaps in their concentration histories.
Metrics can also be strongly correlated. In that situation, several may appear to describe the same exposure contrast, making it difficult to distinguish which feature is most informative. A smaller p-value for one metric is not by itself a mechanistic justification for choosing it. Evaluate the model predictions and the outcome window, report the rationale, and consider whether conclusions are sensitive to other scientifically reasonable metrics. An exposure-response association alone does not establish that changing exposure would change the outcome.
| Metric | What it summarizes | Potentially relevant when | Information it can miss |
|---|---|---|---|
| AUC over an interval | Total concentration over a specified dosing interval | The proposed effect relates to overall interval exposure | How exposure is distributed between peak and trough |
| \(C_{\max}\) | Highest concentration in a specified window | A peak-related process is plausible | Duration of high concentration and total interval exposure |
| \(C_{\text{trough}}\) | Concentration just before a scheduled dose | End-of-interval exposure or sustained coverage matters | Peak size and the full profile between doses |
| Cumulative exposure | Integrated concentration over a longer specified window | The proposed effect depends on accumulated exposure | When within the window exposure occurred and the profile's peaks |
Common Mistake and Limitations
Wrong reasoning: “AUC is the most complete metric because it uses the whole concentration curve, so it must be the best predictor.” Correction: AUC preserves total area but not the timing or shape of exposure. If the clinical effect depends on a peak or on maintaining a trough, AUC alone can conceal the relevant difference. Conversely, choosing \(C_{\max}\) because it is easy to interpret is not justified if the proposed effect reflects exposure across the interval.
No single metric is universally best. Each compresses a concentration-time profile and therefore discards information. Metric choice also depends on reliable estimation: a theoretically appropriate metric may be uninformative if the relevant window is poorly observed or the PK predictions are uncertain. The conclusion should be limited to the exposure summary and window actually analyzed.
Check Your Understanding
Answer each question, then compare with the brief solutions below.
- A 10-hour dosing interval has AUC of 50 mg·h/L. What is the interval average concentration, with units?
- Two profiles have equal AUC over the same interval. Must they have equal \(C_{\max}\) and \(C_{\text{trough}}\)? Explain.
- A proposed effect depends on concentration immediately before each dose. Which of the four main metrics is the most direct candidate, and what must be specified?
- Daily mean concentrations are 1.5 and 2.5 mg/L over two successive 24-hour periods. Calculate cumulative AUC over the 48-hour window.
- Why might a sampled maximum be a poor estimate of \(C_{\max}\)?
Solutions. (1) \(50/10=5.00\) mg/L. (2) No. Equal AUC implies equal average concentration for the same interval length, but different profiles can have different maxima and troughs. (3) \(C_{\text{trough}}\), with the scheduled dose timing and relevant dosing interval specified. (4) \((1.5\times24)+(2.5\times24)=36+60=96\) mg·h/L. (5) Sparse sampling can miss the true peak, so the largest observed concentration may be lower than the profile maximum.