From Concentration-Time Profiles to an AUC Ratio
A simulated concentration-time profile contains more information than a single exposure summary: it shows when concentrations rise and fall, as well as their magnitude. An AUC ratio condenses the profile into a comparison of total exposure under two specified regimens. To make that comparison meaningful, the profiles must refer to clearly defined control and interaction conditions and must be integrated over compatible time windows.
In a drug-drug interaction (DDI) prediction, the control condition commonly represents the victim drug administered without the perpetrator, while the interaction condition represents the victim drug administered with the perpetrator. The ratio is not a property of the concentration profile alone. It depends on the dose, regimen, population or simulated individual, concentration measure, and time interval chosen for both conditions.
For example, if concentrations are expressed in mg/L and time in hours, the AUC is in mg·h/L. An \(AUCR\) of 1 means the two specified AUCs are equal; a value greater than 1 means the interaction-condition AUC is larger than the control AUC; a value below 1 means it is smaller. These descriptions apply to the stated endpoint and conditions, not automatically to every dose, time window, or patient.
Choose the Exposure Window Before Integrating
For a single-dose prediction, the endpoint may be \(AUC_{0-\infty}\), if the profile is followed sufficiently far and the remaining tail can be estimated appropriately. For repeated dosing at steady state, a common endpoint is \(AUC_{\tau}\), the area over one dosing interval of length \(\tau\). These are different endpoints: \(AUC_{0-\infty}\) after one dose is not interchangeable with \(AUC_{\tau}\) at steady state.
The control and interaction profiles must use the same endpoint definition. For an \(AUC_{\tau}\) comparison, integrate each profile over the same dosing interval, such as 0 to 24 hours when \(\tau=24\) hours. For a partial AUC, use the same specified start and end times in both conditions. Comparing a control \(AUC_{0-24}\) with an interaction \(AUC_{0-12}\), for instance, mixes exposure change with a change in the length of the observation window.
At steady state, a repeated-dose concentration profile is periodic: corresponding points in consecutive dosing intervals have the same concentration. A simulated interval should therefore be taken after the model has reached the intended steady state if the target is \(AUC_{\tau}\) at steady state. If the prediction concerns a particular transient period, such as a specified day after perpetrator dosing begins, use that period in both conditions as defined by the study question; do not label a transient-window area as steady-state \(AUC_{\tau}\).
Numerical Integration of Simulated Profiles
A simulation usually reports concentrations at discrete times rather than every instant. With observations \((t_i,C_i)\), the linear trapezoidal rule estimates the area between adjacent times by assuming concentration changes linearly between those points. The estimated total AUC is the sum of the trapezoid areas.
Each term has units of concentration multiplied by time. The formula uses the actual time differences, so unevenly spaced samples are allowed. It does not make the underlying simulated profile linear; it defines a numerical approximation based on the reported points. Accuracy depends on whether the sampling grid captures the profile’s important changes. More points may be needed where concentrations change rapidly or where the curve has pronounced curvature.
Worked Example: Integrating Two Matched Profiles
Suppose a hypothetical simulation gives victim-drug concentrations at the same times in a 12-hour comparison window. The victim dose and route are identical in the control and interaction conditions. Concentrations are in mg/L.
| Time (h) | Control (mg/L) | With perpetrator (mg/L) |
|---|---|---|
| 0 | 0 | 0 |
| 2 | 4 | 3 |
| 4 | 6 | 5 |
| 8 | 3 | 3 |
| 12 | 1 | 1.5 |
The control AUC is the sum of four trapezoids:
For the interaction condition:
Thus \(AUCR=36/40=0.900\). Over this specified 12-hour window, the interaction-condition AUC is 90.0% of the control AUC, or 10.0% lower. This result describes the AUC comparison; it does not establish that concentrations were 10.0% lower at every time.
Constructing a Matched DDI Contrast
The intended contrast is usually the change associated with adding the perpetrator to a defined victim-drug regimen. In a simulation, this means running or extracting two conditions that differ in the interaction treatment while holding relevant design choices consistent. At a minimum, use the same victim-drug dose, route, formulation, concentration measure, integration window, and subject or covariate values when calculating a within-subject or within-simulated-individual contrast.
The perpetrator regimen must also be represented as intended. If inhibition or induction changes over time, the interaction profile should reflect the specified perpetrator dosing history and the chosen observation window. It is not valid to compare a control profile generated under one schedule with an interaction profile taken from an unrelated time point simply because both are labelled “control” and “interaction.”
A dose difference changes the interpretation. If the victim dose differs between conditions, the raw AUC ratio combines the effect of the dose difference with the effect of the interaction. Dose normalization may be defensible under a suitable linear pharmacokinetic assumption, but it is not an automatic correction: with nonlinear pharmacokinetics, exposure need not scale proportionally with dose. For a direct DDI contrast, the clearest comparison generally uses the same victim dose in both conditions.
Worked Example: Steady-State AUC Over a 24-Hour Interval
Consider hypothetical steady-state profiles for a victim drug dosed every 24 hours. The samples cover exactly one interval, and the concentrations at 0 and 24 hours agree within each condition, as expected for corresponding points in a periodic steady-state profile. The dose, route, individual, and sampling times match.
| Time (h) | Control (mg/L) | With perpetrator (mg/L) |
|---|---|---|
| 0 | 2 | 1 |
| 4 | 8 | 7 |
| 8 | 5 | 4 |
| 12 | 3 | 2 |
| 24 | 2 | 1 |
Using the linear trapezoidal rule, the control \(AUC_{\tau}\), with \(\tau=24\) hours, is
The interaction-condition area is
Therefore, \(AUCR=68/92=0.7391\). For this simulated 24-hour steady-state interval, the interaction-condition AUC is about 73.9% of the control value. The interval is integrated in full in both conditions, including the 12-hour final segment; treating the last sampling gap as four hours would produce an incorrect area.
Paired Profiles and Population Summaries
When the model includes multiple matched individuals, first calculate each individual’s control and interaction AUC using the same rules. A within-individual ratio is \(R_j=AUC_{\mathrm{interaction},j}/AUC_{\mathrm{control},j}\), where \(j\) indexes the simulated individual. This preserves the intended paired contrast: each individual serves as their own control for the simulated comparison.
For a positive AUC ratio, a common summary across individuals is the geometric mean ratio. It is calculated by averaging the logarithms of the individual ratios and then exponentiating. This differs in general from dividing the arithmetic mean interaction AUC by the arithmetic mean control AUC. The two calculations answer different questions, especially when baseline AUCs vary across individuals.
Worked Example: Summarizing Two Paired Simulated Individuals
Suppose two hypothetical individuals have concentrations at 0, 12, and 24 hours in matched control and interaction conditions. The endpoint is steady-state \(AUC_{\tau}\) for \(\tau=24\) hours, and each profile has matching concentrations at the interval boundaries.
| Individual and condition | Concentrations at 0, 12, 24 h (mg/L) |
|---|---|
| Individual A, control | 1, 3, 1 |
| Individual A, interaction | 1, 5, 1 |
| Individual B, control | 2, 4, 2 |
| Individual B, interaction | 2, 6, 2 |
For individual A, the control AUC is \(12(1+3)/2+12(3+1)/2=24+24=48\) mg·h/L. The interaction AUC is \(12(1+5)/2+12(5+1)/2=36+36=72\) mg·h/L, giving \(R_A=72/48=1.50\).
For individual B, the control AUC is \(12(2+4)/2+12(4+2)/2=36+36=72\) mg·h/L. The interaction AUC is \(12(2+6)/2+12(6+2)/2=48+48=96\) mg·h/L, giving \(R_B=96/72=1.333\).
The geometric mean ratio is
For comparison, the ratio of arithmetic mean AUCs is \((72+96)/(48+72)=168/120=1.40\). It is not the same calculation as the geometric mean of the paired ratios. With only two hypothetical individuals, these values simply illustrate the distinction; they do not provide a population-level uncertainty assessment.
Why the Profile-Based Calculation Matters
AUC ratios are often predicted from a mechanistic summary, such as a change in clearance, when the required assumptions are appropriate. For a linear system with the same dose and bioavailability, total exposure after a dose is inversely related to clearance. In that special setting, an exposure ratio may be derived from a clearance ratio. However, the simulated-profile calculation directly evaluates the chosen concentration-time endpoint and does not require the entire profile to be a scaled copy of the control profile.
A short justification shows why a uniform scaling changes AUC by the same factor. If the interaction profile satisfies \(C_{\mathrm{interaction}}(t)=kC_{\mathrm{control}}(t)\) throughout the same interval, then
But interaction profiles need not be uniformly scaled. An interaction can alter the timing or shape of concentrations, and an incomplete or partial AUC may respond differently from a full-interval AUC. Calculating both areas from their profiles captures those differences within the chosen endpoint. It does not, by itself, explain which mechanism caused them.
A clearance-based shortcut and profile integration are therefore complementary, not interchangeable in every situation. The shortcut can be efficient when its assumptions and endpoint align with the prediction. Profile integration is useful when the model produces time-varying concentrations and the study endpoint is explicitly defined over a time window. Neither method makes a mismatched control or an inappropriate interval valid.
Diagnostics, Pitfalls, and Limitations
- Check the time grid. Confirm that both profiles cover the full requested interval, use the correct time units, and include the interval boundaries. A missing late segment can materially bias a full-interval AUC.
- Check alignment. Verify that each concentration is paired with its actual sampling time and that control and interaction conditions use the same window. If the grids differ, integrate each on its own reported times over the same endpoints; do not pair values merely by row number.
- Check the regimen and state. Make sure the simulated profile corresponds to the stated single-dose, transient, or steady-state condition. At periodic steady state, compare corresponding dosing intervals rather than an interval from one condition with a noncorresponding interval from another.
- Inspect sparse or rapidly changing regions. Large gaps around steep rises, peaks, or sharp declines can make linear interpolation inaccurate. Add simulation output points where needed and compare estimates using a sufficiently refined grid.
- Keep the contrast interpretable. If victim dose, formulation, or route differs, state that the raw ratio combines those differences with the interaction. Do not attribute the entire change to the perpetrator without justification.
- Do not confuse total and partial AUCs. The same integration method can calculate either, but the resulting ratios represent different endpoints. A ratio over a specified early window is not a substitute for the full-interval ratio.
The trapezoidal estimate is only as reliable as the simulated concentrations and time resolution supplied to it. Integration cannot correct a flawed model, an inappropriate concentration measure, or an incorrectly specified perpetrator schedule. The ratio is also specific to the simulated conditions; translating it to a different dose or population requires additional justification.
Check Your Understanding
Answer each question using the endpoint and conditions stated. Solutions are provided after the questions.
- A control profile covers 0–24 hours and an interaction profile covers 0–12 hours. Can their AUCs be divided to obtain a matched 24-hour AUC ratio? What must be done instead?
- For two successive concentration points, \(C_i=2\) mg/L at 3 hours and \(C_{i+1}=6\) mg/L at 7 hours, what is the trapezoidal area for that segment?
- A matched control AUC is 80 mg·h/L and the interaction AUC is 100 mg·h/L. Calculate and interpret the AUC ratio.
- Why can the ratio of arithmetic mean AUCs differ from the geometric mean of individual paired AUC ratios?
Solutions. (1) No. The windows have different lengths; obtain or define profiles over the same 0–24-hour interval before calculating a 24-hour ratio. (2) The segment area is \((2+6)(7-3)/2=16\) mg·h/L. (3) \(AUCR=100/80=1.25\), so the interaction-condition AUC is 25.0% greater for the stated matched endpoint. (4) The ratio of arithmetic means weights AUC values through the group totals, whereas the geometric mean summarizes individual ratios on the logarithmic scale; with variation in baseline AUCs or individual effects, the results need not coincide.