Why Compare Victim-Drug Exposure?
A drug interaction study asks whether exposure to a victim drug changes when a perpetrator drug is present. The perpetrator may alter processes that affect the victim drug’s clearance or bioavailability. A higher or lower area under the concentration–time curve (AUC) is an empirical interaction effect; by itself, it does not establish which process changed or why.
The central comparison is exposure with the perpetrator against exposure without it. A paired design is especially informative when each participant receives the victim drug under both conditions, because the comparison can then account for stable differences between participants. Translating an exposure ratio into a dose recommendation requires further assumptions: the relationship between dose and exposure must be sufficiently proportional, and the intended exposure target and clinical context must be specified.
Design and Estimand
A clear comparison needs conditions that differ in the intended perpetrator exposure while keeping other relevant features as comparable as possible. In a paired study, each participant receives both conditions, typically with an appropriate washout and control of period, dose, sampling, and concomitant treatments. The washout must be long enough to avoid material carryover from the first condition into the second. The perpetrator exposure should also be adequate for the interaction question being studied.
For AUC, specify the dosing route and the AUC interval, such as AUC over a dosing interval at steady state or AUC from zero to infinity after a single dose. If doses differ between conditions, a dose-normalized comparison may be needed, but that normalization assumes exposure is proportional to dose over the relevant range. Inadequate sampling, uncertain terminal-phase estimation, noncomparable AUC definitions, or substantial deviations from the planned conditions can compromise the comparison.
Exposure ratios are commonly analyzed on the log scale. Let \(Y_i=\ln(R_i)\). The mean of the paired log ratios, when exponentiated, estimates the geometric mean exposure ratio. This is a natural summary when effects are multiplicative: a doubling and a halving are equally distant from no change on the log scale. It is not the same as taking the arithmetic mean of the untransformed ratios.
Worked Example: A Paired Exposure Ratio and Confidence Interval
Worked Example: A Paired Exposure Ratio and Confidence Interval
Suppose six hypothetical participants receive the same victim-drug dose under both conditions. Their paired AUC ratios \(AUC_P/AUC_C\), rounded for display, are 1.774, 1.846, 1.922, 2.082, 2.167, and 2.255. For calculation, suppose these ratios were constructed as \(2\exp(d_i)\), with log deviations \(d_i-\ln(2)\) equal to \(-0.12,-0.08,-0.04,0.04,0.08,0.12\). The calculations use the stated log deviations, not the rounded display ratios.
The deviations sum to zero, so the mean log ratio is \(\bar{Y}=\ln(2)=0.6931\). The estimated geometric mean ratio is therefore \(\exp(\bar{Y})=2.000\). In this hypothetical paired sample, victim-drug AUC is estimated to be twice as high with the perpetrator.
The sample variance of the log ratios equals the sample variance of the deviations. The sum of squared deviations is \(2(0.12^2+0.08^2+0.04^2)=0.0448\). Thus \(s_Y^2=0.0448/(6-1)=0.008960\), and \(s_Y=\sqrt{0.008960}=0.09466\). The standard error is \(SE=s_Y/\sqrt{n}=0.09466/\sqrt{6}=0.03864\).
Assuming the six participants’ log ratios are independent observations from a population with an approximately normal distribution, use a \(t\) interval with \(df=5\). The 95% interval on the log scale is \(0.6931\pm 2.571(0.03864)\), or approximately \([0.5938,0.7925]\). Exponentiating both limits gives \([1.811,2.209]\).
Under these assumptions, the 95% confidence interval for the population geometric mean AUC ratio is approximately 1.811 to 2.209. The interval describes uncertainty in that population ratio; it does not mean there is a 95% probability that the fixed population ratio lies in this particular interval. Because the interval is entirely above 1, it is consistent with increased victim-drug exposure in the perpetrator condition.
From AUC Ratios to Clearance and Bioavailability
The interpretation of an AUC ratio depends on route. For linear pharmacokinetics after intravenous administration, AUC is dose divided by clearance. For oral administration, AUC also depends on the fraction of the dose reaching systemic circulation, represented by bioavailability \(F\). These relationships show both what an AUC comparison can estimate and what it cannot identify without additional information.
For the same dose in both conditions, the intravenous AUC ratio is \(AUC_{P}/AUC_{C}=CL_C/CL_P\). Thus, under linear pharmacokinetics, an intravenous ratio of 2 corresponds to clearance with the perpetrator being half the control clearance. For oral dosing, the ratio is \((F_P/F_C)(CL_C/CL_P)\). An oral AUC increase could therefore reflect reduced clearance, increased bioavailability, or a combination. Oral AUC alone cannot separate those contributions.
Worked Example: Interpreting an Oral AUC Ratio Conditionally
In a hypothetical oral study, the victim drug is given at 100 mg in both conditions. Suppose mean AUC is 500 mg·h/L without the perpetrator and 1,000 mg·h/L with it. The ratio is \(R=1{,}000/500=2.00\). If bioavailability is unchanged and pharmacokinetics are linear, then \(CL_P/CL_C=1/R=0.500\): clearance with the perpetrator is estimated to be half of control clearance.
If the objective were to reproduce the control AUC while the perpetrator is present, and proportionality between dose and AUC holds, the dose calculation is \(Dose_P=Dose_C/R=100/2.00=50.0\) mg. The predicted AUC at 50.0 mg under the perpetrator condition is \(1{,}000(50.0/100)=500\) mg·h/L.
This is a conditional calculation, not a stand-alone clinical recommendation. If the perpetrator raises bioavailability instead of reducing clearance, the same oral ratio still predicts the dose ratio under dose proportionality, but the clearance interpretation is not justified. The calculation also does not establish that matching the control AUC is the appropriate therapeutic target.
Using Additional Route Information
When intravenous and oral comparisons are available under suitable conditions, their ratios can help distinguish a clearance contribution from a change in bioavailability. For the same victim drug, dose-normalized comparisons and compatible conditions are needed. The intravenous ratio estimates \(CL_C/CL_P\) under the linear model; dividing the oral ratio by the intravenous ratio then estimates \(F_P/F_C\). This decomposition remains model-dependent and does not, by itself, identify a specific enzyme or transport mechanism.
Worked Example: Combining Oral and Intravenous Ratios
Suppose a hypothetical study estimates an oral AUC ratio of 1.80 and an intravenous AUC ratio of 1.50 for the perpetrator versus control condition. With linear pharmacokinetics and comparable conditions, the intravenous result gives \(CL_C/CL_P=1.50\), so \(CL_P/CL_C=1/1.50=0.667\). Clearance is estimated to be about 66.7% of control clearance.
The oral ratio satisfies \(R_{\mathrm{oral}}=(F_P/F_C)(CL_C/CL_P)\). Therefore \(F_P/F_C=R_{\mathrm{oral}}/R_{\mathrm{IV}}=1.80/1.50=1.20\), an estimated 20.0% increase in bioavailability. As a consistency check, \(1.20\times1.50=1.80\), the observed oral ratio.
If the aim were to match the control oral AUC, the dose ratio under dose proportionality would be \(1/1.80=0.5556\). A 100 mg control dose would correspond to \(100/1.80=55.6\) mg with the perpetrator in this simplified scenario. The estimate does not imply that such a dose is safe, feasible, or clinically optimal.
Justification for the Log-Scale Interval
A \(t\)-interval is formed for the mean log ratio because the paired participants, not the individual condition measurements considered separately, are the units contributing independent comparisons. If the population log ratios have mean \(\mu_Y\) and variance \(\sigma_Y^2\), the sample standard deviation \(s_Y\) estimates their spread, and the standard error of \(\bar{Y}\) is \(s_Y/\sqrt{n}\). Under independent, approximately normal log ratios, the standardized quantity follows a \(t\) distribution with \(n-1\) degrees of freedom.
The interval is for \(\exp(\mu_Y)\), the population geometric mean ratio, because exponentiation is increasing and transforms the log-scale limits back to the ratio scale. With small samples, strong skewness or outliers in the log ratios can make the normal approximation unreliable. Investigators should inspect paired log ratios and their distribution, consider influential participants and data-quality issues, and report the sample size and uncertainty rather than relying only on a point estimate.
Diagnostics, Alternatives, and Limitations
Before interpreting the estimate, verify that each participant’s condition-specific AUCs are paired correctly and that AUC definitions, doses, and units are consistent. Examine participant-level ratios or log ratios for extreme values, missing pairs, and patterns by study period or sequence. A large ratio may reflect a real interaction, but it can also result from a concentration-time profile with inadequate sampling, an erroneous record, or a problematic terminal-phase estimate. Data review should investigate such observations rather than remove them solely because they are inconvenient.
The paired log-ratio analysis is appropriate when the scientific estimand is a multiplicative exposure effect and the distributional assumptions are reasonable. An analysis of raw arithmetic differences answers a different question and can be dominated by participants with larger AUCs. A parallel-group comparison may be necessary when participants cannot receive both conditions, but it does not remove stable between-person variability through within-person pairing. Population pharmacokinetic modeling can use concentration-time data and represent structural or covariate features, while the paired AUC comparison summarizes an exposure contrast; the methods answer related but not identical questions.
Even a precise AUC ratio has limits. It does not alone describe changes in peak concentration, timing of peak, active metabolites, pharmacodynamic response, or safety. A dose adjustment based on inverse proportionality is only a scenario for matching AUC when exposure is linear in dose, the interaction is stable, and the target is appropriate. Nonlinear pharmacokinetics, dose-dependent bioavailability, changing perpetrator exposure, or clinical constraints can invalidate a simple proportional adjustment. A measured interaction effect should be interpreted in the context of the victim drug’s therapeutic range and relevant clinical evidence.
Check Your Understanding
Answer each question, then compare with the concise solutions below.
- A paired study estimates a geometric mean victim-drug AUC ratio of 0.75 for perpetrator versus control. What direction and approximate magnitude of exposure change does this represent?
- At the same oral dose, AUC is 600 mg·h/L in control and 900 mg·h/L with a perpetrator. Under dose proportionality, what dose would match the control AUC if the control dose were 120 mg?
- Why is a clearance ratio not identifiable from an oral AUC ratio alone when bioavailability may change?
- In a paired log-ratio analysis, what observations are treated as the independent units: the two AUC measurements per participant or the participant-level paired log ratios?
Solutions. (1) Exposure is estimated to be 25.0% lower with the perpetrator, because \(1-0.75=0.25\). (2) The exposure ratio is \(900/600=1.50\), so the dose is \(120/1.50=80.0\) mg. (3) For oral dosing, the ratio combines \(F_P/F_C\) and \(CL_C/CL_P\); without additional information or assumptions, those factors cannot be separated. (4) The participant-level paired log ratios are the independent units, assuming participants are independent; the two condition measurements within a participant form one pair.