Washout Periods and Carryover Control

Why Washout Requires More Than Counting Half-Lives

In a crossover drug-drug interaction (DDI) study, each participant receives treatments in more than one period. The interval between periods is intended to reduce the influence of the preceding treatment on the next one. That influence can persist in several ways: the perpetrator drug may remain in the body, an active metabolite may continue to act, the victim drug may remain from the prior period, or an enzyme or pharmacologic response may take time to recover after the drug concentration has fallen.

A washout duration should therefore be tied to a specified tolerance for residual exposure or effect, not chosen by a universal rule such as “five half-lives.” Five half-lives leaves about 3.1% of a starting concentration under a simple one-compartment exponential-decay model. That may be negligible for one drug and still relevant for another, especially when a low concentration can substantially inhibit or induce a pathway.

For a DDI study, assess both the victim and perpetrator drugs, their relevant active metabolites, and persistent changes in the affected biological system. The calculations below are planning tools. Their assumptions—especially exponential decay, the choice of terminal half-life, and the relationship between concentration and effect—must be checked against the drug and study context.

Concentration-Based Washout

Suppose that after the last dose a drug concentration declines according to a single exponential with terminal half-life \(t_{1/2}\). If \(C_0\) is the concentration at the start of the washout, the concentration after elapsed time \(t\) is \(C(t)=C_0e^{-kt}\), where \(k=\ln(2)/t_{1/2}\). This approximation assumes no further drug input, a stable terminal elimination rate, and a relevant concentration that follows the measured or modeled terminal phase. It can be unreliable with multiphasic disposition, prolonged absorption, nonlinear elimination, or an active metabolite that disappears more slowly.

Choose a concentration threshold \(C_{\mathrm{tol}}\) that is justified for the study question—for example, a concentration below which a perpetrator is not expected to produce a material interaction, or a victim-drug residual small enough not to distort the next period. Solving \(C(t)\leq C_{\mathrm{tol}}\) gives a minimum elapsed time. If \(C_0\leq C_{\mathrm{tol}}\), the concentration criterion is already met at the start; otherwise the logarithmic expression gives the required time.

Recovery of Enzyme Activity and Other Persistent Effects

A perpetrator can alter enzyme activity or abundance in ways that outlast its measurable plasma concentration. When the perpetrator is no longer driving the change, a common first-order recovery approximation describes normalized activity \(E(t)\), relative to baseline activity of 1, as returning toward baseline at rate constant \(k_{\mathrm{deg}}>0\). This model assumes that recovery is governed by a single turnover process and that no continuing inhibitor, inducer, or other driver remains.

The equation describes recovery from either reduced or increased activity: \(E_0<1\) represents activity below baseline, while \(E_0>1\) represents activity above baseline. To require activity to be within an absolute tolerance \(\delta>0\) of baseline, use the absolute deviation. If activity is already within tolerance, no additional recovery time is required by this criterion. Otherwise, solve for the time at which the deviation falls to \(\delta\).

The tolerance \(\delta\) is an absolute difference on the normalized activity scale. A tolerance of 0.10 means activity from 0.90 to 1.10 is acceptable. If the activity-to-exposure relation is nonlinear, being within that activity band does not necessarily mean the victim-drug exposure is within an acceptable band; the pharmacologic consequence should be considered as well.

Other persistent effects can be handled in the same conceptual way if a defensible recovery model and a meaningful tolerance are available. For example, residual pharmacologic effect might be modeled as a declining effect-compartment concentration or as a first-order return of a biomarker toward baseline. The model must represent the mechanism relevant to the next period; a plasma concentration alone may not capture a slowly resolving downstream response.

Combining Requirements and Checking Carryover

When several independent criteria must each be met—such as low perpetrator concentration, low victim-drug concentration, and enzyme activity within tolerance—the planned washout must satisfy the slowest requirement. If the relevant residual effects add together or interact, however, checking each against a separate threshold may not control their combined impact. In that case, specify a joint acceptable limit or evaluate a conservative bound on the combined effect.

Carryover can be investigated with measurements and study patterns. Useful checks include predose concentrations of the prior victim or perpetrator drug, active metabolites, and biomarkers of pathway activity. In crossover data, inspect outcomes by period and treatment sequence, and consider whether the period-two outcome is consistent with a residual effect from the period-one treatment. Such patterns are diagnostic clues, not automatic proof of a causal carryover effect: sequence, period, and treatment can be difficult to disentangle, and formal carryover tests may have low power.

Design Choices, Diagnostics, and Limitations

A pharmacokinetic washout calculation is often more informative than applying a fixed number of half-lives because it connects duration to an explicit residual concentration threshold. A biological recovery calculation is needed when enzyme turnover or another persistent effect governs recovery. A parallel-group design avoids within-participant carryover between treatment periods, but it does not provide the same within-participant comparison and may require more participants to achieve comparable precision. A fixed-sequence design also cannot make an inadequate interval harmless: residual treatment effects can still influence later observations.

The main diagnostics are whether observed predose concentrations, biomarkers, and period-specific outcomes agree with the planned assumptions. A systematically elevated predose concentration after one treatment suggests the pharmacokinetic washout criterion was not met. A biomarker that remains displaced from baseline suggests a biological recovery criterion may not have been met. A period-two pattern that varies with the preceding treatment raises concern for carryover, but should be interpreted alongside concentration and mechanistic evidence rather than treated as a standalone test.

• Half-life mismatch: Using a short distribution or apparent half-life when a slower terminal phase or active metabolite governs persistence can underestimate the required time.
• Threshold without rationale: “Negligible” concentration is not a universal value. It should be justified by expected pharmacology, assay capability, and the study objective.
• Ignoring recovery above baseline: The absolute deviation \(|1-E_0|\) is needed for activity that may be either induced or suppressed. If the starting deviation is already at most \(\delta\), the criterion is already met.
• Treating a nonsignificant test as proof: Failure to detect carryover can reflect limited precision or low power. Evidence of absence requires sufficiently informative measurements and a justified acceptable-effect bound.

These calculations remain approximations. Between-participant differences in clearance, enzyme turnover, dose history, and adherence can make a single nominal interval insufficient for some participants. Where feasible, predose measurements can check whether the intended conditions were achieved. If the study relies on a model, sensitivity analyses over plausible half-lives, starting concentrations, and tolerances help show how robust the chosen duration is.