- Published
- Updated
In a crossover study, the same person can receive different interventions in successive periods. If an earlier effect persists, the second measurement may not reflect only the intervention given in that period. Assess this possibility clinically before selecting the design rather than relying on later statistical repair.
Define the question and assumptions
Consider effect duration, condition stability and reversibility together. Do not choose washout solely for scheduling convenience. Earlier experience may also change behaviour, learning or measurement. Suitability requires more than assessing pharmacological persistence; a later statistical test should not be assumed to settle a design problem that was foreseeable before conduct.
Hypothetical study example
In a hypothetical study, an intervention may create lasting behavioural change. A second-period comparison independent of the first intervention could be implausible. Instead of shortening the interval to simplify visits, the team examines whether an alternative such as a parallel design better addresses the intended scientific question.
Design review checks
- Review clinical evidence about effects persisting from an earlier study period.
- Document the rationale and practicality of the proposed washout interval.
- Consider learning, disease changes and interventions occurring outside assigned periods.
Record the decision and its limits
An accepted rationale explains the assumptions supporting crossover. Where carryover is plausible, adding a model term alone is not presented as a sufficient solution. Plan missing periods and discontinuation separately. Keep period data and limitations visible, without claiming that using each person as their own control eliminates every potential bias.
Frequently asked questions
- Is a design always better when each person is their own control?
- No. Stability and carryover assumptions remain important.
- Does washout definitely remove carryover?
- Not always; lasting and behavioural effects also need consideration.