In a crossover trial, a participant may receive one intervention during one period and another during a later period. The comparison is organised differently from a trial where each person remains in one intervention group throughout.
The Cochrane Handbook’s chapter on trial variants explains crossover designs and concerns such as carryover. An earlier intervention can affect a later period, complicating the comparison.
Draw a fictional two-period design with one sequence receiving A then B and another receiving B then A. The order matters, and appropriate randomisation can assign participants to sequences. Simply giving everyone A first would leave order entangled with intervention.
Look for the length of each period and any planned washout interval. Do not assume that a washout completely removes earlier effects just because the protocol includes one. Its suitability depends on the intervention and condition being studied.
The participant’s condition may also change over time. Some research questions are unsuitable for a crossover design, particularly where effects are lasting or the underlying state cannot reasonably support the planned comparison.
When reading the results, check that the analysis recognises repeated measurements from the same people. Two observations from one participant are not automatically equivalent to observations from two unrelated participants.
A clear summary names the sequences, periods, outcome and main limitations. It avoids turning “participants acted as their own controls” into a promise that all bias has disappeared. Understanding the timeline makes the design easier to assess while leaving clinical conclusions to the full body of evidence.
The accompanying visual is an editorial illustration.

