# A crossover trial needs more than two different periods

Draft • Research checked 4 October 2026 • AI-assisted DomDNA editorial content. No independent clinical review or publication is claimed. General education, not individual medical advice.

Trying one option and then another can feel like a fair comparison because the same person experiences both. In research, that is the attraction of a crossover design. It is also the reason the order and timing need careful attention.

The first period can leave something behind. Skills learned, an enduring effect, a changed routine or a shift in circumstances can influence what happens in the second period. A later result may therefore reflect more than the option used at that moment.

## The same person is not the same situation

[Cochrane's chapter on other study designs](https://www.cochrane.org/authors/handbooks-and-manuals/handbook/current/chapter-23) discusses the conditions that make crossover comparisons interpretable, including possible carryover and period effects. Within-person comparison can reduce some between-person differences, but it does not remove time.

A period effect is a difference associated with when an observation happens. Carryover concerns an earlier intervention affecting a later period. Both can complicate a story that treats each period as a clean fresh start.

A washout interval is intended to reduce lingering effects where that is appropriate. Its existence is not proof that everything returned to an unaffected baseline. The interval needs a rationale suited to the intervention and outcome, rather than a generic waiting rule.

## Consider an invented learning task

Suppose a hypothetical study compares two tutorial formats. Each participant uses one format in week one and the other in week two. Everyone learns the same material, and the final task tests that material.

If participants remember what they learned in the first week, the second format receives people who already know something. Swapping the order across participants may help expose an order problem, but it does not make acquired knowledge disappear.

Now change the research question. Suppose each format uses different, suitably comparable material, with careful attention to task difficulty and sequence. The comparison may become more interpretable, though it still needs an analysis that respects repeated observations from the same person.

These are editorial examples, not recommended protocols or descriptions of a published study. They show why the suitability of crossover depends on what can persist between periods.

## Read the sequence, not just the averages

The [randomised crossover reporting extension](https://pubmed.ncbi.nlm.nih.gov/31366597/), published in 2019, addresses design-specific reporting. A useful paper describes the sequence of interventions, the periods and how within-person comparisons were analysed. This extension predates the updated 2025 base CONSORT statement, so it should not be mistaken for that newer general checklist.

If a summary shows only an average for option A and an average for option B, look for the missing temporal structure. Were both orders used? Did participants contribute both measurements? Were results analysed as paired observations? What happened when someone left after the first period?

You do not have to solve the analysis yourself. The purpose is to recognise information that an ordinary two-column chart might omit.

## An N-of-1 trial is a planned comparison

The [CENT explanation for N-of-1 trials](https://doi.org/10.1136/bmj.h1793) concerns prospectively planned repeated crossover research in an individual. It is different from a person trying something during a difficult week, noticing improvement and declaring the experiment successful.

A planned sequence and defined outcomes can make an individual comparison more informative. They still cannot guarantee that context stayed stable, that lingering effects disappeared or that the observation generalises to other people.

This article is not a suggestion to withdraw prescribed care, cycle medicines or create washout periods independently. Clinical N-of-1 work can involve decisions that are entirely outside a casual tracking exercise.

## A useful everyday application

For a low-risk observation such as comparing two note-taking formats, sketch the sequence before looking at the results. Record what could improve simply through practice, what might carry into the next period and whether the task changed.

That small preparation can prevent a misleading conclusion such as “the second format is better” when the second attempt benefited from familiarity. You may decide that the comparison cannot separate the effects cleanly. That is a useful outcome too.

When reading health research, use the same distinction. “Each participant received both options” describes an appealing design feature. It does not settle whether the intervention's effects were reversible, the periods were comparable or the analysis used the design properly.

The sequence is part of the evidence. Keep it attached to the result rather than treating time as an empty space between measurements.

For an educational starting point, the [DomDNA lifestyle quiz](https://domdna.com/quiz) records lifestyle answers, not DNA analysis or a clinical diagnosis.