Crossover Trial

A trial in which every participant is randomised to a sequence of treatments rather than to a single arm: they receive treatment A then B (or B then A), separated by a washout period. Each person serves as their own control, so the comparison is within subjects.

Two treatment sequences with a washout between periods, ending in a within-subject comparison. Because each patient is measured on both treatments, every person’s stable characteristics cancel out of the comparison.

Check three things: is the condition stable across the whole trial (so it doesn’t simply improve over time)? Is the washout long enough to clear the first treatment? Was the sequence randomised and ideally blinded? Without these the period-2 result is contaminated.

Using each patient as their own control strips out between-person variability, so a crossover needs far fewer participants for the same power — efficient and elegant. But it only works for stable, chronic, non-curable conditions; it is useless for acute or curable disease.

• Each patient receives both treatments (own control)
• Needs a washout between periods
• Only for stable, chronic, non-curable conditions

Chronic stable conditions — crossover designs are common in chronic pain (e.g. comparing two analgesics in stable neuropathic pain) and stable asthma (comparing two inhalers, each given for a fixed period with a washout). Each patient tries both, halving the sample size needed. You would never run a crossover in acute trauma or sepsis: the illness is changing fast (a period effect) and there is no way to “wash out” a resolved or fatal event — a parallel-group RCT is required.

• Carryover effects — an inadequate washout lets treatment 1 contaminate period 2.
• Period effects — the underlying condition or environment changes between periods, mimicking a treatment difference.
• Dropouts mid-sequence — a patient lost after period 1 loses the within-person comparison entirely.