Crossover trials

In a crossover trial, participants are randomized to a sequence of treatments. We shall consider the simplest design, in which there are two treatments. As an example, a trial was conducted in patients with asthma to compare laser acupuncture with a sham ('placebo') procedure. Patients were randomized either to acupuncture for five weeks (first period) then sham for five weeks (second period), or sham then acupuncture. One of the outcomes was a symptom score.

This gives us data for each patient both when they were on acupuncture, and when they were on sham treatment. We can compare these for each patient to assess the effect of acupuncture within each patient. This is a very efficient approach to analysis, because when making the comparison between treatment and control we do not have to allow for all the variation that occurs between patients, which we have to deal with in a parallel group trial.

In practice this means that, for the same number of participants, a crossover design is likely to be more powerful. However, crossover trials are not always appropriate. The most important consideration is whether the patients start the second period in a similar state to how they started the first period. If the characteristics of the participant have changed in some way by the time the second period starts, then the comparison of treatments is not fair, and there will be within patient variation. Some questions you should ask yourself are as follows.

• Is the condition of the patients chronic and stable? Crossover trials are common for conditions such as asthma, osteoarthritis and epilepsy. Crossover trials may not be appropriate for progressive diseases or acute conditions that will worsen or improve by the second period. If patients vary from one period to the other there is said to be a 'period effect'.
• Does the intervention provide temporary relief, and not permanent change? For example, surgical interventions are unsuitable for crossover trials if the surgery permanently alters the condition.
• Can the outcome be repeated in the second period if it occurs in the first? For example, crossover trials are certainly unsuitable when the primary outcome is mortality, or pregnancy in infertility studies.
• Might the effect of the first intervention last into the second treatment period? In the acupuncture trial, a three-week 'washout' period was built into the trial between the two treatment periods. This is a common method to minimize 'carry-over' effects and ensure the participant is in the same state at the beginning of each period, though it is not always sufficient.
• Does the trial go on long enough for drugs to have effects and outcomes to occur? For example, a trial in epilepsy with the outcome of number of fits, needs to observe the patients for long enough to make sure that we haven't, by chance, just picked a particularly good or bad time in their illness.

If you have crossover trials in your review, you will need to decide on the following points:

• Should I include crossover trials in the review?
• Should I include crossover trials in any meta-analyses?
• Should I combine crossover trials with other types of trials?
• How should I include crossover trials in a meta-analysis?

The brief answer to the first three questions is

• There is no reason to exclude crossover trials solely because they are crossover trials.

Of course, there may be other reasons why crossover trials might be excluded, for example, if they have treatment periods that are too short or do not have an appropriate wash out period. If you are anticipating cross over studies in your review you will need to set these inclusion criteria in your protocol.