Cross-over trials are suitable for evaluating interventions with a temporary effect in the treatment of stable, chronic conditions. They are employed, for example, in the study of interventions to relieve asthma and epilepsy. They are not appropriate when an intervention can have a lasting effect that compromises entry to subsequent periods of the trial, or when a disease has a rapid evolution. The advantages of cross-over trials must be weighed against their disadvantages. The principal problem associated with cross-over trials is that of carry-over (a type of period-by-intervention interaction). Carry-over is the situation in which the effects of an intervention given in one period persist into a subsequent period, thus interfering with the effects of a different subsequent intervention. Many cross-over trials include a period between interventions known as a washout period as a means of reducing carry-over. If a primary outcome is irreversible (for example mortality, or pregnancy in a subfertility study) then a cross-over study is generally considered to be inappropriate. Another problem with cross-over trials is the risk of drop-out due to their longer duration compared with comparable parallel group trials. The analysis techniques for cross-over trials with missing observations are limited. The assessment of the risk of bias in cross-over trials is discussed in Section 16.4.3.
In considering the inclusion of cross-over trials in meta-analysis, authors should first address the question of whether a cross-over trial is a suitable method for the condition and intervention in question. For example, although they are frequently employed in the field, one group of authors decided cross-over trials were inappropriate for studies in Alzheimer’s disease due to the degenerative nature of the condition, and included only data from the first period (Qizilbash 1998). The second question to be addressed is whether there is a likelihood of serious carry-over, which relies largely on judgement since the statistical techniques to demonstrate carry-over are far from satisfactory. The nature of the interventions and the length of any washout period are important considerations.
It is only justifiable to exclude cross-over trials from a systematic review if the design is inappropriate to the clinical context. Very often, however, it is difficult or impossible to extract suitable data from a cross-over trial. In Section 16.4.5 we outline some considerations and suggestions for including cross-over trials in a meta-analysis. First we discuss how the ‘Risk of bias’ tool described in Chapter 8 can be extended to address questions specific to cross-over trials.