An Introduction to the Cross-Over Trial Design

Cross-over clinical trials are an essential research design used in many scientific areas of inquiry, such as medicine, nursing, and psychology. This type of trial is designed to compare two or more treatments by having the same participants receive treatments or interventions in a pre-determined sequence (cross-over to another treatment or intervention). Through this method, investigators can assess whether any differences between treatments are due to experimental manipulation or other factors like participant characteristics. This paper builds upon the other manuscripts in this series, introduces the basics of a cross-over clinical trial, and discusses the strengths, weaknesses, and methods to avoid potential pitfalls.

What is a cross-over clinical trial?

Similar to the randomized controlled trial (RCT), the cross-over design is also randomized and can have a placebo-control group and be double-blinded (investigator and participant unaware of treatment assignment) (Please see the Efficacy Randomized Controlled Trial article in this series (Capili & Anastasi, 2023). The primary difference between the RCT and the cross-over design is that participants are randomized into groups who receive multiple treatments or interventions in a pre-determined order.

Participants also complete follow-up visits per study protocol, where investigators monitor their progress using pre-determined outcome measures, and the effects of the treatment or intervention are measured. Outcomes are usually measured during and at the end of each treatment cycle (please see Figure 1), and an untreated wash-out period is implemented to minimize the carry-over effect of a treatment (the intervention’s lingering impact on the outcome after the treatment or intervention ends) (Gallin, 2018; Hulley, 2013). Additionally, the wash-out period between treatment or intervention allow the outcome variable to return to baseline before starting a subsequent treatment or intervention, and any differences due to the experimental manipulation are noted.

Figure 1.

Two Group (AB/BA) Cross-Over Design

Cross-over designs are recommended for studies on symptom reduction, specifically for chronic conditions with short-lived and reversible treatment effects. In other words, the experimental treatments or interventions are not curative. However, the design is unsuitable for acute conditions (i.e., pain management for a sprained ankle) where the outcomes may change or resolve regardless of interventions (Sedgwick, 2014). The duration of each cross-over trial varies, depending on the purpose of the study but may last anywhere from days to months.

Another resource to support the understanding of the cross-over design is the Consolidated Standards of Reporting Trials (CONSORT) Extension for Cross-over Designs (Dwan, Li, Altman, & Elbourne, 2019). The cross-over extension is a set of guidelines to improve the reporting quality of cross-over trials. The CONSORT statement for cross-over trials is a 25-item checklist that guides the writing and reporting of cross-over trials in publications. It also includes a flow diagram detailing the recruitment process, providing information on how many participants were enrolled, how many participants completed the study, and why they dropped out. The flow diagram provides transparency regarding the conduct of the study. Therefore, by following the CONSORT Statement for cross-over trials, investigators can ensure transparent and comprehensive reporting of their research, facilitating a better understanding and interpretation of the trial results. See Table 1 for a summary of the CONSORT Extension for Cross-Over Trials.

Table 1.

Key Elements of the CONSORT-Cross Over Extension

Title and Abstract: Specify that the study is a cross-over trial in both the title and abstract, providing an informative summary of the trial design, interventions, primary outcomes, and results. Background: Describe the rationale for cross-over design, including its advantages and potential limitations. Participants: Provide details on participant eligibility criteria, settings, and locations where the trial occurred. Interventions: Describe each intervention administered during the trial, including their sequence and wash-out periods between interventions Outcomes: Clearly define the primary and secondary outcome measures and specify the methods for assessing and analyzing these outcomes. Sample Size: Explain the sample size calculation, considering the cross-over design and accounting for potential drop-outs or incomplete data. Randomization: Describe the method used to randomize participants to different intervention sequences and any steps to ensure concealment of allocation. Statistical Methods: Detail the statistical methods used for analyzing the data, including any specific techniques for handling data from cross-over trials, such as within-subject comparisons or adjustments for a period or carry-over effects. Results: Report the flow of participants throughout the trial, including enrollment, allocation, follow-up, and analysis. Present the results for each intervention and any observed differences accounting for the cross-over design. Discussion: Interpret the trial findings in the context of the cross-over design, addressing any potential limitations, biases, or implications for future research and clinical practice.

Cross-over trial example

For example, if an investigator wants to compare the efficacy of treatment A (valerian root) with treatment B (black cohosh) for hot flushes, a cross-over design could be considered to minimize the risk of confounding. For this study, the investigator will use 140 women randomly divided into two groups, with 70 participants in each group. Group 1 will receive treatment A for four weeks to manage their hot flushes, while Group 2 will receive treatment B for four weeks. Then, there will be a six-week wash-out period where neither treatment A nor B will be administered. Afterward, Group 1 will receive treatment B (black cohosh), while Group 2 will receive treatment A (valerian root) for four weeks. Twice during the AB/BA type cross-over trial, the investigator measures the primary outcome of hot flush reduction. The first measurement is taken after Group 1 receives treatment A, and the second is after Group 1 receives treatment B. It is important to note that outcome measurements can occur more often during an AB/BA cross-over trial (i.e., during the two-week -period and then at the end of the four weeks for the hot flush reduction example).

The AB/BA model is the basic form of the cross-over trial where one group of participants initially receives treatment A, followed by a wash-out period and then treatment B. Then, the same sequence is observed for the second group, but in reverse order. This model also has variations such as ABC, CAB, and BCA regimens. Please see Figure 2.

Figure 2.

Three Group (ABC/CAB/BCA) Cross-Over Design

Strengths of the cross-over trial

A benefit of using the cross-over design is that each participant acts as their control, decreasing individual variability. In statistics, the term is called within-subject. Within-subject is described as investigators testing the same participants completing multiple treatments or interventions at different time points. The within-subject statistics evaluate how outcomes change or stay the same over time as the result of the treatment or intervention in the same participant. The within-subject statistics increase the study’s power (the probability of correctly deciding to reject the null hypothesis when it is false). Please see the article in this series, the Archectiture of a Research Study (Keeler & Curtis, 2023), because the participants act as their controls (Piantadosi, 2017).

For instance, women who consume more caffeine daily may experience more hot flushes than those who avoid caffeinated foods and beverages. Hence the cross-over design reduces the diet variability among the women participating in the hot flush study since the outcome measures are compared using each participant’s outcome data and not between-subjects (or between-groups (Group 1 versus Group 2)). Moreover, the design enables the detection of smaller effect sizes with fewer participants needed (Gallin, 2018). Since each participant receives both treatments or interventions, the study only needs half as many participants as the traditional RCT, where participants only receive one treatment or intervention (Piantadosi, 2017). The effect size in clinical research refers to the quantitative measure of the magnitude or strength of the relationship between variables or the extent of an observed effect in a study (i.e., a 30 percent difference of hot flushes per day in the two treatment groups (valerian versus black cohosh) (Hulley, 2013). Effect size is used to evaluate the practical significance of a treatment or an intervention and provides valuable information about the meaningfulness of the research findings. Effects sizes are often reported alongside statistical significance (p-value) to assist investigators and clinicians in better understanding the real-world implications of the results.

Weaknesses of the cross-over trial

Cross-over designs may not be ideal if the treatment studied has a lasting effect on the outcome. In the above example, it can be challenging to distinguish treatment B’s (black cohosh) impact from treatment A’s (valerian root) if treatment A has a sustained effect even after the treatment has ended. In such cases, using a cross-over design could present challenges.

Furthermore, carry-over effects can be challenging to detect or quantify (Gallin, 2018). In certain situations, cross-over designs may be considered unethical. For instance, switching participants from the effective treatment group to the other group may be unethical if one treatment or intervention is ineffective and the condition is a severe health risk (Gallin, 2018). An example of a condition in which the cross-over trial could be considered unethical or inappropriate is type 2 diabetes. For participants requiring daily use of a pharmaceutical agent to control their blood sugar, it would be unethical to withhold diabetic agents during the wash-out period, wherein participants do not take any medications to control their blood sugar.

Another limitation of the design is that participants enrolled in cross-over studies must be in the study longer. Cross-over trials may not appeal to participants as the study design involves using multiple treatments or interventions instead of just one, resulting in an extended study period and burden for participants (Gallin, 2018). The longer duration for study participation may cause extra difficulty for participants and increase the drop-out rate.

Avoid potential pitfalls

Cross-over designs can be more effective than other research designs if used with the appropriate patient population and treatment or intervention since the design reduces the confounding variables in non-treatment-related factors (i.e., diet, medical history, medication history). Cross-over designs should be considered when studying conditions that involve relapsing/remitting or episodic symptoms, especially if the cycles are short and somewhat predictable (Gallin, 2018). Examples of such conditions include migraines and hot flushes. However, suppose the condition is unstable or progressive; in that case, the cross-over design may not be appropriate as it introduces extra variability due to changes in the condition or disease during the study (Gallin, 2018). The investigator should select a research design that is more suitable for the condition or disease under investigation.

For the type 2 diabetes example presented previously, a method to consider for this condition and using a cross-over design is to permit participants to use their baseline diabetic medications during the entire study, inclusive of the wash-out period (Wang et al., 2016). A caveat to using this approach is to ensure that participants do not get hypoglycemic and their baseline medications will not interact with the study medications.

The longer duration of study participation is another potential issue with this study design. As discussed previously, participant recruitment could be hampered by individuals realizing that the study requires extensive time, which can affect daily schedules and event planning, such as vacations. In addition, some participants may not want to switch study treatments or interventions during one study. However, if the available treatments or interventions in the cross-over study are perceived by participants as two similarly effective treatments or interventions, especially for conditions with predictable or stable clinical traits. Recruitment may not be hindered (Piantadosi, 2017).

The potential for carry-over effects is another concern for this design. Investigators could consider a dose-de-escalating option. This method further reduces the possibility of the carry-over effects of the previous treatment or intervention. Dose-de-escalating entails gradually decreasing the dose or interaction if the intervention is non-pharmacologic in the first treatment, followed by no treatments or interventions during the wash-out phase (Wang et al., 2016). Then once the wash-out phase is complete, participants switch to the subsequent treatment or intervention in the sequence. Before the wash-out, the premise of dose-de-escalating is to reduce the potential for a carry-over effect even further.

Summary

In summary, cross-over trials have several benefits, including the ability to compare two or more treatments or interventions and enabling investigators to use fewer participants than in the traditional RCT. The design also allows controlling for individual variability across treatment groups. However, potential carry-over effects must be considered before utilizing this research method. The cross-over design may only be appropriate for some chronic and stable conditions that involve relapsing/remitting or episodic symptoms.