Within-person studies – what, when and how?

What are within-person studies? In randomized controlled trials (RCT) the unit of randomization is usually the individual; however, there are designs where a larger or smaller unit is used. For example, in cluster randomized trials, general practices rather than patients may be randomly allocated to receive different interventions such as a new education package for preventing hand eczema in nurses.1 If individuals are used in such studies ‘contamination’ could occur. For example, nurses in the control group may access the education package and adopt it themselves. In contrast, smaller units such as lesions or limbs may be randomized so that participants receive two or more treatments on different sites of their body. The latter is a type of self-controlled study where patients act concurrently or sequentially as their own control. These studies are called ‘within-person’ randomized trials.

In this issue of the BJD, Pandis and colleagues report the extension of the Consolidated Standards of Reporting Trials (CONSORT) 2010 statement for transparent reporting of within-person studies.2 In order to understand the need for the CONSORT extension, it is important to revisit the rationale and limitations of within-person designs.

When to use or avoid within-person studies? There are several advantages of a within-person study over the conventional parallel group RCT, the main one being efficiency. Because patients vary so much in their initial disease state and in their response to therapy, substantial numbers of patients on each treatment are usually needed in a parallel group RCT in order to reliably estimate the magnitude of any treatment effect. However, the nature of within-person studies reduces variability between the populations and such studies are therefore useful for rare conditions or where there is a high degree of patient-to-patient variability, as is often the case in dermatology.3, 4 Within-person designs can also be favoured ethically as it gives every participant the opportunity to receive treatment on one part of the body rather than leaving some untreated.3

Despite the benefits, there are design issues specifically related to within-person trials that need to be considered. Firstly, important outcomes related to the whole person such as quality of life cannot be assessed. Secondly, lesions such as those that occur in psoriasis need to be similar on each limb in terms of induration and size, and they should be located on similar body sites for example, one elbow compared with the other elbow rather than elbow and knee. Such tight entry criteria may affect the generalisability of the results and limit the clinical implications of the study.4 As such, the CONSORT extension requires a rationale for using a within-person design and for the identification of body sites.

Furthermore, because of the high degree of collaboration that is often required from the patients, such as impractical treatment modalities, there could be a high dropout rate. Someone dropping out has double the impact in within-person studies compared with other study designs because each patient contributes to both the treatment and control groups.4 Also, as patients have access to both treatments, there is greater risk of contamination (particularly if one treatment is working better than the other). Another potential disadvantage is carry-across effects that occur when the treatment performed in one part of the body affects the treatment responses in other parts of the site for example systemic absorption from a very potent topical steroid. Therefore, for certain interventions, it may be better to consider a sequential rather than concurrent timing of treatments.5 The CONSORT extension requests that the timing of the intervention is reported.

How to analyse within-person studies. The key feature to consider when analysing within-person trials is that the clinical responses within an individual are correlated, and therefore no longer independent values. This correlation must be taken into account in the analysis. As a result, the CONSORT extension requires authors specifically to report appropriate statistical methods for within-person studies. For example, for a continuous outcome such as disease severity score like the Psoriasis Area and Severity Index, generalized linear mixed modelling should be used rather than linear regression.6 Methods have also been proposed to analyse correlated binary outcomes and survival outcomes.6-9 Analyses that ignore the correlations will overestimate the variability, thus artificially increasing P-values and decreasing the chances of observing a significant effect (decreasing the statistical power and increasing the type II error rate).6

Within-person studies have been quite common for topical therapies in dermatology and may continue to be so.4 The CONSORT within-person extension2 will improve the clarity of how teams that report such trials have dealt with the traditional potential pitfalls, as described above, and will lead to better interpretation of the clinical relevance of these studies.

Conflicts of interest

None to declare.