*Geek Box: Within-Person Design
In a within-person design, all participants will be exposed to every intervention or treatment. This may often be achieved by a cross-over study, where participants will undergo Intervention A before Intervention B [or B before A, to minimise an effect of treatment order]. In a within-person study, each participant serves as their own control. This is important where there may be significant individual differences which may influence the results, in particular by resulting in measurement errors.
For example, we know that the composition of bacteria in the microbiota of one person differs significantly from person to person; thus, comparing two groups of people against each could result in distorted averages. By comparing the effects of an exposure in one person to another exposure in the same person, this potential error is accounted for, because the changes are relative to that persons baseline. This minimises the potential for individual differences to generate misleading results.
For nutrition studies, there are a number of attractions to within-person designs. First, responses to diet, and dietary intake generally, varies from person to person: comparing each person against themselves may thus provide more insight into the effect of an intervention. Secondly, in between-person studies where participants in one group are compared to another [often a placebo], double the number of participants are required. For example, if a power calculation indicates that for a between-person study 40 participants would be required for the intervention and 40 for the control – 80 participants total – then by having each subject serve as their own control in a within-person design would mean the study could be conducted with 40 people.
The main potential issue with within-person designs is the effect of treatment order; there may be a carryover effect of undergoing treatment A before treatment B. To overcome this, a good design can randomise participants to a treatment order [A, then B vs. B, then A].