The dog-leg: an alternative to a cross-over design for pragmatic clinical trials in relatively stable populations

Background: A cross-over trial design is more powerful than a parallel groups design, but requires that treatment effects do not carry over from one period of the trial to the next. We focus here on interventions in chronic disease populations where the control is routine care: in such cases we cannot assume the intervention effect is easily washed out in crossing over from the experimental intervention back to the control. Methods: We introduce an alternative trial design for these situations, and investigate its performance. One group is assessed before and after the experimental intervention, whereas two other groups provide respective, independent treatment comparisons in each period. We call this a dog-leg design because of the pattern of assessments in the three groups. The dog-leg design is reminiscent of a stepped wedge design, but with a reduced schedule of assessments and with the notable difference that not all groups receive the intervention. Results: If the correlation between baseline and follow-up is < 0.72, the dog-leg design is more efficient than a parallel groups design with a baseline assessment. The dog-leg design also requires fewer assessments in total than a parallel groups design where participants are only assessed once, at follow-up. Conclusions: The dog-leg design is simple, and has some attractive properties. Though there is a risk of differential attrition in the three arms, the design's good performance relative to alternatives makes it a useful addition to the methodologist's toolkit.