This thesis adds to the methodological literature of lateral wedge shoe insoles (LWIs), a conservative, non-surgical therapy used in the management of patients with symptomatic knee osteoarthritis (OA). It is thought that they reduce loading, and therefore painful symptoms, predominantly in the medial aspect of the knee - the site most affected by those with medial tibiofemoral OA. To date, however, LWIs have shown mixed evidence of efficacy, and the question of whether, and for whom, these insoles work for is unclear. This thesis seeks to explore and address how these trials', and by extension, all OA trials', methodology can be improved. Six publications are included. The first is a systematic review and meta-analysis of LWI clinical trials, suggesting that a placebo control is essential when studying LWIs, due to the substantial placebo effect observed in these studies when a no-treatment comparator is used. When an appropriate *placebo* control is used, there appears to be no overall effect of LWIs when given to a sample with mixed knee OA. The next two publications summarise the SILK trial, a single-visit randomised controlled trial comparing four footwear conditions, including two types of LWIs. The first of these two papers demonstrates an overall beneficial effect of LWIs, but with substantial variation in efficacy between individuals. The second SILK trial paper explores whether the individuals that did have a beneficial effect ('responders'), which made up one third of the study sample, could be differentiated from non-responders using other gait variables. This publication found that ankle-related measurements were associated with responder status. A further two publications demonstrate a novel method to help select which pain outcome should be used as the primary in knee OA trials, and also whether creating composite outcomes of instruments purporting to measure the same underlying construct (pain) increases the ability to detect treatment effects. The method has advantages over existing methods comparing responsiveness: it allows formal statistical testing of outcomes for differences in responsiveness simultaneously, and at multiple time points. The final publication summarises the 'InRespond' trial, a randomised controlled cross-over trial looking at the effect of LWI therapy on knee symptoms. The trial used a stratified medicine approach to recruit treatment-responsive individuals, included an optimally responsive pain outcome (as a secondary outcome), an appropriate placebo control, and a cross-over design, which increases the ability to detect differences with lower sample sizes. The results suggest a larger treatment effect than shown in previous studies, which is statistically significant, and of borderline clinical significance also. Although focused on knee osteoarthritis, the findings from this thesis may be relevant to clinical trials of chronic diseases generally.