Study design and setting
YORVIC (York study of unloading shoes for vascular intermittent claudication) was a single-centre randomised AB/BA crossover trial with a 2-week observational follow-up. Participants were recruited from vascular clinics at York Hospital, and all assessments were conducted at York St John University. Following a screening visit, participants completed two assessment visits up to 2 weeks apart. At each assessment visit, participants completed three standardised walking tests whilst wearing either the unloading shoes or visually-similar control shoes, the order of which was randomly assigned. At the end of the second assessment visit, participants were given either the unloading or control shoes to use in their home environment for 2 weeks, with the instruction to wear them for at least 4 h every day. At the end of this period, participants returned the shoes and a completed survey about them. A sub-sample of participants was also interviewed about their experiences of using the shoes. Participants were able to claim up to £15 per visit towards travel expenses. The study was approved by the NRES Committee for Yorkshire & The Humber - Leeds West (Ref: 15/YH/0107), and prospectively registered (ClinicalTrials.gov: NCT02505503). Written informed consent was obtained from participants prior to enrolment.
Participants
Inclusion criteria were: aged ≥16 years; stable symptoms of intermittent claudication for ≥3 months; resting ankle-brachial index ≤0.9 and/or imaging evidence of PAD; pain-free walking distance <250 m on 6-min walk test with ambulation limited primarily by calf claudication (assessed at screening visit), and; able to read and speak English and provide written informed consent. We excluded people with: absolute contraindications to exercise testing (as defined by the American College of Sports Medicine [7]); critical limb ischemia; lower-limb amputation; co-morbidities that limit walking to a greater extent than intermittent claudication (e.g., severe knee osteoarthritis); ambulation limited by claudication in regions other than the calf; major ankle or foot pathology, and; current or previous (within 6 months) use of shoe inserts, knee or ankle braces or customised shoes prescribed by a health professional.
Interventions
The unloading and control shoes were produced and supplied by an established shoe manufacturer (Chaneco; www.chaneco.co.uk). Shoe size was assessed during the screening visit, and shoes were ordered after eligibility had been confirmed. The unloading shoe was a trainer-type shoe with a black leather upper section, laces, and a specially-designed rocker sole (Fig. 1a). The rocker soles, which were manually shaped according to the specifications of the patent that is owned by the University of York (Patent no.: GB2458741B), comprised three circular curves with arc centres that are positioned at the anatomical ankle, hip and knee, respectively (assuming a vertical lower limb), and so forming a posteriorly-placed apex to the rocker shape. This is designed to influence the line of action of the ground reaction force to pass close to the anatomical joint centres and so reduce the moments needed to be generated for ambulation by the muscles acting across those joints in the lower limb. Additionally, it is designed to place the ankle into a relatively plantarflexed position where the ankle plantarflexors use less energy than, for instance, when placed in dorsiflexion. This is because it also increases the lever arm between the Achilles tendon and the ankle joint; so making propulsion, and therefore calf muscle power generation, more efficient. It is also intended to unload the calf muscles by providing a simultaneous reduction in ankle range of motion in relative plantarflexion but still moving with a near-normal trajectory. To facilitate participant blinding, the control shoes were made to be similar in appearance to the unloading shoes (Fig. 1b). These shoes had the same upper section as the unloading shoes, but a different rocker sole. Here, the apex of the sole was anteriorly-placed, which is not designed to place the ankle in relative plantarflexion during stance phase of gait. Participants were allowed to habituate to wearing each pair of shoes for 5 min before commencing the first walking test.
Assessment procedures and outcome measures
Both assessment visits involved three walking tests that were separated by 20-min periods of seated rest: (i) a 6-min corridor walk test to quantify 6-min walk distance (6MWD) [8], (ii) a usual-pace walk test to measure pain-free walking distance, and (iii) a “figure-of-8” walk test during which gait biomechanical parameters were quantified as described previously [9]. Heart rate (via telemetry: Polar T31 transmitter with Polar FT1 watch, Polar Electro, Oy, Finland), blood pressure (Omron M6 Comfort, Omron Healthcare Europe B.V., Hoofddorp, The Netherlands), and ratings of perceived exertion (Borg 6–20 scale [10]) and leg pain (Borg CR-10 scale [10]) were recorded before and immediately after each test. All participants had a leg pain score of 0 before commencing the next test. For the 6-min walk test, we used a 30-m straight corridor and standardised instructions [11], which included to walk as far as possible within the 6 min. The same course was used for the usual-pace test. The figure-of-8 test was conducted in a gait laboratory. A figure-of-8 was chosen to minimise the potential for fatigue that might have been be caused through participants solely performing all clockwise or all counterclockwise turns. Reflective markers were positioned on anatomical landmarks of the lower extremities using double-sided sticky tape to allow 3D motion analysis [9]. Participants were instructed to walk at their usual pace along a figure-of-8 circuit, without slowing down, for a maximum of 12 min. A force plate (9281EA, Kistler, Germany) positioned in the central straight portion of the course captured kinetic data. The participants were naïve to the force plate, to help ensure a natural walking gait. Infra-red 3D optical motion analysis cameras (Oqus, Qualisys, Sweden) captured kinematic data each time a participant approached and passed over the force plate. Kinetic and sagittal plane kinematic data were exported to Visual 3D motion analysis software (C Motion, Rockville, MD, USA) for processing and analysis. Inverse dynamics were used to determine joint moments and powers. Participants indicated when they experienced the onset of claudication pain and continued walking until pain prevented them walking further. Time-distance variables used to identify gait differences between the two shoe conditions during pain-free walking were walking speed, step length, step cadence, and time in stance phase, swing phase, and double support (% of gait cycle). The potential calf unloading effect of the adapted shoes was also explored using the following variables for the most affected limb: ankle range of motion, peak plantarflexion angle, peak plantarflexion moment (in Nm per kg body mass), and peak plantarflexion power (in W per kg body mass).
After completing the second assessment visit, all participants were given the pair of shoes that they wore during that visit to wear for 2 weeks. During this period, the participants were instructed to wear the allocated shoes as much as possible every day, with a minimum target of 4 h per day [12]. On completing the 2-week period, the participants were asked to return the shoes along with a completed survey about them. In the survey, participants were asked to estimate, on average, how many hours per day they wore the shoes. They also rated the overall level of shoe comfort using an 11-point (0–10) numeric rating scale (with terminal descriptors of ‘extremely uncomfortable’ and ‘extremely comfortable’), and perceived changes in walking ability and physical activity using 5-point (1–5) scales (with terminal descriptors of ‘much worse’ to ‘much better’ and ‘much less physically active’ to ‘much more physically active’, respectively). Finally, participants recorded any benefits, negative aspects, and untoward medical events related to the shoes.
A sub-sample of 12 participants also undertook a telephone-based interview to share their thoughts about the shoes. Purposive sampling was used according to the following criteria: shoe type (unloading and control), age (above and below 65 years), sex, and walking ability (6MWD above and below 350 m). The interviewer sought feedback regarding factors affecting shoe usage, benefits and negative consequences of wearing the shoes, and the design of the shoes. All interviews were audio-recorded, transcribed, and analysed to identify themes.
Adverse events
We recorded all serious adverse events (regardless of cause), and all non-serious adverse events that were believed to have occurred as a result of performing a study assessment, or from using the study shoes. The latter are subsequently termed ‘adverse device effects’.
Randomisation, allocation concealment and blinding
The order of testing for each participant (i.e., unloading shoes first then control shoes, or vice versa) was determined using a computer-generated randomisation sequence created by a statistician at York Trials Unit, who was not otherwise involved in the study. Blocked randomisation with a block size of 8 was used to ensure that the overall order of testing was balanced (ratio 1:1). The allocations were blinded (i.e. labelled AB and BA) before being passed to the trial statistician. Once a participant had completed the screening visit, an investigator emailed the trial statistician who assigned the participant to the next available allocation.
Participants were blinded to allocation by using control shoes that were visually-similar to the unloading shoes and by stating in the participant information sheet that the study was investigating two different types of shoes, rather than comparing normal and adapted shoes. Our attempt to blind the outcome assessor was unsuccessful because they were not naïve to the true purpose of the study and therefore could tell which shoe was the unloading shoe when preparing the participant for the gait analysis. However, the use of standardised testing procedures and objective outcomes (e.g., 6MWD) ensured that the risk of detection bias is low. The researcher overseeing data entry and the statistician remained blinded until the analysis was complete.
Sample size
The primary outcome was 6MWD measured in metres. The cross-over ANOVA square root of the mean squared error for 6MWD was found to be 30 m in a recent trial [13]. A mean difference of 25 m has been suggested as the minimum clinically important difference [14]. Using these values at 90% power and 2-sided 5% significance level in a cross-over design would require 34 participants. Therefore, recruitment stopped once 6MWD had been collected at both assessment visits for 34 participants.
Statistical analysis
Formal analyses were conducted following the principles of intention-to-treat with participant’s outcomes analysed according to their original, randomised testing order irrespective of the order that they actually received the shoes, where data were available. Analyses were undertaken in Stata v13 using two-sided statistical tests at the 5% significance level. Participant baseline data are summarised descriptively overall and by testing order (AB or BA) both as randomised and as analysed in the primary analysis. No formal statistical comparisons between testing orders were undertaken on baseline data. To allow for a possible period effect, analysis of the 6MWD was via a two-sample t-test to compare the difference between assessment 1 and assessment 2 for the two sequences. Dividing the resultant difference (and corresponding 95% confidence limits) by two gives an estimate of the treatment effect (i.e., A minus B) and 95% CI. Pain-free walking distance at usual pace was analysed in the same manner. Kinetic, kinematic and temporal-spatial measures of gait were taken at each assessment visit and calculated for each participant when pain free, at the onset of pain, and at absolute pain. The difference between the measure as assessed at visit 1 and visit 2 was calculated for each participant at each point in time (pain free, pain onset, and absolute pain). These three differences were modelled using a covariance pattern mixed model, with sequence allocation (AB or BA), time and an allocation-by-time interaction as fixed effects and participant as a random effect. The mean differences (and 95% CI) between the two sequences were extracted for the pain-free, onset of pain, and absolute pain time points, and divided by two to obtain an estimate of the treatment effect A-B. Only the data for pain-free walking is presented in this manuscript; all other gait data will be published elsewhere.