Effect of knee unloading shoes on regional plantar forces in people with symptomatic knee osteoarthritis – an exploratory study

Participants in this study were a sample of convenience that comprised a subset of volunteers enrolled in a 6-month randomised controlled trial comparing the effects of unloading and conventional walking shoes on knee OA symptoms [18]. The number of participants in this study was dictated by the number of participants in the control group (conventional shoes) whom were willing to undergo in-shoe plantar pressure data collection upon exit from the trial. For this exploratory cross-sectional study, 21 participants were recruited after completing the final 6-month assessment for the randomised controlled trial. Participants were included in the trial if they were aged 50 years or older, had knee pain on most days of the previous month, reported average pain in the previous week of at least 4 out of 10 on an 11-point numerical rating scale (NRS) (with terminal descriptors of “no pain” and “worst pain possible”), had radiographic evidence of medial compartment knee OA (Kellgren-Lawrence (KL) grade ≥ 2), and had definite medial tibiofemoral osteoarthritis on radiography. Exclusion criteria for the trial have previously been published [18]. Study procedures were approved from the University of Melbourne Human Research Ethics Committee (HREC No. 1239045) and all participants provided written informed consent.

The unloading shoes were black leather commercially available unloading walking shoes (GEL-Melbourne OA [ASICS Oceania]) with triple-density midsoles that were stiffer laterally (Shore A rating of 65) than centrally (Shore A rating 55) and medially (Shore A rating of 45), and with mild (5-degree) lateral-wedge insoles attached to the underside of the sockliners (Fig. 1) [11]. The GEL-Melbourne OA weighed 334 g (based on women’s shoes size 8US). The conventional shoes were black leather commercially available neutral walking shoes (GEL-Odyssey [ASICS Oceania]) that were visually similar to the unloading shoes but possessed a mono-density midsole (Shore A rating of 55) and did not contain the lateral wedge. The GEL-Odyssey weighed 289 g (based on women’s shoes size 8US). During testing for this study, participants were not informed which shoes were the unloading shoes.

An in-shoe plantar pressure measurement system (pedar-X insoles, version 24.3.5 software, Novel gmbh, Munich, Germany) was used to record regional plantar pressures for both feet at 100 Hz [19]. Each insole had 99 capacitive cells distributed throughout the insole.

Participants completed walking trials at their own self-selected speed in the unloading and conventional shoes, presented in random order, over an 8-m walkway. Two sets of photoelectric timing gates four meters apart were used to ensure walking speed was maintained within 5% of each participant’s average. For each shoe condition, data collection commenced approximately two meters after the beginning and before the end of the walkway to ensure steady state gait. Three practice trials were permitted for each shoe condition and discrete data from all recorded stance phases of the affected limb (usually 2–3) of the fourth trial were averaged per participant. For participants with bilateral knee symptoms, data from the most symptomatic knee was collected.

Insole pressure data from every individual cell were exported into custom Matlab (Mathworks Inc., USA) code, and converted from pressure (1 kPa = 0.1 N/cm²) to force for analysis, using the known (different) area of each cell: force (N) = pressure (N/cm²) x cell area (cm²). Medial and lateral sub regions were analysed. For these, the foot was initially separated into three regions: forefoot (40% of the total foot length), midfoot (30% of foot length) and heel (30% of foot length) [20]. The toe sub region was disregarded, because data from this region is highly variable [19]. To determine the medial and lateral regions of the heel and forefoot, we then divided each row of sensors in those regions in half. If a row contained an odd number of sensors, data from the centre sensors were disregarded.

Force across all cells in each sub region were summed at each point in time (Fig. 2). As force data was analysed paired across shoe conditions for each participant, no normalization for body size was applied. Centre of pressure data for the whole foot were exported separately via the FGT file format, as the manufacturer would not supply geometric locations of each sensor for CoP to be calculated. The insole XY coordinate system has its origin medial to the most posterior cell, with positive X direct laterally (through the base of the posterior cell), and positive Y anteriorly (through the most medial cell).

Peak total force (N) during stance was extracted for the following regions: lateral heel, medial heel, lateral forefoot, and medial forefoot. The ratio of the peak medial heel total force to the peak lateral heel total force during stance was also calculated [12]. Centre of pressure antero-posterior (x) and medio-lateral (y) positions (in mm) were extracted at 25% of stance phase [7]; this point was used because it roughly coincides with the first peak in the KAM [21]. The arch index was calculated as the ratio of midfoot loaded area relative to total loaded area excluding the toes [22], at midstance.

Age, sex, height and body mass were recorded, and body mass index (BMI, in kg/m²) was calculated. Participants reported symptom duration and laterality of symptoms. All participants underwent a semi-flexed weight-bearing postero-anterior knee radiograph to determine OA severity using the KL grading system (Grade 2 represents mild OA, Grade 3 moderate, and Grade 4 severe). Foot posture was assessed using the Foot Posture Index (FPI), a valid and reliable tool that scores six features of foot posture from − 2 (more supinated) to + 2 (more pronated) [23]. Total scores range from − 12 (highly supinated) to + 12 (highly pronated). Scores were categorised into highly supinated (< − 4), supinated (from − 4 to − 1), neutral (from 0 to + 5), pronated (from + 6 to + 9) and highly pronated foot posture (> + 9) [23]. Vertical and medial-lateral mobility of the midfoot was assessed with the foot mobility magnitude (FMM) [24], calculated as the square root of the sum of the squared difference in arch height from non-weight bearing to weight bearing and the squared difference in midfoot width from non-weight bearing to weight bearing. Foot mobility was also assessed using the navicular drop test, which is the difference in navicular height (in mm) between non-weight bearing and weight bearing [25]. Overall average pain while walking over the previous week was measured via an 11-point NRS [26]. Difficulty with physical functioning over the previous 48 h was measured by the function subscale of the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC, Likert version 3.1), with total scores ranging from 0 (no dysfunction) to 68 (maximum dysfunction) [27].

Descriptive statistics are presented as means with standard deviations and numbers with proportions as appropriate. Mean differences in variables between shoe conditions were calculated, along with 95% confidence intervals. The main effect of ‘condition’ (i.e. unloading or conventional shoes) on variables was evaluated using a mixed linear random effects model with ‘condition’ as a fixed effect and ‘participant’ as random effect. To investigate if ‘foot posture’ modified the effect of shoes on plantar forces, we firstly dichotomised the FPI score as either neutral or pronated (pronated/highly pronated; no participant had a supinated or highly supinated foot posture). We then added foot posture to the mixed linear model as an additional fixed effect to test interaction between ‘condition’ and ‘foot posture’. Similarly, we also tested interaction between ‘FMM’ and ‘navicular drop’, respectively with shoe condition in separate models. FMM and navicular drop were added as continuous variables. All analyses were performed using STATA version 14.1, with an alpha level of less than 0.05 considered as statistically significant.

Participant characteristics are presented in Table 1. Approximately half of the participants were female (57%). Participants were generally overweight [28] and mean symptom duration was 8 years. A range of radiographic disease severities was observed, with 29% of participants having mild, 43% having moderate, and 29% severe radiographic OA severity. Approximately half of the participants had a neutral foot posture (43%), and half had a pronated foot posture (57%) based on FPI score. Mean (SD) walking speed was not significantly different between conventional shoes (1.36 m/s (0.19)) and unloading shoes (1.35 m/s (0.20)).

The effects of unloading shoes on in-shoe regional plantar forces, relative to conventional shoes, are summarized in Table 2. Force in the lateral heel and forefoot regions significantly increased with unloading shoes (both P < 0.001), with corresponding decreases in the medial heel (P = 0.001) and forefoot (P = 0.005) force (Fig. 3). As a result of these changes, the medial-lateral heel force ratio also decreased significantly with unloading shoes (P < 0.001). Unloading shoes significantly shifted whole foot CoP anteriorly (P = 0.034) and laterally (P < 0.001). There was no difference in arch index between the shoes (P = 0.093). Whilst all but three participants demonstrated an increase in lateral heel force with unloading shoes relative to conventional shoes (Fig. 4), there was a large individual variation in response. However, FPI, FMM and navicular drop were not found to significantly moderate the effect of unloading shoes for any variable (Table 2).