Changes in foot pain, structure and function following bariatric surgery

This study recruited a convenience sample of people with foot pain from bariatric surgery waiting lists at two tertiary hospitals in Adelaide, South Australia between January 2015 and June 2017. All patients on the lists were invited to participate in this study. Participants were recruited either immediately before surgery (treatment group) or when they were added to the waiting list (control group). All participants underwent baseline measures and were reassessed at six-month follow-up. Baseline measures for participants in the treatment group were recorded 2–3 weeks prior to surgery, designed to reduce the impact of non-surgical weight loss that occurs with meal replacements prior to surgery. Bariatric surgery procedures (sleeve gastrectomy, laparoscopic adjustable gastric band or Roux-en-Y gastric bypass) were based on the clinical requirements of each patient, and not directed by this study. The study received ethical approval before commencement by the Southern Adelaide Clinical Human Research Ethics Committee (Project ID 211.14).

People aged ≥18 years were recruited if they reported current foot pain of ≥30 mm on a visual analogue scale [22], indicating moderate pain (at least), which had to have been present for ≥3 months. Participants were excluded if: pregnant, history of previous bariatric or foot surgery, systemic inflammatory disease, loss of peripheral sensation in the feet, known infectious disease, cancer, non-ambulatory.

Self-reported age and gender were recorded at baseline. Body weight and height were measured to the nearest 0.1 kg and 0.1 cm, respectively using an electronic stadiometer (with shoes, socks, and bulky clothing removed) (Seca 284, Germany). From these data, BMI (weight (kg)/height (m²)) was calculated [23].

Foot pain and disability were assessed using the Manchester-Oxford Foot and Ankle Questionnaire (MOXFQ). The MOXFQ is a reliable and valid 16-item questionnaire that comprises three separate underlying dimensions: walking/standing problems (seven items), foot pain (five items), and social interaction (four items). Item responses are each scored from 0 to 4, with 4 representing the most severe state [24]. The MOXFQ-index is a summary score that incorporates all three domains and was used for this study as it investigates overall foot pain and disability and is graded on a 0 to 100 scale, with 100 being the most severe state.

The Center for Epidemiologic Studies Depression scale (CES-D) consists of 20-items that assess depressive symptoms [25]. All items are graded via a four-point (0 to 3) Likert scale, giving a possible range of 0–60. The sum score was used as a continuous variable to assess depressive symptoms at baseline and follow-up.

Dynamic plantar pressure data were collected with the MatScan® (Tekscan, USA) platform system at baseline and follow-up. The platform is a 5 mm-thick floor mat (432 × 368 mm) incorporating 2288 resistive sensors (sensor size = 0.70 cm², 1.4 sensors/cm²) with data sampled at a rate of 40 Hz. Analyses conducted using the MatScan® platform have been previously shown to have good accuracy [26] and moderate to good inter-session reliability in adults, for total peak pressure and maximum force ICC (95% CI) of 0.58 (0.28 to 0.75) and 0.92 (0.84 to 0.96), respectively [27]. The MatScan® platform was positioned in a level walkway. Following individual step calibration, which involved entering the participant’s body weight followed by a period of standing on the MatScan® platform, balancing on one foot. The participants were then asked to walk barefoot with their usual gait pattern across the MatScan® platform. Data were collected using a midgait protocol, whereby participants were instructed to take two steps, striking the platform on their third step, before continuing to walk for a further three steps. The midgait protocol has been found to have moderate to excellent reliability, for total peak pressure and maximum force ICC of 0.52 to 0.97 and 0.36 to 0.73, respectively [28]. Data were collected from three complete, valid trials of the right foot. A valid trial was determined when participants’ gait pattern was not perturbed and when no ‘targeting’ of the plantar pressure platform occurred. Individual “masks” were manually constructed to determine plantar pressures for the whole foot and also five regions of the foot; heel, midfoot, forefoot, hallux and lesser toes, using the Research Foot software (version 6.51). Measures of regional contact area (cm²), maximum force (kgf), peak plantar pressure (kPa) and contact time (ms) were calculated for all trials to obtain a mean value. The right foot was chosen to ensure that the assumption of independence of data was met [29] and additional measures such as mean pressure or pressure-time integral were not collected given the interdependence between these measures and peak pressure [30, 31]. Contact time was used as a proxy for walking speed [32].

Weight-bearing dorsoplantar and lateral radiographs were taken at baseline and follow-up using a standardised technique on a digital radiograph unit (Ysio, Siemens Healthcare, Germany). Dorsoplantar views were taken with the participant standing above the image receptor on both feet, the central beam was angled at 15 degrees towards the calcaneus and the centering point was aimed at the base of the third metatarsal. Lateral radiographs were taken with the participant standing on a platform, the image receptor was positioned on the medial aspect of the foot and the centering point was at the base of the metatarsals from a lateral to medial direction.

Foot posture was assessed using four radiographic angles described by Murley et al. [33]. The calcaneal inclination angle is the angle between the inferior aspect of the calcaneus and the supporting surface, while the calcaneal-first metatarsal angle is the angle on the dorsum of the foot taken between the inferior calcaneal angle and a line parallel to the midshaft of the first metatarsal. Both of these angles were measured from the lateral view. The calcaneal-first metatarsal angle was used as a measure of foot posture in the regression analyses, having been shown to strongly correlate with clinical measures [33].

Two further angles were taken from the dorso-plantar radiograph: the talo-navicular and talo-second metatarsal angle. The anteromedial and anterolateral extremes of the talar head and the bisection of the proximal articular surface of the navicular formed the talo-navicular angle. The talo-second metatarsal angle is formed between the bisection of the shaft of the second metatarsal and the line perpendicular to the anteromedial and anterolateral extremes of the head of the talus.

Ankle joint dorsiflexion was measured using the technique described by Munteanu et al. [34], which has very good intra- and inter-rater reliability (ICC (95% CI) 0.88 (0.75–0.94) and 0.92 (0.86–0.96)) when taken in a weight-bearing position with the knee in full extension. A digital protractor (Gain Express, USA) was placed on the anterior aspect of the tibia and the angle between the ground and the leg was recorded. The measure was repeated in duplicate and a mean was obtained.

Frontal plane range of motion of the hindfoot (ankle and subtalar joints) was measured using the technique described by Menadue et al. [35]. Participants are seated in an upright position with the foot positioned overhanging the examination table. The foot was moved from maximal abduction to maximal adduction with the total range of motion recorded with a goniometer (Physio-Med, AUS), this measure was repeated in duplicate and a mean was obtained.

Descriptive statistics (frequencies and means) were obtained. All data distributions were checked for normality via the inspection of histograms and the Shapiro-Wilks test prior to inferential statistical analysis. Differences between the treatment and control groups at baseline were assessed using chi-squared tests for categorical data and t-tests or Mann-Whitney U tests for continuous data that was normally or non-normally distributed, respectively. Multivariable linear regression was used to analyse the association between foot pain severity and mechanical (foot posture, plantar pressure, BMI) and non-mechanical (depressive symptoms) variables, along with age and gender at baseline and follow-up. Within group differences in contact area, force and peak plantar pressure between baseline and follow-up were analysed with either the paired-samples t-test or the Wilcoxon signed-rank test. As repeated measures were taken over six-months, a multilevel repeated model using the xtmixed command in STATA was used to assess the associations between MOXFQ-index and peak plantar pressure and other covariates (ankle joint range of motion [36], contact time [37], age and gender [38]) for both groups over six-months while adjusting for the interaction between group and follow-up time. This type of regression model allows for the correlations of observations within subjects. In all analyses, p values (two-sided) less than 0.05 were deemed to be statistically significant. All data analyses were performed with SPSS V25 (IBM SPSS Statistics, Armonk, NY, USA) and STATA V15 (StataCorp College Station, TX, USA).

This study recruited 45 participants (38 women), with a mean (standard deviation) (SD) age of 45.7 (9.4) years. Twenty-nine participants had bariatric surgery, undergoing a Roux-en-Y gastric bypass (n = 13), sleeve gastrectomy (n = 11), or laparoscopic adjustable gastric band (n = 5). Sixteen participants remained on the waiting-list and were used as controls. There were no significant differences in the characteristics between those undergoing bariatric surgery and those in the control group (Table 1). All participants underwent baseline measures, one participant who underwent bariatric surgery was lost to follow-up as they were uncontactable. Two additional participants from the treatment group were unavailable for follow-up plantar pressure data collection and four participants from the control group were unavailable for follow-up plantar pressure and foot posture data collection.

In multivariable regression analysis, only depressive symptoms were significantly associated with foot pain severity measured by MOXFQ at baseline (β = 0.7, 95% CI 0.2 to 1.2, p = 0.008). Age, gender, BMI, foot posture and whole-foot plantar pressure were included in the model, but these were not significantly related to foot pain severity, although age and BMI were trending towards statistical significance (Table 2). Multivariable regression analysis of the treatment group at follow-up found foot pain severity was also associated with depressive symptoms (β = 1.2, 95% CI 0.8 to 1.7, p <  0.001), while age, gender, foot posture and whole foot plantar pressure were not significantly associated with foot pain. There were no significant associations between foot pain severity and any of the listed variables in the control group (Table 3). The treatment group had a significant reduction in foot pain of 35.7 points (95% CI -42.2 to − 28.8, p <  0.001), while the control group had a small non-significant reduction of 4.0 points (95% CI -15.2 to 7.1, p = 0.454) (Table 4). Nine people (one-third) in the treatment group had complete resolution of their foot pain at 6-months follow-up.

The 29 participants who underwent bariatric surgery had a mean reduction in weight of 24.3 kg (95% CI -27.5 to − 21.1, p <  0.001), representing a mean per cent change in body weight of 19.9% (range 0.9 to 31.8%). The 16 participants who did not have bariatric surgery had a mean weight gain of 1.2 kg (95% CI -2.5 to 4.9, p = 0.877). In the treatment group, there were significant reductions in force and peak pressure in multiple regions, although the midfoot was the only site to exhibit a statistically significant mean reduction in contact area of 2.1 cm² (95% CI -3.9 to − 0.3, p = 0.022). Representative peak plantar pressures of participants before and after bariatric surgery are depicted in Fig. 1. In the control group, the only region to demonstrate a statistically significant change in plantar pressure was the whole foot variable, which increased by a mean 13 kPa (95% CI 0 to 25, p = 0.046). Depressive symptoms reduced in both groups, although only the treatment group had a significant mean reduction of 5.9 points (95% CI -10.3 to − 1.5, p = 0.007). There was no significant change in the ankle joint range of motion or foot posture in either the treatment group or the control group between baseline and follow-up. The crude changes between baseline and follow-up are detailed in Table 4.

The change in foot pain severity between baseline and follow-up was not associated with peak pressure when adjusted for other covariates. Both group (β = − 11.2, 95% CI -23.8 to 1.4, p = 0.081) and follow-up time (β = − 7.7, 95% CI -18.5 to 3.2, p = 0.168) variables were associated with a reduction in foot pain, but only the group*time (6-months) interaction was statistically significant (β = − 21.1, 95% CI -40.8 to − 13.5, p <  0.001) (Table 5).