An observational study was carried out at the University Hospital Leuven following ethical approval of the local ethical committee and all participants provided informed consent (ML9038).

Between 2015 and 2022, 76 patients consulting our university hospital with symptomatic and radiographically confirmed arthritis of STJ and MTJ were invited to participate in the clinical motion analysis laboratory to collect a prospective database. All selected patients had a lack of satisfactory symptomatic response regarding conservative treatment, including foot orthoses, shoe modification and physiotherapy treatment (including massage, joint mobilization, strengthening, proprioceptive exercises). Since considerable heterogeneity existed within this population with respect to the medical background of the foot pathology, a further selection of patients was considered based on the following inclusion criteria: i) diagnosis of unilateral talonavicular, calcaneocuboid and subtalar osteoarthritis (at least a grade 2 Kellgren-Lawrence osteoarthritis score confirmed by two senior orthopedic surgeons) [19], ii) the ability to walk 100 m barefoot without rest or walking aids. Exclusion criteria were: i) radiographic signs of osteoarthritis in other foot joints (Kellgren OA score > 1), ii) uni- or contralateral joint fusions in the lower limb, iii) presence of systemic and or neurological diseases (e.g. rheumatoid arthritis, hemophilia, Charcot-Marie Tooth) or a history of osteoarthritis in any of the lower limb joints based on their available medical history and on clinical assessment, iv) people younger than 18 years. Finally, only 10 patients met the predetermined in- and exclusion criteria and were therefore selected for the current study.

Reference data serving as the control group in the current study was selected from an existing database of asymptomatic male adults. Selection of the specific control data was based on the age and BMI of the patient data.

Clinical and radiographic data were extracted from the patient electronic medical record. Standard weightbearing plain radiographs of the ankle and foot were used. Imaging was performed at the first consultation and at maximum 3 months before their gait analysis. Concerning the ankle, an anteroposterior, a mediolateral orientation and a mortise view were used. Concerning the foot, an anteroposterior, a mediolateral and a ¾ internal rotation view. Two senior foot surgeons evaluated the radiographs and a consensus meeting was organized regarding the scoring of the different joints on the Kellgren-Lawrence scale.

Gait analysis was performed in the institutional’s Clinical Movement Analysis Laboratory. Data were obtained when the participants walked along 10-m walkway, surrounded by an optoelectronic motion capture consisting of 10 T-10 cameras (100 Hz, Vicon Motion Systems Ltd., Oxford, United Kingdom). A force plate (Advanced Mechanical Technology, Watertown, MA, USA) and a superimposed pressure plate (Footscan, dimensions 0.5 m ∙ 0.4 m, 4096 sensors, 2.8 sensors/cm2; RSscan International, Paal, Belgium) were an integral part of the walkway. Both plates were dimensionally matched and were built into the floor. An RSscan 3D synchronization box was used to synchronize and calibrate the plantar pressure and force data.

The force and pressure plate data were sampled at 200 Hz.

Multi-segment foot kinematics and kinetics were assessed by placing retroreflective markers (Ø = 10 mm) to the participants’ feet and shanks following the marker placement protocol of the Instituto Ortopedico Rizzoli (IOR) Foot Model [20]. Subsequently, the patients were instructed to walk along the aforementioned walkway at their own pace until at least five representative trials were recorded.

Data processing incorporated manual marker labelling and definition of the individual gait cycles using Nexus software (Vicon Motion System Ltd, Oxford Metrics, Oxford, UK). Following this post-processing routine, the IOR-4segment-model-1 described by Deschamps et al. [21] was applied. This model calculates 3D intersegment joint rotations between the following adjacent segments: shank-calcaneus (Sha-Cal), the calcaneus-midfoot (Cal-Mid), midfoot-metatarsus (Mid-Met), hallux and metatarsus (Hallux); as well as the following non-adjacent segments: calcaneus and metatarsus (Cal-Met). The following terminology was used with respect to the aforementioned adjacent inter-segment angle calculations (joints): tibio-talar joint between shank and calcaneus, mid-tarsal joint between calcaneus and midfoot, tarso-metatarsal joint between midfoot and metatarsus, and the first metatarsophalangeal joint (MTP 1) between hallux and metatarsus.

Joint centers were respectively defined as the midpoint between both malleoli (tibio-talar joint), the midpoint between the navicular and cuboid bone (mid-tarsal joint), the second metatarsal base (tarso-metatarsal joint), and the projection of the MTP1 marker halfway to the floor (MTP1). Subsequently, ground reaction forces and moments (captured by the force and pressure plate) were distributed over the different segments of the IOR-4segment-model-1 using a validated proportionality scheme [22]. For every time frame of the gait cycle, the resulting pressure in each of these segments, compared with the total pressure, provided the proportion of the total ground reaction force to each corresponding segment. Then, we calculated inertial parameters based on the mass of each segment and their geometric solids. The mass of the foot was distributed at a 30/30/30/10 (rearfoot/midfoot/forefoot/hallux) percent rate. Joint kinetics were computed starting from the distal joint and progressing proximally, using Newton–Euler equations using an in-house custom inverse dynamic analysis program (ACEP-Manager, Matlab2016a, The Mathworks, Natick, US). Following data-processing, normalization of all waveforms for a full stance phase was performed.

Regarding the joint kinematics the following outcome variables were investigated: 1) range of motion (RoM), defined as the difference between the maximum and minimum value in a kinematic waveform, and calculated for three subphases of stance including loading response (0–20% stance phase), midstance and terminal stance together (21–83% stance phase) and pre-swing (84–100% stance phase) [23].

Finally, peak internal joint moment, peak dorsiflexion and plantarflexion angular velocity and peak power generation and absorption (joint moment multiplied with angular velocity) at the different joints were quantified as kinetic outcome measures.

All patients completed the Short-Form-36 (SF-36) [24] and an adapted Foot Function Index (FFI) validated for a Dutch speaking population [25, 26], and scored their pain experience using a visual Analog Scale (VAS). The SF-36 quantified physical and mental well-being in eight health concepts. A high score indicated a good outcome. The FFI quantified foot health, pain and foot health related quality of life with a score between 0 and 100. A higher score indicates a lower functioning person in terms of pain, disability and activity restriction.

IBM SPSS Statistics 20 (IBM Corp, Armonk, NY, USA) was used to perform the statistical analyses. Data were assessed for normality with the Shapiro–Wilk test. An independent samples t-test (if the assumptions for normality are achieved) or a Mann Whitney U test (if the assumptions for normality are not achieved) was used to compare the demographic parameters. One-way ANCOVA tests were computed to analyze group differences between the zero-dimensional parameters of the control and patient group. Walking speed was considered as a covariate since the control group had a significantly higher walking speed. To guard against inflation of type I error but maintain statistical power across the multiple comparisons made up on the multiple variables it was decided to adjust the conventional alpha level and adopt a P-value of < 0.01. Mean differences (with 95% confidence intervals) and effect size (ES) were also calculated. The effect size for Cohen’s d value was calculated and interpreted as follows: d = 0.20 (small effect), d = 0.50 (medium effect), d = 0.80 (large effect), and d = 1.30 (very large effect) [27]. Trends were still considered within the usual benchmark, α = 0.05.

The degree of kinematic coupling was evaluated by calculating the cross-correlation coefficient of the angular displacement curves of adjacent segments across the stance phase [28]. Based on previous publications [29], it was decided to analyse the coupling between four inter-segment rotations: i) Sha-Cal Inversion/Eversion with Cal-Met Dorsiflexion/Plantarflexion, 3) Sha- Cal Inversion/Eversion with Cal-Met Inversion/Eversion, ii) Sha-Cal Inversion/Eversion with Cal-Met Adduction/Abduction. The following qualitative benchmarks were used when evaluating this cross-correlation coefficient: 1) strong coupling > 0.7 or <  − 0.7), 2) moderate coupling between ( −)0.3 to ( −)0.69 and weak coupling between − 0.3 and 0.3 [30].