Do different multi-segment foot models detect the same changes in kinematics when wearing foot orthoses?

A convenience sample of healthy individuals with no congenital, or acquired pathology of the nervous or musculoskeletal systems, no deformities or serious injuries of the pelvis or lower limbs and feet, no self-reported lower limb/foot pain were included in the study. This study was approved by the Ethical Committee of the Faculty of Physical Culture, Palacký University in Olomouc, Czech Republic (reference number 3/2018) and all participants provided written informed consent prior to data collection.

Participants walked at a self-selected walking speed along a 15 m walkway whilst wearing a correctly sized pair of ProTouch Drop Shot trainers (IIC-INTERSPORT, Bern, Switzerland) under two conditions; no orthosis and a pair of foot orthoses with a medial heel bar, positioned under the sustentaculum tali which aimed to minimise foot eversion during the stance phase (see Fig. 1). The OFM and CAST foot kinematics were captured simultaneously at 200 Hz using an eight camera Vicon Vantage V5 (Oxford Metrics, UK) motion capture system.

The CAST lower limb model was used to measure lower limb kinematics [18]. Anatomical markers were placed on the anterior and posterior superior iliac spines, medial and lateral femoral epicondyles, and medial and lateral malleoli. The shank segment was defined proximally by the knee joint center, estimated from the medial and lateral femoral epicondyle markers, and distally by markers placed on the medial and lateral malleoli. Clusters of non-collinear markers were placed on the lower leg (i.e. shank) and the thigh. Marker placement for both foot models is displayed in Fig. 2. Retroreflective markers for the OFM were positioned according to Stebbins et al. [7] with a following modifications. Instead of a marker wand positioned on the posterior aspect of the calcaneus, the lateral calcaneus (HF2) marker was used for hindfoot tracking, the sustentaculum tali marker was moved inferior to fit on the shoe surface, and the hallux was not tracked in this study due to not being able to place a marker on the hallux under shod conditions.

The proximal and distal CAST MFM hindfoot was defined by the medial and lateral malleoli (MMAL, LMAL) and sustentaculum tali (STAL) and lateral aspect of the calcaneus (LCAL, at the same distance from the most posterior point as STAL [7]) markers, respectively. The hindfoot was tracked using markers positioned on the medial and lateral hindfoot (HF1, HF2) in addition to the marker positioned inferior on the posterior aspect of the calcaneus (CAL1), and STAL and LCAL markers. The proximal and the distal forefoot were defined by the STAL and LCAL markers, and the first and the fifth metatarsal markers (D1MT, D5MT), respectively. The forefoot was tracked using markers positioned on the base of the 1^st and 5^th metatarsals, and between the base of the 3^rd and the 4^th metatarsal (P1MT, P5MT and P3MT, respectively). The two most anteriorly placed markers (FF1, FF2) were not used by either model in this analysis (Table 1).

A single static trial with participant standing with a comfortable relaxed posture was captured and used to calculate local coordinate systems of the segments. Anatomical markers were removed after the static trial and all tracking markers remained in place during both conditions. For each participant, twenty-five trials were collected, of which seventeen trials had complete marker trajectories for all markers, and were used for the analysis. The two orthotic conditions were randomized by participants picking different coloured balls from a bag which represented the with and without orthotic conditions. A five-minute familiarization and wash-out period was used between each condition.

Kinematic data were processed in Vicon Nexus 2.8 and exported to C3D format. Heel strike and toe-off were identified manually based on linear acceleration, velocity and visual inspection of the heel and toe marker trajectories. The kinematic data were filtered with a fourth-order low-pass Butterworth filter with a 6 Hz cut-off frequency. The CAST lower limb model and both the CAST MFM and the OFM were applied in Visual 3D (C-motion, USA) using segment optimization pose estimation. For the shank, hindfoot and forefoot segment coordinate systems of both foot models, the motion about a medio-lateral (X), an antero-posterior (Y) and a vertical (Z) axis was plantar/dorsiflexion, inversion/eversion and abduction/adduction, respectively. The X–Y-Z Cardan rotation sequence equivalent to the segment coordinate system was used to calculate joint kinematics [23], and data were normalized to 100% of the stance phase.

Statistical analysis was carried out using spm1d package version 0.4.3 (http://​www.​spm1d.​org/​) in Python version 3.8. The D’Agostino-Pearson K2 test was used to assess the time series data normality. Data were not normally distributed; therefore, the non-parametric version of vector field analysis, statistical non-parametric mapping (SnPM) was used [24]. SnPM paired t-tests (p < 0.05), with the number of iterations set to 10,000, were used to explore the effect of the medial heel bar over the stance phase for each foot model for each participant in all three anatomical planes using observations from both feet. The hindfoot relative to shank and forefoot relative to hindfoot segments in both feet were compared between models. The segments for each foot model had comparable markers and thus hypothesised that outputs would be similar, despite having subtly different anatomical axes definitions. The stance phase was split into the first double support (DS1), single support (SS) and the second double support (DS2) phase, identified from the gait events, and statistically significant kinematic effects of the medial heel bar, represented by the suprathreshold clusters in the SnPM analysis, were compared between the two foot models for each sub-phase separately. For each sub-phase of the gait cycle, possible kinematic outcomes between foot models were; the same kinematic effect (blue), no effect for both models (turquoise), a unique effect of either the CAST MFM (green) or OFM (orange), or an opposite effect of the two foot models (red). The visual inspection of the SnPM analysis outputs showed considerable number of small, arguably clinically irrelevant suprathreshold clusters (waveform areas showing statistically significant differences). The authors could not find any data to identify minimal clinically relevant suprathreshold cluster size specific to this type of analysis, therefore, based on the visual inspection, only differences that were significant for more than 5% of the stance phase were considered meaningful in the different sub-phases and were included in the analysis (Fig. 3).

At the hindfoot, there was good agreement between the foot models, with both foot models detecting the same kinematic changes to the orthotic condition in 60.9% of cases, with 25.5% of cases demonstrating no kinematic differences in both foot models. The CAST MFM was able to detect unique kinematic changes in 7.8% of the cases in the sagittal plane (10.9% in DS1, 10.9% in SS, 1.6% in DS2), 5.7% of cases in the coronal plane (4.7% in DS1, 7.8% in SS, 4.7% in DS2), and 7.8% of the cases in the transverse plane (10.9% in DS1, 6.3% in SS, 6.3% in DS2). The OFM was able to detect unique kinematic changes in 14.1% of the cases in the sagittal plane (14.1% in DS1, SS, and DS2), 3.6% of the cases in the coronal plane (4.7% in DS1, 6.3% in DS2), and 3.1% of the cases in the transverse plane SS, with an opposite effect being observed in 1.6% of the cases in the sagittal plane in DS2.

At the forefoot, the same overall effect of the medial heel bar was observed in 51.3% of the cases with both models being able to detect the same kinematic change in 48.3% and no kinematic change in 3.0% of the cases. The CAST MFM was able to detect unique kinematic change in 10.9% of the cases in the sagittal plane (14.1% in DS1, 10.9% in SS, 7.8% in DS2), 7.3% of the cases in the coronal plane (10.9% in DS1, 3.1% in SS, 7.8% in DS2), and 32.3% of the cases in the transverse plane (17.2% in DS1, 29.7% in SS, 51.6% in DS2). The OFM was able to detect unique kinematic change in 13.0% of the cases in the sagittal plane (14.1% in DS1, 17.2 in SS, 7.8 in DS2), 4.2% of the cases in the coronal plane (1.6% in DS1, 7.8% in DS2, 3.1 in DS2), and 9.4% of the cases in the transverse plane (9.4% in DS1, 10.9% in DS2, 7.8 in DS2). The opposite effect was observed in 31.8% of the cases in the sagittal plane (25.0% in DS1, 37.5% in DS2, 32.8% in DS2), 4.7% of the cases in DS1, SS, and DS2 in the coronal plane, and 32.3% of the cases in the transverse plane (46.9% in DS1, 35.9 in SS, 14.1% in DS2).

At the hindfoot, the overall agreement of both models across all planes over the entire stance phase was very good, potentially due to the similarities in the markers used to track the hindfoot segment, despite the subtle differences in the segment coordinate system definitions. Also the use of the same biomechanical conventions could contribute to the good agreement of the models as it was suggested to be more crucial than the design of relevant marker sets in the comparison of different gait protocols [25]. Both models were in very good agreement in detecting the reduction of hindfoot eversion, which is consistent with past research on the effect of medial heel bars on foot [26] and hindfoot [5] kinematics, and rearfoot posting [27‐29]. The CAST MFM detected slightly more (~ 2% of cases) unique kinematic changes than the OFM. No opposite effect was present. The small differences between models could be explained by an extra tracking marker in the CAST MFM which may improve the precision of the measures of that segment, however the effect of the differences in the segment coordinate system definitions cannot be ruled out.

In the transverse plane, both models were in very good agreement in detecting increased hindfoot abduction, which was reported to be the effect of medial heel bar on hindfoot [5] and foot [26], across all the phases. The CAST MFM was able to detect a unique change in more cases (approximately ~ 8%) across all stance sub-phases, while the OFM was able to detect a unique kinematic change (~ 3% of cases) and only in SS phase. In the sagittal plane, there was reasonable overall agreement in the ability of both models to detect increased dorsiflexion (~ 77%), which was previously identified as the effect of a medial heel bar on the hindfoot [5] and the foot [26], across all phases, however the OFM detected a unique kinematic change in approximately twice as many number of cases (~ 14%, see Table 2) across all the phases than the CAST MFM. Hindfoot results imply a minor shift between hindfoot local coordinate system axes of both models leading to a different distribution of the three dimensional joint rotations across the anatomical planes, which could potentially explain the subtle differences in the sensitivity of the models to be able to detect kinematic changes in a given plane across all phases.

At the forefoot there was a marked difference in overall agreement of both foot models across all planes over the entire stance phase (approximately 50% of cases), whilst there was a marked increase (~ 23% of all cases) in opposing kinematic change between the MFM. However, this measurement of response due to the foot orthosis differed significantly across anatomical planes. The highest level of agreement of both models was observed in the coronal plane, where previous findings demonstrated the medial heel bar increased inversion of the forefoot during the entire stance phase [5]. On the contrary, the level of agreement between the CAST MFM and the OFM was rather poor in the sagittal and transverse planes (~ 45% and ~ 25% of cases, respectively) with the opposite effect detected in approximately a third of the cases in both planes. These differences could be explained by the different number and position of tracking markers with the OFM using an extra tracking marker. While the three CAST MFM tracking markers are positioned above the base of the metatarsals, two out of the four OFM tracking markers are positioned in a more anterior position (mid-point between 2^nd and 3^rd metatarsal heads and 5^th metatarsal head), which could possibly explain the difference in detected kinematic change as the OFM is partially tracking a more distal part of the foot. In the transverse plane, the median medial heel bar effect on the peak abduction of the forefoot was reported to be less than 0.3º degrees [5], which together with different tracking marker placements could contribute to the high percentage of disagreement between the two models.

In foot orthoses research it is common practice to report the material, density, shape and inclination of foot orthoses’ elements [27, 28] in order to explore a particular prescription and its variations [29]. Knowing the features of various MFMs is no different. The medial heel bar used in this study has been shown to decrease eversion of the hindfoot in healthy adults, which may benefit patients with abnormal hindfoot pronation [5]. Both the CAST MFM and the OFM showed very good agreement in the ability to detect kinematic changes as a result of the medial heel bar in the coronal plane (90.6% in the hindfoot, 83.6% in the forefoot), which implies both models could be used interchangeably. However, recent work has shown that the effect of foot orthoses [30] and specifically designed shoes [31] is not necessarily universal and responders and non-responders can be identified from biomechanical data. In the coronal plane of the hindfoot, the CAST MFM and the OFM detected unique kinematic changes in 5.7% and 3.6% of the cases, respectively, which shows a level of disagreement between the models in the identification of possible biomechanical responders and non-responders in 9.3% of the cases. The level of disagreement in the identification of possible biomechanical responders and non-responders in the coronal plane of the forefoot was 11.5% of cases with the CAST MFM and the OFM being able to detect a unique kinematic change in 7.3% and 4.2% of cases, respectively. This level of difference could potentially have an impact on the identification of biomechanical responders and non-responders to the hindfoot medial posting intervention. This is an important area for future work which is required to explore the ability of two models to identify responders and to determine the most appropriate model, if any. As there are different versions of the OFM producing slightly different kinematic outputs [20], the level of disagreement may change depending on the version used.

This study had some limitations. All the participants were healthy individuals therefore the magnitude of the effect and potentially the ability of both models to detect kinematic change may differ for people requiring orthotic interventions. Previous comparisons of the OFM and Rizzoli Foot Model have shown differences present in kinematic outputs which also depended on gait type [20]. Due to its position, the motion of the markers placed on the midpoint between the 2^nd and 3^rd metatarsal heads and on the 5^th metatarsal head could be susceptible to the deformation of the shoe, especially during the heel rise, which could have an effect on the OFM forefoot kinematics.