In the present study, we performed gait analysis of 25 patients with persistent impairments after foot and ankle injuries during hospitalization in a rehabilitation clinic. The experimental procedure consisted of four short walking trials, wearing a lightweight 3D-accelerometer. We compare walking with and without orthopaedic shoes. The cadence remained unchanged between conditions. On the other hand, the results indicated that footwear improved local dynamic stability (LSD) of the gait (defined by using Lyapunov exponents). A significant effect was observed along the three axes, but it was more pronounced in the medio-lateral direction. Moreover, a substantial pain reduction was also observed.
In a previous study, we observed that orthopaedic shoes improved gait symmetry and reduced pain in patients with unilateral ankle injury . The gait symmetry was defined as the difference between healthy side and injured side in the accelerometric signal. In the present study, it was possible to include patients with unilateral and bilateral impairments, because local dynamic stability is computed over consecutive gait cycles. Therefore, this parameter may be used in a broader spectrum of pathological gait than symmetry assessment. The study was conducted in a clinical setting: therefore there are strong evidences that the method we used is readily applicable by practitioners: collecting acceleration data during short walking tests is a simple and rapid procedure (5 min). Unlike most kinetics/kinematics measures, the short walking tests can be performed by the regular medical staff directly on site. However, LDS calculation still requires custom software and advanced computer skills.
Non-linear indexes of gait variability have attracted growing interest as clinically valid indicators for the follow-up of various pathologies [20, 21]. LDS has already been shown relevant in the evaluation of various conditions such as in patients with peripheral arterial disease , in unilateral transtibial amputees , in knee osteoarthritis , or in Anterior Cruciate Ligament (ACL) rupture . LDS has been also used to study the effect of aging and fall risks in elderly [24–26]. The relative change in LDS induced by orthopaedic shoes lay between 7% and 10%. In order to evaluate the relevance of this change, we compared it to similar results in the literature: Manor et al.  reported that desensitization of the foot soles (induced by ice-exposure) led to LDS decrease by -40%. Sinitksi et al. , by applying visual and mechanical perturbations to walking individuals, found that LDS was reduced by about -25%. By impairing balance control with application of randomly varying Galvanic Vestibular Stimulation (GVS), van Schooten et al.  reported an effect on LDS of -11%. We reported that treadmill walking significantly increased LDS as compared to over-ground walking by +9% . The effect of aging on LDS, defined as relative difference between young and older adults, has been found to be about -70%  or -40% . On the other hand, some studies failed to report significant effects on LDS: for instance, no LDS change was reported after a single bout of resistance exercise (effect of fatigue: 0%). Walking without arm swing led to a non-significant slightly higher LDS (+3%). The values of the present study are on the lower end of the range of reported effects; in absolute value they are similar to the stabilizing effect of treadmill or to the destabilizing effect of GVS: although orthopaedic shoes did not stabilize gait in a large extent, it should be noted that most of the experimental studies that we examined report the effect of destabilizing interventions [13, 27, 28]. Conversely, we evidenced a non-invasive therapeutic intervention that immediately improve LDS, what, as far as we know, has no equivalency in the literature..
Although the study was conducted in a single centre, included patients came from all over the west part of Switzerland. Apart the gender imbalance, they were therefore well representative of the chronic foot&ankle injured population in Switzerland. Although there was a substantial diversity among the diagnoses (midfoot, hindfoot and ankle injuries), included patients were treated by a homogenous and constant team of specialists (physical medicine practitioners and orthotists), which ensured that the orthopaedic shoes had a constant design during the study. The AOFAS score is widely used to assess the degree of impairment in this type of patient, and can serve as a basis to generalize the results to other comparable populations. The average score at hospitalization time (57) is low as compared to other studies [30, 31], what reflects important disabilities and substantial pain that justified hospitalization. The significantly higher score at the end of stay (76) was the result of the multiple therapeutic interventions, including prescription footwear. Deeper analysis of AOFAS results goes beyond the scope of the present study and would require further investigations.
Speed was not assessed during the trials. Therefore, under the null hypothesis that footwear adaptation had no effect, we have to exclude the possibility that a significant speed difference occurred “by chance” between conditions, that would have induce a change in LDS. It is well established that SF is tightly related to walking speed [32, 33]. Because we observed that cadence was not significantly modified between conditions, it is likely that speed remained also unchanged. However, if a significant effect of orthopaedic shoes is hypothesized, it is possible that a step length change occurred, which induced increased speed without cadence modification. Effectively, some results in the literature support the fact that foot orthoses might improve step length more than cadence . As a result, it is possible that the LDS increase was concomitant to step length improvement. It should be noted that the relationship between LDS and speed is still a matter of debate . The results seem to show that there is little LDS modification around preferred walking speed  or that LDS decrease with walking speed (higher λs[36, 37]): consequently, it is very unlikely that the reported LDS increase was exclusively induced by higher speed.
Due to short acclimation time with new orthopaedic shoes (1-3 days), we did not evaluate the long-term beneficial effects of prescription footwear. Further follow-up studies are therefore needed. However, in a clinical context, it is crucial to evaluate the treatment effects rapidly, in order to adapt further interventions . Consequently, LDS is likely a responsive index that would be pertinent for the early evaluation of footwear outcome.
Classically, non-linear analyses of the gait variability (such as Lyapunov exponent, or Detrended Fluctuation Analysis), have been performed over long continuous walking trials [38–40]. It has been shown that short term LDS could be assessed with a good reliability with a trial of at least 2 minute  or 150 consecutive strides . It has been suggested that type II errors could be frequent with short walking samples. However, increasing the number of subjects  and multiplying the number of measured walking episodes  has been proposed to attenuate this effect. Consequently, two episodes of 20 strides and 25 patients have been used in the present study. Other studies successfully used LDS in short walking trials of 30 strides [22, 24, 37], 16 m walking distance , or 8 strides . Although it is likely that long duration trials could provide a higher precision, the results show that LDS was responsive to footwear change despite the low number of analyzed strides.
Treadmill walking has been shown to influence walking variability and stability [39, 40], questioning its use for gait evaluation. In addition, many patients cannot walk for a long period because of fatigue and pain. It has been shown that LDS was modified while turning as compared to walking in a straight line . A straight path should therefore be used to evaluate LDS in order to avoid potential bias induced by turns. Based on these considerations, the experimental design of the present study was to submit patients to four 30s trials along a straight corridor. It was a balance between a) the need to maximize the length of the trials in order to increase the precision on LDS, b) the necessity to avoid long and painful walking sessions, and c) the space limitation of the building (70 m corridor).
We observed that footwear effect was stronger for the medio-lateral axis (d = 0.55), and barely significant for antero-posterior direction (p = 0.04). Other studies also highlighted discrepancies between the different axes . With a comparable method (trunk acceleration), Chang et al.  observed significant differences in LDS between normal and compliant surfaces for medio-lateral and vertical axes, but not for antero-posterior axis, i.e. the same trend as in our results. Furthermore, it has been observed that, under dual tasking conditions (Stroop test), LDS was responsive along the three axes, with a lower sensitivity in AP direction . It is unclear whether these differences could have a biomechanical significance, however, it is worth noting that theoretical studies emphasize the role of lateral stability in falls avoidance .
Concerning pain, we measured a significant (p < 0.01) and practically relevant positive effect of the orthopaedic shoes (medium effect size: -0.76). Recent recommendations have stated that pain reductions ≥30% —a threshold close to our results (28%)— appear to reflect a least moderate clinically important differences .
What are the underlying mechanisms that could explain the observed improvement in local dynamic stability? Ankle injuries, and more specifically lateral ankle sprains, are often associated with chronic ankle instability. Both mechanical (joint laxity) and functional (proprioceptive disorders) instabilities have been described . Several studies related ankle instability to postural control deficit  and altered kinematics and kinetics . It has been shown that biomechanical changes induced by ankle instability are due to altered neuromuscular control , even at supraspinal level . The role of foot orthoses in the treatment of ankle instability has been emphasized . It has been proposed that improvement of tactile sensation could be a mechanism by which orthoses are beneficial to postural control . Furthermore, ice-induced plantar desensitization led to decreased LDS (approximately -40%), what bring evidence of relationships between foot sensitive feedbacks and gait stability. The functional impairments and disabilities of subjects included in the present study (AOFAS score: 57) goes far beyond the after-effects of simple lateral ankle sprain, such as functional instability. Most of the studied patients suffered from important foot and/or ankle pain while walking. Many exhibited mechanical restrictions at the ankle joint and/or at the subtalar joint. Moreover, it is obvious that severe foot and ankle fractures impair the proprioceptive and sensitive feedbacks at the foot and ankle level. In addition, it can be hypothesized that pain interferes with the motor control both at peripheral and supraspinal level (avoidance reflex). The results of the present study and previous results demonstrated that orthopaedic shoes reduce pain . Taking together those considerations and evidences in the literature about ankle instability, we hypothesize that the improvement in local dynamic stability was mainly due to pain reduction and/or enhanced proprioceptive/sensitive feedbacks provided by the orthopaedic shoes. The improvement in tactile sensations may have allowed the patients to better control their injured foot and hence to reduce noise in the motor control, that resulted in a better dynamic stability of gait. This hypothesis is also reinforced by the fact that most patients received low orthopedic shoes, which did not provide a strong mechanical support to the ankle: therefore, the explanation of a pure mechanical stabilization should be discarded. Because each patient received specific prescription footwear with multiple adjusted parameters, it is of course still unclear what is the exact cause of the observed results and which combination of adaptations would be more efficient for gait stabilization. However, the advice could be, if enhanced gait stability is part of the therapeutic goals, that more stress should be put on pain relief and foot comfort than on pure mechanical foot stabilization. The benefits of enhanced foot control may be an increased confidence in walking, with a concomitant reduction of stumbling and hence a lower risk of re-injury, but that remains to be investigated.