Understanding gait alterations: trunk flexion and its effects on walking neuromechanics Free

Upright bipedal walking is one of the most highly automated motor acts that humans perform and distinguishes humans from other mammals. It necessitates several specific adaptations of the locomotor apparatus (Huw Crompton et al., 1998; Spoor et al., 1994), including the erect posture of the trunk and head. Such an erect posture makes walking gait mechanically efficient because the centre of mass of the body (CoM) vaults over the supporting relatively straight limb like an inverted pendulum (Cavagna et al., 1977; Willems et al., 1995). This mechanism limits energy expenditure by means of a transduction between the kinetic (E_k) and potential (E_p) energy of the CoM (Cavagna et al., 1977; Dewolf et al., 2017; Willems et al., 1995). A major part of the muscular work is used to redirect the CoM from one pendulum arc to the next during walking (Kuo et al., 2005; Ruina et al., 2005). To do so, the propulsion of the trailing limb at the end of stance and the weight acceptance of the leading limb at the beginning of stance are coordinated to avoid collisional loss at foot contact (Dewolf et al., 2022; Dewolf and Willems, 2019; Donelan et al., 2002; Hiebert et al., 1996; Meurisse et al., 2019a). As a result, the lower-limb muscle activation pattern is pulsatile, with muscle activity occurring briefly at specific phases of the cycle to re-excite the intrinsic oscillations of the system when energy is lost (Ivanenko et al., 2004; Lacquaniti et al., 2012).

The head and trunk segments represent up to 65% of the body's mass (Dempster, 1955). Therefore, deviations from the erect posture, even small ones, have inevitable mechanical consequences. For instance, Müller et al. (2017) demonstrated that during walking with a flexed trunk, the CoM is displaced forward and slightly downward, with repercussions on stability (Aminiaghdam et al., 2017a, 2018). Also, the associated posterior shift of the hip relative to the CoM leads to a flatter angle at foot contact of the leading leg and a steeper trailing leg angle at toe-off and induces a modification of the two peaks of vertical ground reaction force (GRF) (Aminiaghdam et al., 2017a,b) and a modification of joint moments of the lower limbs (Leteneur et al., 2009), suggesting a modification of the step-to-step transition. These biomechanical modifications lead to alterations in the muscle activation patterns of the lower limb (Alghamdi and Preece, 2020; Grasso et al., 2000), with more crouched postures resulting in greater activation of both proximal and distal muscles, but with a stronger effect on the activation of the thigh and gluteal muscles compared with the shank muscles (Hora et al., 2024). Although these studies provided valuable insights, they did not explore the modification of the step-to-step strategy during the double support phase of walking. Understanding trunk control and its impact on gait is important for clinicians and scientists, as various pathological and ageing-related conditions tend to increase trunk inclination during walking. For example, there is a substantial body of research demonstrating increased trunk flexion during gait with ageing (Dewolf et al., 2021a, 2019b) and with pathologies such as Parkinson's disease (Pongmala et al., 2022) or knee osteoarthritis (Preece and Alghamdi, 2021). In older adults, there is a reduction in mechanical power generated by the plantar flexor muscles during the push-off phase of walking (Delabastita et al., 2021; Winter et al., 1990), an unanticipated step-to-step transition (the redirection of the CoM occurring after foot contact) (Dewolf et al., 2022; Meurisse et al., 2019a,b), and a widening of distal muscle activations (Dewolf et al., 2021a,b).

While the inter-relationship between posture and locomotion is well recognized, the isolated effects of trunk flexion on the neuromuscular control of gait are not well understood. In particular, it is not clear whether the trunk flexion and the associated modification of lower-limb kinematics affect the step-to-step transition of walking and, with it, the activation pattern of lower limb muscles observed in able-bodied human gait. To answer this question, we studied the effects of small (from 10 deg) and large (up to 40 deg) changes in upper body inclination in 20 young and healthy participants. The GRFs, the activity of five lower-limb muscles and the joint angles of lower limb segments were recorded. We hypothesized that trunk flexion would result in a more flexed limb posture during stance, a greater peak of force by the leading limb (and a smaller peak under the trailing one), and an anticipated activation of the distal extensors, consistent with the compensations observed in older adults.

Twenty healthy young adults (10 males and 10 females, age: 29.4±4.6 years, mass: 69.3±11.5 kg, height: 1.69±0.11 m, means±s.d.) volunteered to participate in this study. The sample size was estimated a priori based on the difference of 75% of the rate of F_back/F_front (η²=1.51) (Dewolf et al., 2022). Considering a power analysis of 0.85 (1−β) and one-way ANOVA with an alpha level of 0.05, 7 participants would be sufficient. The software used for the sample size was G-Power (v.3.1.9.7). All participants had no history of gait pathologies, neurological disease or orthopaedic problems that would affect how they walked. Informed consent was obtained from all participants, and the procedures used for this study were approved by the ethics committee of Finis Terrae University, Chile (ref. no. 22-114). The study followed the guidelines of the Declaration of Helsinki. Preliminary results have already been published in a conference paper (Nunez-Lisboa and Dewolf, 2023).

Participants were asked to walk wearing their shoes on an instrumented treadmill at 4 km h⁻¹. The fixed speed imposed in our study allows us to isolate the effect of trunk inclination, as previous research has shown that preferred walking speed can be influenced by trunk inclination (Aminiaghdam et al., 2018). The selected walking speed was reported as the comfortable and economical average walking speed based on the net energy consumption (Cavagna and Kaneko, 1977). Five different gait conditions were tested. Participants were first asked to walk normally, with a natural trunk orientation (called ‘normal walking’). Then, they were asked to walk with an anterior trunk flexion of 10, 20, 30 and 40 deg relative to the vertical. The motion capture system FreeMoCap v.1.0.25 (https://freemocap.org/) was used to track the estimated vertical and anteroposterior coordinates of the acromial apex, greater trochanter, lateral femoral condyle, lateral malleolus and fifth metatarsophalangeal joint at 60 Hz. The participants were filmed in the sagittal plane, and the trunk flexion, defined as the orientation of the line crossing the acromial apex and greater trochanter relative to the vertical axis, during the different gait conditions was controlled by a mobile app (OnForm, v.2.03.0). Before data collection began, participants took several trials with oral feedback from the experimenter until they adopted the correct position. Participants were instructed to bend the hips to achieve the target trunk flexion instead of rounding at the lower or upper back, following the methodology described by Kluger et al. (2014). No instruction was given regarding arm movements. The participants walked then for at least 1 min per condition, with 2 min of rest between each trial to avoid muscular fatigue. Data were recorded for 10 s once steady walking was reached and, on average, 7.3±0.5 (mean±s.d.) strides were analysed for each participant.

The treadmill (Stellar, h/p/cosmos, Nußdorf, Bayern, Germany; belt surface: 1.6×0.65 m; mass: ∼240 kg) was instrumented with four force transducers (Arsalis^®, Genappe, Belgium). As the transducers were placed under the body of the treadmill, the force transducers measured the three components of GRF exerted by the treadmill belt under the foot (Willems and Gosseye, 2013): F_v, F_f and F_l are, respectively, the vertical, fore aft and lateral component of the GRF. Data were sampled at a frequency of 500 Hz. Two algorithms were used to reconstruct the three components of the force under the left and right foot (Bastien et al., 2019; Meurisse et al., 2016).

Electromyographic (EMG) activities from five muscles of the right lower-limb were recorded at 2 kHz using a Delsys Trigno Wireless System (Boston, MA, USA): vastus medialis (VM), vastus lateralis (VL), tibialis anterior (TA), medial gastrocnemius (MG) and lateral gastrocnemius (LG). We selected these muscles to have a good comparison of activity between proximal and distal muscles of the lower limbs. The EMG activity of other relevant muscles, such as the biceps femoris, rectus femoris and gluteus maximus, has previously been reported in a similar task (Grasso et al., 2000). Furthermore, our focus was specifically on step-to-step transitions, where extensor muscles have been shown to significantly contribute to CoM propulsion and where the recorded muscles are particularly active (Dewolf et al., 2019a; Ivanenko et al., 2008). The location of the electrode was based on suggestions from SENIAM (the European project of surface EMG, seniam.org). The signal quality was verified visually by checking the EMG signals during voluntary contractions before walking. Acquisition of the EMG and GRF were synchronized with a trigger module from Delsys.

The foot contact (FC) and toe-off (TO) events were estimated from the displacement of the centre of pressure on the belt (Meurisse et al., 2016). A stride was defined as two successive right FCs. Stance phases were measured as the time between TO and FC of the same leg. The double contact phase (DC) started from the front leg FC to the back leg TO.

The vertical and antero-posterior coordinates of the head of the fifth metatarsal, the external malleolus, the lateral condyle of the knee, the greater trochanter and the apex of the shoulder of the right side of the body were recorded. The coordinates were filtered with a 4th order Butterworth filter, with a 7 Hz cut-off frequency. The angle of each segment (trunk, thigh, shank and foot) relative to the vertical (elevation angle) was computed in the sagittal plane. From these angles, the hip, knee and ankle joint angles were calculated. The hip, knee and ankle joint angles during standing were subtracted from those of the joint measured during walking tasks. Each stride of each participant was interpolated over 100 points. For the three joint angles and the trunk elevation angle, the range of motion (ROM) and the mean values were measured. Joint angles were measured relative to the subject's standing position, with 0 deg indicating this posture.

The fore–aft and vertical velocity of the CoM were determined from the fore–aft and vertical components of the GRF using the procedure described in detail in Dewolf et al. (2017). In short, the fore–aft acceleration of the CoM was calculated as a_f=F_f/m, where m is the participant's body mass. The vertical acceleration of the CoM was calculated as a_v=(F_v−mg)/m, where g is the acceleration due to gravity. The vertical (V_v) and the forward velocity (V_f) of the CoM were calculated by time-integration of a_v and a_f, respectively, plus an integration constant, which was computed so that the average velocity over a stride was equal to zero. The vertical and forward displacements of the CoM (S_v and S_f, respectively) were then computed by time-integration of V_v and V_f.

The time of occurrence of the minimum of forward (V_f,min) and vertical velocities of the CoM (V_v,min) at the beginning of the double contact phase were detected. The step-to-step transition strategy was defined as follows: the transition was considered as anticipated when V_v,min occurred before FC and unanticipated when V_v,min occurred after FC. Also, the peaks of vertical force under the front leg (F_front) and the back leg (F_back) were measured during the DC. In addition, the negative (F_f,brake) and positive (F_f,propulsion) peak of the left and right fore–aft force were measured.

The raw EMG signals were high-pass filtered (30 Hz), rectified and low-pass filtered with a zero-lag 3rd order Butterworth filter (10 Hz). The time scale was normalized by interpolating individual gait cycles over 400 points. For each condition and each stride, the full width at half maximum (FWHM) was calculated as the period during which the EMG activity exceeded half of its maximum (Martino et al., 2014). Each EMG waveform's centre of activity (CoA) was calculated as the vector's angle that points to the CoM of the circular distribution (Martino et al., 2014). Also, the mean activation was calculated.

For each dependent variable, the normality of residuals was visually assessed using Q–Q plots. A mixed-effects model with a post hoc Fisher LSD test was performed to evaluate differences between normal and the 10, 20, 30 and 40 deg conditions. Both numerator and denominator degrees of freedom (d.f.) were reported; the numerator d.f. reflect the number of comparisons, and the denominator d.f. adjust for the sample size and variability. In cases where sphericity was violated, the Greenhouse–Geisser correction was applied to adjust both d.f. (Wright and Wolfinger, 1997). Furthermore, the Friedman test with post hoc Dunn's test was used for dependent variables that did not meet the normality assumption. Cohen's d-effect size (ES) was calculated for each comparison, interpreted as 0.2 (small), 0.5 (medium) and 0.8 or greater (large). All statistical analyses were conducted using GraphPad Prism 9 software (GraphPad Software, San Diego, CA, USA) with an alpha level set at 0.05.

Fig. 1A,B illustrates the digital joint identification at the sagittal plane, the typical curves of the vertical and fore–aft forces of the back and front leg, and raw EMG data during one stride in all conditions. Fig. 2 demonstrates the average comparison of walking conditions with trunk flexion. The anterior trunk flexion during walking differed between conditions (F_2.857,50.72=252.1, P<0.0001; Fig. 2). Specifically, compared with normal walking, the degrees of anterior trunk flexion were higher at 10, 20, 30 and 40 deg (post hoc: P<0.0001, ES=4.31; P<0.0001, ES=7.61; P<0.0001, ES=6.44; P<0.0001, ES=8.55, respectively), and the average trunk elevation angle was close to (but smaller than) the angle requested to the participants (normal=−1.55±1.92 deg; 10 deg=7.5±2.2 deg; 20 deg=18.3±3.1 deg; 30 deg=24.5±5.3 deg; 40 deg=33.3±5.3 deg). The stride duration was affected by the trunk flexion (F_1.644,31.24=17.40, P<0.0001; Fig. 2). Specifically, compared with normal walking, the stride duration was lower at 10, 20, 30 and 40 deg (post hoc: P=0.045, ES=0.26; P=0.005, ES=0.30; P=0.001, ES=0.83; P<0.0001, ES=1.25, respectively). However, no difference was observed in the stance duration (F_2.439,45.72=1.865, P=0.158; Fig. 2). Specifically, compared with normal walking, the stance duration did not change at 10, 20, 30 and 40 deg (post hoc: P=0.938, ES=0.14; P=0.778, ES=0.12; P=0.734, ES=0.32; P=0.247, ES=0.59, respectively).

Fig. 3A illustrates the average waveform of the hip, knee and ankle joint angles during normal and flexed conditions over a stride. Notably, the joint angle's time course changed significantly as trunk flexion increased during the stride. Trunk inclination significantly affected the average angles of all lower-limb joints (hip: F_1.200,118.8=4404, P<0.0001; knee: F_1.058,104.8=124.6, P<0.0001; ankle: F_1.588,157.2=404.3, P<0.0001) (Fig. 3B). Furthermore, trunk inclination had a significant impact on the mean ROM of the hip, knee and ankle joint (hip: F_3.180,56.44=165.7, P<0.0001; knee: F_2.816,49.99=28.60, P<0.0001; ankle: F_2.776,49.27=15.48, P<0.0001) (Fig. 3B). Compared with the normal condition, the average joint angle of the hip was higher at 10, 20, 30 and 40 deg of trunk flexion (post hoc: P<0.0001, ES=2.05; P<0.0001, ES=4.17; P<0.0001, ES=5.36; P<0.0001, ES=7.34, respectively), while the average joint angle of the ankle and knee was higher at 20, 30 and 40 deg of trunk flexion (ankle post hoc: P=0.204, ES=0.37; P<0.0001, ES=1.14; P<0.0001, ES=1.68; P<0.0001, ES=2.21, respectively) (knee post hoc: P=0.080, ES=0.59; P<0.0001, ES=1.75; P<0.0001, ES=2.40; P<0.0001, ES=3.09, respectively).

Fig. 4A presents typical traces of V_v (top) and V_f (bottom) during one stride in all walking conditions. The black circles represent the minimum (V_v,min) and maximum (V_v,max) of V_v, which define the step-to-step transition. The black circles also indicate the minimum (V_f,min) and maximum (V_f,max) of V_f. With increasing trunk flexion angle, V_v,min occurred later in the stride (χ²=68.98, d.f.=4, P<0.0001). This indicates that the step-to-step transition was anticipated during normal walking and walking with a 10 deg flexion (V_v,min occurred before FC) and unanticipated at higher trunk flexion angles (V_v,min occurred after FC) (post hoc: 20 deg: P<0.0001, ES=1.64; 30 deg: P<0.0001, ES=2.47; 40 deg: P<0.0001, ES=2.88) (Fig. 4B). In a parallel observation, as the trunk flexion angle increased, V_f,min was noted to occur before heel strike (F_1.534,28.77=3.139, P=0.0703), coinciding with a monotonic change in the peak of vertical and fore–aft forces exerted by the back leg and by the front leg (Fig. 4C). With increasing trunk flexion, F_back decreased (F_2.165,41.13=84.31, P<0.0001). Compared with normal walking, the mean F_back decreased at 10, 20, 30 and 40 deg (post hoc: P<0.0001, ES=1.03; P<0.0001, ES=1.85; P<0.0001, ES=2.54; P<0.0001, ES=2.72, respectively). In contrast, F_front increased with increasing of trunk flexion (F_2.021,38.40=16.04, P<0.0001). The mean F_front increased at 20, 30 and 40 deg compared with normal walking (post hoc: P=0.002, ES=0.67; P<0.0001, ES=1.09; P=0.0004, ES=0.99, respectively) (Fig. 4C). Additionally, F_f,propulsive decreased with increasing trunk flexion (F_2.351,44.08=32.64, P<0.0001). Specifically, compared with normal walking, the mean F_f,propulsive decreased at 10, 20, 30 and 40 deg (post hoc: P=0.037, ES=0.42; P=0.001, ES=0.95; P<0.0001, ES=1.61; P<0.0001, ES=2.10, respectively). Conversely, F_f,brake was unaffected by trunk inclination (F_2.023,37.94=0.4022, P=0.674). Specifically, there were no significant changes in mean F_f,brake at 10, 20, 30 and 40 deg (post hoc: P=0.192, ES=0.21; P=0.278, ES=0.22; P=0.787, ES=0.07; P=0.609, ES=0.13, respectively) (Fig. 4C).

Fig. 5 demonstrates typical traces of the pendulum mechanism variables. Specifically, Fig. 5A illustrates a typical E_k, E_p and E_CoM trace during one stride under each condition. The CoM work (W_CoM) significantly differed between conditions (F_1.292,24.55=7.572, P=0.007, Fig. 5B). Compared with normal walking, W_CoM was lower at 10 and 20 deg (post hoc: P=0.0004, ES=0.91; P=0.006, ES=0.84, respectively) (Fig. 5B). Additionally, W_CoM did not change at 30 and 40 deg (P=0.415, ES=0.26; P=0.096, ES=0.58, respectively). The modification of W_CoM may partly be explained by the change of pendulum-like energy exchange (Fig. 5B). Indeed, the inclination of the trunk affects the %R (F_1.387,26.35=11.82, P=0.0008). As compared with normal walking, the %R was higher at 10 and 20 deg (post hoc: P=0.0004, ES=0.87; P=0.015, ES=0.81, respectively), with no changes at 30 deg (P=0.405, ES=0.39), and was significantly lower at 40 deg (post hoc: P=0.025, ES=0.84).

Fig. 6 presents the average (across strides and participants) rectified muscle activity of lower limb muscles (VM, VL, TA, GM and GL) during one stride in each condition. The angle of trunk flexion significantly influenced the mean activation of most muscles (VM: F_1.563,19.15=12.99, P=0.0006; VL: F_3.018,46.03=4.153, P=0.010; TA: F_1.530,16.06=8.180, P=0.0058; GL: F_1.619,23.87=5.733, P=0.013), except for GM, which was not significantly affected (F_1.787,22.79=1.583, P=0.227; Fig. 6B, top). Specifically, compared with normal walking, the mean activation of GM did not change at 10, 20, 30 and 40 deg (post hoc: P=0.335, ES=0.41; P=0.432, ES=0.33; P=0.954, ES=0.23; P=0.904, ES=0.21, respectively). Furthermore, compared with normal walking, the mean VM muscle activity increased at 10, 30 and 40 deg (post hoc: P=0.035, ES=0.16; P=0.005, ES=0.65; P<0.0001, ES=1.20, respectively). However, VM activity did not change at 20 deg of trunk flexion (post hoc: P=0.263, ES=0.24). Additionally, the mean VL muscle activity increased at 10 and 40 deg (post hoc: P=0.020, ES=0.31; P=0.031, ES=0.65, respectively) and did not change at 20 and 30 deg (post hoc: P=0.118, ES=0.40; P=0.189, ES=0.51, respectively). The mean TA activity was higher at 40 deg (post hoc: P=0.0174) and did not change at 10, 20 and 30 deg (post hoc: P=0.686, ES=0.05; P=0.352, ES=0.15; P=0.091, ES=0.38, respectively). Lastly, the mean activation of GL was higher at 30 and 40 deg (post hoc: P=0.039, ES=0.51; P=0.008, ES=0.65, respectively), and did not change at 10 and 20 deg of trunk flexion (post hoc: P=0.353, ES=0.10; P=0.114, ES=0.35).

In most muscles, the duration of muscle activation also varied with trunk flexion (Fig. 6B, middle). The trunk inclination affected the FWHM of VM, VL, GM and GL (VM: F_2.510,30.12=12.46, P<0.0001; VL: F_2.288,37.76=4.604, P=0.0129; GM: F_3.108,52.06=3.217, P=0.028; GL: F_2.991,56.07=5.515, P=0.002). In contrast, the FWHM of TA was not significantly affected by trunk angle (F_2.170,40.69=0.2317, P=0.811). Specifically, compared with normal walking, the mean FWHM of TA did not change at 10, 20, 30 and 40 deg of trunk flexion (post hoc: P=0.976, ES=0.0; P=0.993, ES=0.0; P=0.636, ES=0.11; P=0.670, ES=0.11, respectively). Furthermore, compared with normal walking, the FWHM of VM was higher at 10, 20, 30 and 40 deg of trunk flexion (post hoc: P=0.005, ES=0.47; P=0.004, ES=0.93; P=0.0007, ES=1.43; P=0.0001, ES=1.57, respectively). The FWHM of VL was higher at 40 deg of trunk flexion (post hoc: P=0.021, ES=0.89), and did not change at 10, 20 and 30 deg of trunk flexion (post hoc: P=0.656, ES=0.12; P=0.928, ES=0.03; P=0.361, ES=0.27, respectively). The FWHM of GM was higher at 20, 30 and 40 deg (post hoc: P=0.015, ES=0.95; P=0.017, ES=0.84; P=0.014, ES=0.81, respectively), and did not change at 10 deg (post hoc: P=0.147, ES=0.47). The FWHM of GL was higher at 20, 30 and 40 deg of trunk flexion (post hoc: P=0.016, ES=0.47; P=0.0004, ES=0.76; P=0.045, ES=0.44, respectively), and did not change at 10 deg of trunk flexion (post hoc: P=0.071, ES=0.23).

The trunk flexion also influenced the timing of activation of extensor muscles (VM: F_2.800,30.80=16.09, P<0.0001; VL: F_2.158,38.31=12.95, P<0.0001; GM: χ²=40.6, d.f.=4, P<0.0001; GL: F_2.272,30.67=19.07, P<0.0001) but not of flexor muscles (TA: F_2.881,54.01=0.3533, P=0.778) (Fig. 6B, bottom). Compared with normal walking, the activation of VM occurred earlier at 30 and 40 deg of flexion (post hoc: P=0.0003, ES=1.19; P=0.0008, ES=1.55) and did not change at 10 and 20 deg of flexion (post hoc: P=392, ES=0.11; P=0.889, ES=0.03, respectively). Furthermore, the activation of VL occurred later at 30 and 40 deg of flexion (post hoc: P=0.0006, ES=1.29; P=0.0031, ES=1.41, respectively) and did not change at 10 and 20 deg of trunk flexion (post hoc: P=0.392, ES=0.05; P=0.344, ES=0.67, respectively). For distal extensors, the changes were much larger: compared with normal walking, the CoA of GM and GL occurred earlier at all trunk flexion angles (GM: 10 deg: P=0.021, ES=65; 20 deg: P=0.0001, ES=1.15; 30 deg: P<0.0001, ES=1,30; 40 deg: P<0.0001, ES=1.69; GL: 10 deg: P=0.0002, ES=0.88; 20 deg: P=0.0002, ES=1.29; 30 deg: P=0.0003, ES=2.25; 40 deg: P<0.0001, ES=2.20).

Fig. 7A presents the mean activation of the extensor muscles (VM, GM and GL) during the initial and final 20% of the stance phase. The initial 20% of the stance phase was influenced by trunk flexion in most muscles (VM; F_2.646,28.44=18.70, P<0.0001; GM: F_2.586,34.27=4.714, P=0.009; GL: F_2.900,52.92=4.928, P=0.004). Compared with the normal condition, the mean activation of VM during this phase was higher at 10, 30 and 40 deg of trunk flexion (post hoc: P=0.040, ES=1.21; P=0.0006, ES=4.15; P=0.0003, ES=5.52, respectively) and did not change at 10 deg of trunk flexion (post hoc: P=0.096, ES=0.99). In contrast, VL was not significantly affected (F_1.878,26.30=1.150, P=0.329). Compared with normal walking, VL did not change at 10, 20, 30 and 40 deg of trunk flexion (post hoc: P=0.921, ES=0.02; P=0.095, ES=0.32; P=0.301, ES=0.15; P=0.195, ES=0.21, respectively). The mean activation of GM was higher at 20 and 30 deg (post hoc: P=0.032, ES=0.86; P=0.016, ES=0.91, respectively) and did not change at 10 and 40 deg of trunk flexion (post hoc: P=0.188, ES=0.48; P=0.127, ES=0.73, respectively). The mean activation of GL was higher at 30 and 40 deg (post hoc: P=0.047, ES=0.65; P=0.016, ES=0.77, respectively) and did not change at 10 and 20 deg (post hoc: P=0.412, ES=0.19; P=0.084, ES=0.54, respectively).

Regarding the final 20% of the stance phase, discernible alterations were observed in VM (F_2.630,32.21=3.016, P=0.0501). Specifically, VM remained significantly unaltered at 10, 20, 30 and 40 deg of trunk flexion (post hoc: P=0.622, ES=0.19; P=0.695, ES=0.17; P=0.114, ES=0.81; P=0.061, ES=0.86, respectively) (Fig. 7A). Similarly, VL did not show significant changes (F_{0.06431,0.9325}=1.210, P=0.1214). Compared with normal walking, VL did not report significant changes (post hoc: P=0.884, ES=0.03; P=0.756, ES=0.06; P=0.142, ES=0.42; P=0.161, ES=0.39, respectively). Furthermore, the mean activation of GM did not present significant changes (χ²=0.5, d.f.=4, P=0.969). Specifically, GM did not change at 10, 20, 30 and 40 deg of trunk flexion (post hoc: P=0.932, ES=0.06; P=0.682, ES=0.01; P=0.969, ES=0.14; P=0.622, ES=0.14, respectively). In contrast, GL showed significant changes (χ²: 12.33, d.f.=4, P=0.015), increasing at 40 deg of trunk flexion (post hoc: P=0.017, ES=0.64) but with no change at 10, 20 and 30 deg of trunk flexion (post hoc: P=0.656, ES=0.14; P=0.806, ES=0.08; P=0.066, ES=0.55, respectively).

This study aimed to determine the effect of trunk flexion on the neuromuscular control of gait. Specifically, we analysed the effect of trunk flexion on lower-limb kinematics, the step-to-step transition of walking and, with it, its impact on lower-limb muscle activation patterns during walking. The results are discussed in the context of the biomechanical principles underlying foot-support interactions during step-to-step transitions in bipedal upright walking.

Given that the trunk and head segment constitute approximately 65% of the body mass (Dempster, 1955), walking with the trunk anteriorly flexed modifies the position of the CoM, which is displaced forward and slightly downward (Müller et al., 2017). As a result, even a small change in trunk orientation can influence the mechanical demands of walking (Figs 3 and 4; Aminiaghdam et al., 2017b; Hora et al., 2024; Müller et al., 2017), in turn affecting the neuromechanical control of locomotion. As illustrated in Fig. 7B, trunk flexion alters the spatial relationship between the CoM and the FC point of the back leg at the end of the single stance (Warrener et al., 2021), resulting in different leg orientation (Aminiaghdam et al., 2017b). This results in a reduction of the vertical and fore–aft forces under the back leg during propulsion (Figs 4C and 7B; Aminiaghdam et al., 2017b; Grasso et al., 2000). Additionally, the GRF is more vertical at FC because F_front increases at weight acceptance without changes in F_f,brake (Fig. 4C). Similar modifications of vertical impulses have been observed in older adults (Delabastita et al., 2021; Dewolf et al., 2022, 2019b) and in trunk flexion-induced gait analysis (Müller et al., 2017). We documented here that it has a clear impact on step-to-step transition: an unanticipated transition was revealed at 20, 30 and 40 deg of trunk flexion (Figs 1B, 4B and 7B), accompanied by a reduction in the mechanical recovery of walking at 40 deg of trunk flexion (Fig. 5B). The unanticipated transition occurs when V_v,min appears after FC. This absence of modification of transition at 10 deg (and in turn of mechanical work; Fig. 5B) may be due to the minimal changes in the energy fluctuations (Fig. 5A). Instead, with a greater trunk inclination, the vertical displacement of the CoM starts to increase and the modification of the ratio between potential and kinetic work over one stride (W_p/W_k) induces a reduction of recovery between energies (Cavagna et al., 1976a,b). The detailed results of W_p/W_k are shown in Fig. S1.

As for the modification of muscle activation during the stride (Figs 1 and 6), it is noteworthy that with increasing trunk flexion, the activation of extensor muscles generally increases, except for GL and GM (Fig. 6B). A similar result has also been described recently in crouched walking (Hora et al., 2024), showing lower increases of activation in the lower leg muscles with trunk inclination compared with the thigh muscles. As suggested by Hora et al. (2024), the lower increase in activation of ankle muscles might reflect the smaller changes in ankle joint kinematics compared with the hip and knee joints (Fig. 3A,B). When specifically examining the activation of the extensor muscles in the front leg, VM increases its activity mainly during the initial 20% of the stance phase (Fig. 7A), confirming the greater contribution of VM during weight acceptance. However, during the initial 20% of stance, GM and GL also increase with increased trunk flexion, suggesting an antigravitational role (Hamner et al., 2010; Honeine et al., 2013) when the collision with the ground of the front leg becomes more important. This increase in distal extensor activation is related to their earlier activation (CoA; Fig. 6B, bottom) during the walking cycle. In addition, we noticed a change in the period during which muscle activity remains consistently above half of its maximum level (Fig. 6B). Here, we evidenced a higher FWHM with an increasing trunk inclination (Fig. 6B, middle) in proximal (VM and VL) and distal muscles (GM and GL). These prolonged activations probably suggest the need for enhanced stability and control as the body adapts to the forward trunk inclination. Similar compensatory mechanisms are observed in the gait pattern of patients with balance deficit, such as cerebellar ataxia (Martino et al., 2014). In summary, when the lower limbs are more flexed and the CoM is shifted forward and downward, increased muscle activity is required to generate the necessary joint torque to accomplish the walking task. This is also related to the more flexed lower-limb joint during stance (Fig. 3A,B; Aminiaghdam et al., 2017b). Indeed, as trunk flexion increases, the participants exhibit a linear increase in hip, knee and ankle (dorsi-) flexion, while the ROM of these joints remains consistent throughout the step (Fig. 3A,B). This most likely induces muscle compensation reminiscent of crouched walking (Grasso et al., 2000). Indeed, Grasso et al. (2000) observed that imposing greater knee flexion during stance leads to a higher muscle activation of lower-limb muscle. As for walking with a flexed trunk, the increment of muscle activity can be related, among others, to the lower mechanical energy recovery through the pendular energy exchange (see figure 7 in Li et al., 1996). Additionally, Grasso et al. (2000) also observed that the time sequence of activation of different muscles varied substantially as a function of imposed posture. In particular, as in the present study, a greater activation of the knee extensor (VL) was observed at the end of the stance phase, and a greater activation of the ankle extensor (GM) at the beginning of the stance phase, consistent with our findings. In contrast, our study found greater activation of the knee extensor (VM) at the beginning of the stance phase (Fig. 7A).

This suggests that trunk inclination and the lower-limb kinematics change in parallel and cannot be studied separately. Similar observations were previously made by Grasso et al. (1999), who showed that in Parkinson's disease patients in the apomorphine OFF condition, the flexion of the trunk is paralleled by the flexion of the legs. In the ON condition, the trunk and the limbs extend together. These authors proposed that the basal-ganglia circuitry may provide a spatiotemporal framework for the control of both posture and movement in a gravity-based frame.

Also, a recent study (Mesquita et al., 2023) suggests that altering the foot interaction with the ground affects the walking pattern similarly to trunk flexion. By imposing different foot positions during stance (e.g. maintaining a flat-foot during the whole stance phase), the authors described that deviating from the classical heel-to-toe pattern of walking leads to a higher F_front, a smaller F_back, an anticipated activation of distal extensor muscles and a higher activation of proximal muscles. They argued that the specific use of the heel-to-toe rolling pattern allows delay of the activation of distal extensors relative to proximal extensors, ensuring a smooth step-to-step transition. This is a specific adaptation of the locomotor apparatus to bipedalism. Indeed, in quadrupeds, the step-to-step transition is not as crucial as in bipeds because the mechanical cost of the redirection is reduced in proportion to the number of contacts used to achieve a given redirection of the CoM (Ruina et al., 2005). In addition, the change of the hindlimb configuration at the end of stance (knee flexed) may allow them to propel themselves forward using primarily the proximal extensors. Interestingly, deviating from the erect posture of the trunk, another specific adaptation of the locomotor apparatus to bipedalism, results in similar modifications of walking pattern.

These indications provide arguments to the fact that the erect posture has evolved to optimize gait, allowing the CoM to vault over the supporting straight limb. Indeed, based on the fluctuation of the mechanical energy of the CoM, normal walking is often compared to an inverted pendulum model based on the fluctuation of the mechanical energy of the CoM (Cavagna and Kaneko, 1977; Willems et al., 1995). Maintaining a flexed trunk during walking potentially affects this pendulum-like mechanism as it modifies the position of the CoM and the limb posture during stance. Here, when comparing normal walking and walking with a 40 deg trunk inclination, we found a significant decrease of the energy transduction between E_p and E_k (%R) and with it a medium effect (ES=0.56) on the mechanical work done to move the CoM (W_CoM) (Fig. 5B). Furthermore, the pattern of walking with imposed trunk flexion is more similar to that of bipedal chimpanzee gait than to normal walking. Indeed, the kinematics and mechanics of walking with imposed trunk flexion overlap widely with the gait pattern of chimpanzee bipedal walking. Chimpanzees walk with a crouched posture with flexed limbs (Johnson et al., 2022) and similar CoM mechanics and recovery rates can be observed for chimpanzee bipedal walking (Demes et al., 2015). Understanding the link between posture and gait mechanics across different species may help researchers better interpret the effects of evolution on bipedal walking.

More surprisingly, the opposite was observed when comparing normal with walking with a 10 and 20 deg flexed trunk, i.e. an increase in %R and a decrease in W_CoM. This last result suggests that the recovery of energy only gives partial information concerning the metabolic cost of walking as the collisional loss occurring at contact is more likely to be a better determinant of the cost of locomotion. Nevertheless, it allows a better understanding of the mechanical–bioenergetic interaction of walking. In general, humans tend to prefer economical walking patterns. However, Hunter et al. (2010) showed that when walking on a shallow downhill slope, their participants did not take optimal advantage of the propulsion provided by gravity to decrease energetic costs, but instead preferred a more stable and costlier gait pattern. Here, it could be that, because of the shift in the hip relative to the foot, a 10 and 20 deg trunk flexion allowed the participant to take advantage of the propulsion provided by gravity and thus to reduced the mechanical cost of walking at the expense of stability. When the trunk becomes more flexed, it can be observed that E_k and E_p become more in phase after foot contact (Fig. 5A). This suggests that the pendulum-like exchange is progressively reduced, and other mechanisms such as elastic energy storage potentially come into play (Dewolf et al., 2016).

In conclusion, this research sheds light on the impact of trunk flexion on gait dynamics and its neuromuscular control. The study reveals that even subtle changes in trunk orientation, as observed in older adults (Dewolf et al., 2022, 2021a), lead to adjustments in lower limb kinematics, step-to-step transitions, GRFs and lower-limb muscle activation patterns (Fig. 6B). Notably, increased trunk flexion introduces mechanical challenges as a result of the modification of CoM position (Müller et al., 2017). These findings underline the complex interplay between mechanics and neural control of gait, enhancing our understanding of human bipedal walking and its biomechanical underpinnings.

We express our appreciation to R. Vargas-Foitzick for their insightful conversation, which sparked the initial idea for this research.