The sample size was estimated using G*power software, which determined that an effect size of 0.5 and a statistical power of 80% (α = 0.05) are necessary to detect a significant difference (Zhang et al., 2002), requiring at least 51 trials in each group. The study finally recruited 128 elderly participants, with ineligible individuals, dropouts and invalid data excluded, included 36 participants, each of them completed five trials, resulting in 166 trials, which satisfied the requirement for statistical significance.

The inclusion criteria for the participants were as follows: female, right-leg dominance; good physical health; absence of back and pelvic system issues, neuromuscular disorders or balance impairments; and normal cognitive function. The sedentary group (SG) was defined as those who reported no exercise habits and an average sedentary time greater than 6 h per day in the past 7 days, and the Tai Chi group (TC) had more than 6 years of training.

Participants with a Mini-mental State Examination (MMSE) functional score of 24, frailty, lower limb injuries or surgeries within the past 6 months, a history of falls in the previous 2 years and fatigue or poor condition prior to the experiment were excluded.

All participants were informed about the specific test procedures and signed written informed consent forms. The study was approved by Henan Sports Science and Technology Centre (Henan Anti-Doping Centre), China.

Kinematic data were obtained using eight 3D near-infrared high-speed cameras (Qualisys, Sweden, model: Oqus 600) with a sampling frequency of 100 Hz. Fifty-two infrared reflective marker points (diameter of 14 mm) were attached to the participants to collect GI kinematic information. The placement method of the markers follows the Helen Hayes model. Kinetic testing was conducted using four Kistler force plates (90 cm × 60 cm ×10 cm, Switzerland, model: 9287C), with a sampling frequency of 1,000 Hz. A remote-controlled start signal lamp was self-constructed and used.

Participants were instructed to refrain from engaging in strenuous physical activities for 3 days prior to the experiment and to avoid consuming stimulating beverages, such as alcohol or coffee, which might affect the nervous system on the day before the experiment. Sufficient sleep was ensured. On the day of formal testing, participants were required to familiarize themselves with the experimental procedures and exercise them thoroughly. After the experimental preparations were completed, the participants performed a 5-min warm-up activity before proceeding to the formal testing. The starting action involved participants standing on a force plate with standardised footwear and clothing, with each foot placed on a force plate. They were instructed to stand naturally in a static position with their feet shoulder-width apart. The signal indicator is located at the endpoint, which is opposite to the walking direction of the subject 1.2 m above the horizontal plane and 6 m away from the starting position of GI.

The GI walking test was performed under the experimental condition designed in this study. The participants stood naturally on the force plate for 3–5 s and then initiated a step forward voluntarily. Trials started with the subject in quiet standing, with their feet placed in their preferred natural stance. Following a verbal cue of ‘anytime’, the subjects waited for a self-selected time interval (no less than 3 s) then initiated forward stepping at their self-chosen pace. Five valid trials were collected and reviewed online to ensure a steady-state baseline before the ‘anytime’ instruction. Trials without a period of steady-state standing were discarded and repeated.

Data of gait evaluation were analysed using Visual3D software (C-Motion, USA, v6.01.36). Kinematic and kinetic data were filtered and denoised by using a lowpass filter, with cut-off frequencies of 6 Hz for kinematics and 20 Hz for kinetics.

The GI process is divided into two phases: the APA phase and the locomotion (LOC) phase. This study focused on analysing the stability of the LOC phase. The MoS at the onset of LOC as well as the forward and lateral control stability at the end of LOC were computed.

Next to the position and the velocity of the CoM, we can introduce the ‘extrapolated centre of mass (XcoM)’, as shown in Eqs 1, 2, where ω₀ is a constant related to stature. XcoM allows the requirements for stable walking to be formulated simply based on the inverted pendulum model of balance. ω 0 = g / l (1) XcoM = CoM + V C o M ω 0 (2)

Where XcoM represents the XcoM position, V C o M is the CoM velocity and normalised by the eigenfrequency ω 0 , l is the leg length measured from the lateral malleolus to the greater trochanter and g represents the gravity acceleration (Hof et al., 2005).

The minimum M-L margin of stability MoS_ML and A-P margin of stability MoS_AP were computed as follows in Eqs 3, 4: MoS ML = BoS x − XcoM x (3) MoS AP = BoS y − XcoM y (4)

MoS_ML and MoS_AP were calculated as the distance between XcoM and the boundaries of the base of support (BoS) during the single stance from the foot markers, namely, the fifth metatarsal–phalangeal joint for the lateral border and the medial malleolus for the medial border (Hof et al., 2005; Osada et al., 2022), as illustrated in Figure 1.

We calculated lateral stability with a constant offset control by positioning the right foot B_ML to the right of the XcoM, which is related to the desired step width SW in Eq. 5 (Hof, 2008). B M L = S W e ω 0 t + 1 (5) Where B M L is the stability of lateral control, S W is the step width and t is the time of LOC.

The simplest stable control of CoP position could be made by positioning the CoP at a constant distance behind the ‘constant offset control’ of XcoM. The offset value B_AP is represented as a constant distance. The equation is as follows in Eq. 6 (Hof, 2008). B A P = S L e ω 0 t − 1 (6) Where B A P is the stability of forward control, S L is the step length and t is the time of LOC.

Foot placement was determined based on step length, step width and foot inclination angle, as illustrated in Figure 1.

Statistical analysis was performed using SPSS software (version 22, SPSS Inc., Chicago, IL USA). Shapiro–Wilk test was used to verify the normal distribution of parameters; if the parameters were normally distributed, the Independent Samples t-test was applied to compare data between SG and TC groups, if the parameters were not normally distributed, the Mann–Whitney U test was applied. Effects were considered to be significant at p < 0.05.

The study included 17 older females who did not engage in regular exercise (age: 65.59 ± 3.66 years; height: 1.62 ± 0.53 m; mass: 65.41 ± 12.25 kg; BMI: 26.04 ± 7.79; MMSE scores: 28.7 ± 0.40) and 19 older females who participated in long-term Tai Chi exercise (age: 65.58 ± 3.63 years; height: 1.58 ± 0.43 m; mass: 60.94 ± 7.27 kg; BMI: 24.32 ± 2.72; MMSE scores: 28.9 ± 0.49). Table 1 provides detailed information about these individuals.

Table 2 indicates statistically significant differences in lateral stability between the SG and TC groups at the beginning of the single-support phase during the transition from double to single support during GI. The MoS_ML result in SG is significantly lower than that in the TC group (p = 0.19, η ² = 0.342, 95%CI: −0.011, −0.001). M-L and A-P stability also show statistically significant differences at the end of GI when the swing leg touches the ground and shifts from single to double support. The B_ML (p = 0.002, η ² = 0.455, 95%CI: −0.013, −0.003) and B_AP (p = 0.001, η ² = 0.473, 95%CI: −0.009, −0.002) results in SG are significantly lower than those in the TC group.

Table 3 shows that the TC group has a greater step length, step width and step speed than SG at the end of the GI process when the swing leg heel makes contact with the ground and transitions from single to double support. The statistical analysis reveals significant differences between the two groups, with a more pronounced difference observed in SL and SW. The foot inclination angle (SG: 33.097 ± 29.219; TC: 31.333 ± 20.671, p = 0.821, η ² = 0.352, 95% CI: 13.893, 17.42) is higher in SG than in the TC group.

GI involves internal self-perturbations of balance, known as endogenous ‘action equilibration’ (Bouisset and Do, 2008). MoS, derived from dynamic stability theory and the human inverted pendulum model, is a measure of walking stability by representing the shortest distance between a given CoM position–velocity state point and the backward imbalance boundary. Gait stability can be evaluated by calculating stability margins in the A-P and M-L directions, with the M-L dynamic MoS as a key indicator of walking stability.

The results show a statistically significant difference in M-L stability between SG and TC groups at the moment the swing leg is lifted off the ground during the transition from double-to single-leg support. M-L stability is an important predictor of fall risk, and the results suggest that sedentary older women are potentially at risk of falling. Nakano et al. also suggested that age-related changes in M-L balance control may increase the risk of lateral falls during GI (Nakano et al., 2016). Breniere and Do. (1986) reported that steady-state gait conditions are achieved at the end of the first step when walking at a habitual speed. The goal of GI is to place participants under steady-state gait conditions during the first step. At the end of the first step, the M-L and A-P stability in SG are significantly lower than those in TC, indicating that Tai Chi significantly improves the stepping gait stability during GI. At the onset of GI, sufficient APAs will ensure that the CoM is repositioned over the new support base (standing/supporting leg), preventing the body from tilting towards the swing leg after transitioning from bipedal to unipedal stance. If APAs fail to effectively reposition the CoM onto the new support surface, then walkers will need to rely on CPAs to reduce the effect of this instability, typically by taking more side steps with the stepping foot to increase the range of the support surface (Zettel et al., 2002). Some adaptation effects are consistent with the role of motor predictions in coupling posture and movement. The CNS anticipates the direction and magnitude of the initial lateral body movement, corresponding to the expected position of the M-L foot placement. Foot placement is adapted by altering posture and movement before stepping to ensure stability during stepping. These adaptations during movement may directly alter stepping performance. Disruptions in frontal plane postural control prompt movement adjustments to maintain stability and efficiency (e.g., maintaining step length) during the first step (Mille et al., 2007).

This study compares the displacement results of the CoP during GI at different angles (45° left anterior, 45° right anterior and 90° right lateral). A 16-week Tai Chi intervention significantly increases the displacement and speed of lateral steps in sedentary older adults. Tai Chi may improve dynamic postural control during the initial phase of GI in older adults by altering the movement characteristics of the CoP. Another study revealed that a 12-week Tai Chi exercise can improve posture control on unstable platforms and reduce the risk of falls in older individuals with sarcopenia (Huang et al., 2023). However, a previous research that compared the minimum postural sway ability of older adults suggested that a 15-week Tai Chi intervention did not improve measures of platform postural stability (Wolf et al., 1997). Our study includes stability calculations and reports that during the end stage of step advancement, the M-L and A-P stability of older women in the TC is significantly higher than that in sedentary older adults. Hence, Tai Chi exercise has a positive effect on improving the stability of GI in older adults. The different results of Tai Chi in improving postural stability may be due to the specific movement characteristics of Tai Chi, such as stable shifts in the centre of gravity (CoG) and slow, smooth transitions in step movements, which may manifest stability in controlling step movements.

A potential method used to improve postural stability during GI is modifying the initial foot position (IFP). The human body achieves safe and efficient step advancement by adjusting the step length and position of the swing leg foot at ground contact. Considering the significance of appropriate IFP for GI and recovery response, the observed changes in step length during the first step of older fallers may serve as an important predictor of posture-related problems (Azizah Mbourou et al., 2003). During GI, coarse balance control is achieved through the placement of the swing leg’s foot (Jian et al., 1993). This study found statistically significant differences in the step length, width and speed of older females in the SG and TC group, with SG exhibiting significantly shorter step length and width. These findings align with previous research, which indicated that step length and step speed significantly decrease with age during GI (Muir et al., 2014) and that older fallers demonstrate significantly shorter first steps (Azizah Mbourou et al., 2003). Additionally, older adults can modulate their step width to maintain dynamic balance in the M-L direction (Nakano et al., 2016). Muir et al. (2014) compared age-related changes in the first four steps of gait, including the GI phase, for three age groups (20–25 years, 65–79 years and 80–91 years) walking at self-selected speeds. The results showed significant decreases in step speed and length across age groups. Higher variability in step length and step width may reflect increased foot placement errors and/or decreased CoG control among older adults. Healthy individuals aged 80–91 years walked at a slower pace with shorter steps. Furthermore, older fallers had significantly shorter first steps during GI and significantly longer double support time (Azizah Mbourou et al., 2003). Our results further confirm that older adults may utilise their foot placement at touchdown to regulate stable control at the end of the stance phase. A faster gait requires longer anticipation time and shorter execution time (Brenière et al., 1987). No statistically significant differences were observed in the foot inclination angle between the two groups. Inclination of the foot to the outer side (IFP) may be a feasible option to increase ankle muscle activation and improve age-related decline in postural stability (Jeon et al., 2021). Reinmann et al. (Reimann et al., 2017) proposed that foot placement and ankle lateral strategies may be two independent mechanisms that are coupled and coordinated in time to compensate for unstable M-L APAs and stiffened stance leg in the frontal plane to maintain body sagittal stability (Van Dieёn et al., 2008). The foot placement position is actively regulated by the CNS to maintain lateral balance (Bauby and Kuo, 2000; Donelan et al., 2004). The human body maintains gait stability by coordinating the M-L foot placement position and the control of the CoM (O’Connor and Kuo, 2009). The CNS anticipates the initial lateral displacement of the body and adjusts the foot placement accordingly to ensure its alignment with the expected position. The foot position is adapted by altering the pre-step posture to ensure stability during foot strikes (Collins and Kuo, 2013). Even if the foot placement position is slightly inappropriate at foot strike, it can still be compensated by adjusting the M-L position of the CoP, which is known as the ‘ankle lateral strategy’ (Hof et al., 2007). Changes in the length of the first step may be an important predictor of postural problems in older adults considering the importance of appropriate foot placement for GI and recovery responses.

This study compared the GI stability characteristics and foot placement features between two representative older groups, namely, sedentary older women and older women with long-term regular Tai Chi exercise. The results provide a reference for future studies on GI stability in specific population subgroups. However, this study did not investigate the coordination of limb spatial positioning and ground reaction forces. In future research, we will consider the coordination of GI and potential differences in force control that may lead to an increased risk of injury. Previous studies indicated that gender is a major risk factor for falls, with older women being more prone to falls. The present work only included older women participants and did not further categorize them by age group. Further studies will consider additional age groups and physical activity groups for research purposes.