Participants

Seventeen healthy participants (8 females and 9 males) who were all recreational runners took part in the study. The demographic information of participants is summarised in Table 1. As inclusion criteria, participants needed to be able to walk without a walking aid and to be able to understand Cantonese or English instructions. Exclusion criteria included any history of musculoskeletal injuries or surgeries in the past 6 months, neurological disorders, cardiopulmonary conditions, or open wounds. Participants who were professional running athletes (International or Olympic group/ a minimum of 6 training sessions per week), with hydrophobia, and/or obesity were also excluded from the study. All participants were informed about the purposes, procedures and potential risks of the study before data collection. Then, the written informed consent forms were obtained from them. Before the experiment, participants were asked to fill in the Physical Activity Readiness Questionnaire (PARQ) to ensure their capability of participating in this research.

Design and Procedures

This was a cross-sectional study design. Participants were asked to perform 2-min running sessions using a motorized treadmill on land, and in three different water depths: 1) waist level (anterior superior iliac spine), 2) mid-thigh level, and 3) mid-shin level. The lateral view and the posterior view of running action were videotaped using an iPhone 11 at 30 frames/s to qualitatively view the gait. The recording devices were placed 1m laterally and 10m posteriorly away from participants, positioned at the knee crease level to prevent angulation of the video. The muscle activities of the rectus femoris, tibialis anterior, biceps femoris, and medial gastrocnemius of both legs during running were recorded using a 16-channel sEMG system (aktos surface EMG transmitters, Myon, Switzerland) and a customized data logger at a 2000Hz sampling rate. The sEMG signals were exported using EMGandMotionTools Version 6.0.4.0 (Cometa, Milan, Italy). The sagittal joint angles were recorded using the Inertial measurement unit (IMU) system (aktos-t 3-axial sensors, Myon, Switzerland) with a sampling rate of 286Hz and a proprietary bidirectional 2.4GHz transmission protocol.

Data collection procedures began with the measurement of anthropometric data as follows: (1) body weight using an electronic scale (BC-730b, Tanita, Japan), (2) body height using a measuring tape, and (3) blood pressure using a sphygmomanometer (omron HEM-7211). The BMI was determined through the calculation of participants’ height and weight.

After the anthropometric measurements, the electrodes were attached to the skin of each individual to record the rectus femoris (RF), biceps femoris (BF), tibialis anterior (TA) and medial gastrocnemius (mGas) muscle activity according the SENIAM recommendations (Hermens et al., 2000) after standard skin preparation procedures by shaving the hair around and using alcohol swab to reduce skin impedance. These muscles were selected based on their role in the gait. Waterproof dressing 3M™ Tegaderm™ Transparent Film Roll 16004 was applied over the electrodes to reduce the influence of water on EMG signals and for waterproof protection for the electrodes (Carvalho et al., 2010).

Then, each participant performed the muscle isometric maximal voluntary contraction (MVC) for subsequent EMG data normalization. During the MVC task, participants increased the force gradually till reaching the maximum effort after 3-5s, then held it for 3s and relaxed for another 3s. Two repetitions of MVC trials were performed with a 60-s rest interval between the two consecutive trials. MVC testing procedures are listed in Table 2.

Afterwards, the inertial measurement unit (IMU) sensors were attached to the participant’s body (mid-thigh, mid-shin, and dorsum of feet) according to the altos-t system and covered by the same waterproof dressing, followed by IMU calibration.

Participants were asked to enter the motor driven treadmill (LIFESTYLE The Ultimate Hydrotherapy System, HYDRO PHYSIO^®□) after being instrumented with sEMG and IMUs. Each participant underwent an adaptation period in order to get used to the equipment and data collection. All participants had no prior experience with ATR. Subsequently, a 1-min practice trial with 110 bpm by a metronome app (soundbrenner) was used during both land treadmill testing and aquatic treadmill testing at the waist level to familiarise participants with running (Jung et al., 2019). No horizontal water flow or artificial current was used during ATR.

In the experiment, participants completed 2-min test trials, with hands off the rails, on land, and aquatic treadmill adjusted to three water levels (with participants in standing): waist (ASIS), mid-thigh and mid-shin level. The test speed was selected based upon the speed that elicited 110 steps per min paced by the same metronome app. Meanwhile during running, sEMG and gait kinematics were collected and analysed. A 5-min rest interval was provided after each test trial to minimize fatigue.

The Borg rating of perceived exertion (RPE) (0-10 scale) (Health, 1998) was measured every minute to record the effort and the level of fatigue during the study.

Statistical Analysis

Raw EMG signals were processed with a bandpass filter (cutoff frequencies between 20 and 300 Hz) and a root-mean-square sliding window (50ms time constant) (MatLab2017a; Mathematical computing software, Natick, MA, USA), with the left peak hip flexion angle used to divide the period of each gait cycle. After initial 10 gait cycles, amplitude of EMG signals for the muscles from 10 gait cycles was selected and averaged. Similarly, raw MVC data were filtered and smoothed with the same method. The largest mean MVC value during contraction was selected as the MVC value for each muscle. Then, mean sEMG data for the 10 gait cycles were normalised to these MVC values and expressed as %MVC. sEMG activities during the stance phase and the swing phase were analysed separately. As for kinematic data, the peak hip flexion angle was analysed using IMU to indicate initial contact as the start of the stance phase. Stance and swing phases were differentiated using the peak hip flexion angle and the peak hip extension angle for MVC data as well.

Muscle activity and kinematic statistical calculations were computed using Matlab 2017a and SPSS 21, respectively. Descriptive data were calculated for demographic data (age, body height, body weight and BMI) as well, moreover, %MVC for each muscle per water depth condition was calculated. The data are expressed as means ± standard deviations. Total mean of sEMG activities in terms of %MVC was analyzed using a 2 (leg side: left and right) x 2 (task type: stance and swing) x 4 (muscle type: RF, BF, TA and mGas) x 4 (treadmill running condition: land, mid-shin, mid-thigh and waist level) repeated measures ANOVA. The alpha level was set at 0.05 for all statistical computation. Post-hoc comparisons with least significant difference procedures were used to determine any significant differences between the specific levels of the factors.

For gait kinematics, one-way ANOVA was used to compare the peak angle of different joints and the swing/stance ratio across different water depths. The alpha level was set at 0.05 for all statistical computation. Post-hoc comparisons with Fisher's least significant differences were used to identify any significant interactions and differences between the specific levels of water depths.

Muscle activity response

%MVC of RF, TA, BF and mGas during different phases are presented in Table 3. The results of repeated measures ANOVA showed that the muscle type (F(3,57)=8.607, p<0.001), task type (F(1,19)=18.151, p<0.001), and treadmill running condition (F(3,57)=8.080, p<0.001) were significant factors that influenced %MVC. However, the leg side was not the case. Also, there was a significant interaction between the muscle type and the task type (F(3,57)=28.320, p<0.001) in %MVC.

In the stance phase, for both right and left sides, there was no significant difference in %MVC of RF, TA and mGas across the land and three different water depths in spite of an increasing trend of muscle activity with an increasing water level (p>0.05). The difference in %MVC of BF was significant across the land and three different water depths for the left side (p=0.008), but not significant for the right side (p=.085). As reflected from the post-hoc test, significant differences were present when comparing land and mid-shin (p=0.015), mid-thigh (p = 0.018) and waist (p=0.011) levels, respectively.

During the swing phase, both right and left sides, RF showed a significant difference in %MVC between water and land conditions (p<0.05). Post-hoc testing revealed significant differences on the right side when comparing land and mid-shin levels (p 0.001), the mid-thigh level (p= 0.001), and the waist level (p=0.003), also between mid-thigh and mid-shin (p=0.050), and waist (p=0.030) levels. Meanwhile, on the left side, statistically significant differences could be found when comparing land and mid-thigh levels (p=0.020), land and waist (p=0.006), mid-shin and mid-thigh (p=0.003) levels. These showed an increasing trend in muscle activities until the mid-thigh level, but dropped in the waist level in both legs. The muscle activity of RF during the swing phase in all water levels was higher than the land level. For TA, there was no significant difference observed on the right side (p=0.319) in terms of %MVC across four conditions. However, statistically significant differences could be seen on the left side (p=0.001) with the post-hoc differences revealed when comparing land and mid-shin (p<0.001), mid-thigh (p<0.001) and waist (p=0.002) levels. For BF, the difference of %MVC was not significant on the right side (p=0.074), but significant on the left side (p=0.012). Post-hoc tests for the left side revealed significant differences between land and mid-shin (p=0.021), land and waist (p=0.021), mid-shin and mid-thigh (p=0.030), mid-thigh and waist (p=0.36) levels. For mGas, there was no significant difference observed on the right and left sides in terms of %MVC across four conditions (p>0.05).

Peak angle

One-way ANOVA was used to compare the peak hip, knee and ankle angles across different water depths for the entire stride and the results are presented in Table 4.

A statistically significant difference was found in the left hip peak angle among the land and three different water depths (p=.033). Significant differences were found in all four comparisons between land and the mid-shin level (p=.032), land and the mid-thigh level (p=.036), mid-shin and waist levels (p=.034) and between the mid-thigh and waist levels (p=.038) using post-hoc comparisons. However, no significant difference was found when comparing land and waist conditions, and when comparing mid-shin and mid-thigh conditions. Generally, it showed a trend of increasing the peak angle of hip flexion from land to the thigh level, but decreased in the waist level.

Similarly, a statistically significant difference was shown in the right hip peak angle when comparing land and three different water depths (p=.049). A significant difference was found between mid-shin and waist levels (p=.049), yet no significant difference was found in comparisons of peak angles in other water levels.

As for the peak knee angle, no statistically significant difference was shown in both the left (p=.595) and the right leg (p=.481). Also for the peak ankle angle, no statistically significant difference was found for the left (p=.633) and the right leg (p=.335).

Swing/stance ratio

One-way ANOVA was used to compare the stance/swing ratio across land and different water depths and statistical analysis is presented in Table 5. There was a significant increase in the swing/stance ratio across land and different water depths (both sides: p=.000) on both sides. Post-hoc analysis revealed that on the left side, there was a significant increase in the swing/stance ratio when comparing land and mid-thigh (p=.006), land and waist (p=.000), mid-shin and waist (p=.000), also mid-thigh and waist (p=.009) levels. As for the right side, post-hoc analysis showed significant differences between land and waist (p=.000), mid-shin and waist (p=.000) as well as mid-thigh and waist levels (p=.047). The trend reflected an increase in the ratio with rising water depths.

Borg rating of perceived exertion (RPE)

The RPE values between land treadmill and aquatic treadmill running at different depths showed a significant difference between groups (p=.017). In subgroup analysis, land running compared with shin immersion aquatic treadmill running and waist immersion aquatic treadmill running showed significant changes (p=.003 and p=.015, respectively).

Relationship of muscle activity and peak joint angles across water levels

This study indicates that ATR significantly increased selected muscle activities at different levels of water immersion when compared to land treadmill running. Immersion depth is an important component of hydrotherapy. It is clinically relevant to examine the effect of water depth on the gait. Previous studies have discussed effects of immersion depth on either muscle activities or kinematics, without correlating both. A lack of evidence for the optimum water depth with consideration of muscle activation and kinematics underwater makes clinical decision difficult. Acknowledging the relationship between immersion depths, muscle activation and the gait pattern would help identify potential benefits and provide useful guidelines to make clinical decisions for those using ATR.

Additionally, different water depths during ATR could influence many aspects of muscle activity. For example, there was an increase in muscle activity of the RF till the mid-thigh level, but a decrease in the waist level. This can be explained by an increase in water depth, a longer period of the swing phase and acceleration of an added water mass when performing ATR (Silver et al., 2014). This contributes to an increase in muscle activities in the RF as more hip flexion range is required to work against the effects of hydrodynamic resistance in the swing phase (Silvers et al., 2008). Notably in the mid-thigh level, in order to reach an economical locomotion, effects of hydrodynamic resistance are reduced with greater hip joint flexion range, inducing more knee flexion range (Kato et al., 2001). This also explains why the largest peak hip flexion angle was observed in the mid-thigh level. Our results also show the largest peak knee angle in the mid-thigh level, albeit non-significant. This is consistent with a previous study showing no significant difference in the peak knee angle due to a smaller and streamlined surface area of the shank (Jung et al., 2019). Moreover, as half of the leg is immersed in water, the RF is responsible for overcoming the skin friction drag while lifting the leg above the water surface and the hydrodynamic resistance to perform hip flexion, in addition to knee extension. This leads to the highest %MVC of the RF in the mid-thigh level. However, the effect of buoyancy offsets the effect of hydrodynamic resistance in the waist level, giving rise to more body weight support and thus less recruitment of the RF to flex the hip during the swing phase. Also, as the limbs (i.e., lower extremity) were fully immersed while running, overall muscle activity of the RF decreases.

Importantly, our results confirm that both the BF (stance) and TA (swing) show an increasing muscle activity trend with increasing water depth which is coherent with other studies. The BF is a biarticular muscle which can perform hip extension and knee flexion. Silvers et al. (2014) observed that the BF during walking after heel contact acts as hip extensors together with the gluteus maximus to perform hip extension. Takeru et al. (2002) also reported the BF’s role as hip extension instead of vertical support during the stance phase. Since hip extension is needed when transitioning from heel strike to the pre-swing phase, therefore this can explain our results as higher BF activity is needed to counteract stronger drag force with an increasing water level. These results are in contrast with the significant decrease in muscle activity suggested by Liebenberg et al. (2011). It was reported that reduction in effective body weight (i.e., increased body weight support) resulted in a reduction in muscle activity (Liebenberg et al., 2011). However, it is important to note that in that study, a lower body positive pressure treadmill was used to provide body weight support and the study did not address the effect of hydrodynamic resistance as it only provided body support to imitate the buoyancy effect provided in the water environment. Therefore, a combination of hydrodynamic resistance and buoyancy may affect the pattern of muscle activities.

For the muscle activity outcome observed in the TA, our results showed that the muscle activity increased with increasing water depth. It is partly consistent with the study showing a significant difference between land and waist level running at 0.8 m∙s^-1 (Takeru et al., 2002). This pattern could be explained by the role of the TA to overcome more hydrodynamic resistance to dorsiflex the ankle in preparation for foot strike during the swing phase (Kato et al., 2001). Also, it is supported by the evidence of an increased swing/stance ratio which may prolong the muscle activation of the TA in the swing phase. However, our results are inconsistent with Silvers et al. (2014), showing no significant difference when comparing ATR and LTR.

Gait kinematics in different water levels Overall decrease in hip ROM in the waist level

There was a decrease in overall hip range of motion in the waist level compared to the other three conditions. It could be the result of increased resistance and buoyancy due to increased water depth (Jung et al., 2019). More resistance makes weight transfer over the standing foot more difficult. Besides, more upthrust force due to increased buoyancy may also interfere with weight transfer. Therefore, a decrease in hip extension occurs during the mid to late stance phase, reducing the hip ROM.

Swing/stance ratio

In our study, an increasing swing/stance ratio across different water depths was observed. Our findings support the qualitative results of other studies. When performing ATR at the waist level, more body support can be provided with buoyancy which can subsequently influence the flight phase during running. This enhances the dynamic balance, enabling participants to maintain balance even with longer single-leg stance time (Baezner et al., 2008). In addition, since fewer body parts are affected by lower air drag force with increasing water depth, hydrodynamic resistance in the form of drag is also higher (Machado, 2010). Hydrostatic resistance, caused by limb movement speed and the limb frontal surface area, may decrease velocities during limb movement (Macdermid et al., 2017). When combining the effects of buoyancy and drag forces during ATR at a constant speed, it apparently prolongs the swing phase duration (Silvers et al., 2014). This is in line with the qualitative findings showing the delay pattern shown in ATR at the waist level, meaning that the peak angle of joints at the waist level usually comes after other levels.

Although the conditions varied greatly, it is meaningful to note that there were many similarities in muscle activity and kinematics between conditions. Humans tend to have a preferred gait pattern. It is natural for humans to accommodate their gait to compensate variations in human postures, and to adapt a natural gait pattern under different conditions for equilibrium and stability (Carlos et al., 2017). It could be due to the fact that one of our limitations of this study was the under-control speed. The control of speed could have constrained participants in using a narrow range of gait patterns. Subsequent research should be conducted which would allow participants to select a preferred speed at each water depth. Likewise, our understanding of the influence of water depth on muscle activity would be enhanced by having participants exercise at the same exercise intensity for each depth. The importance of the present study is that we demonstrated muscle activity at different depths when speed of locomotion was controlled.