This study was approved by the institutional review board of our university and informed consent was obtained from all subjects. In total, 15 female subjects with hip OA and 15 female healthy controls were recruited. Inclusion criteria of the hip OA group were women who were scheduled to undergo unilateral total hip arthroplasty for treatment of moderate to severe OA and aged < 85 years. The severity of OA was determined on radiography according to the Kellgren and Lawrence (KL) grade  in all cases. Exclusion criteria of the hip OA group included a history of (1) immunosuppression or autoimmune deficiency, (2) inflammatory arthritis, (3) local or systemic infections, (4) knee arthritis and/or total knee arthroplasty, or (5) symptomatic spinal cord disease. None of the healthy controls had a history of bone fracture or surgery in the lower limbs, history of neurological, respiratory, or cardiovascular diseases, musculoskeletal disorders within the past 6 months, or previous history of trauma.
Participants wore specifically designed shorts with sensors while using the LBPP treadmill. The height of the chamber was fixed to accommodate the participant, and sensors from the shorts were attached to the LBPP treadmill. The height of the chamber was set equal to that of the greater trochanter of the participant femur (Fig. 1). Calibration to determine the correlation of gravity and the internal pressure of the chamber was performed for each participant, as previously described . Participants walked at a self-selected speed under 100% bodyweight (BW) conditions on the LBPP treadmill (Anti-Gravity Treadmill M320, Alter G, Inc., Fremont, California, USA). The walking speed under 75% and 50% BW conditions were consistent with the 100% BW condition. Participants walked 30 s under three conditions selected randomly (100% BW, 75% BW, and 50% BW) for the testing procedure. Before recording the walking trials, participants were asked to familiarize themselves with walking on the LBPP treadmill for three minutes and were given 90 s to adapt to each unweighting condition (Fig. 1). Participants in the hip OA group were asked to assess their hip pain using a numeric rating scale (NRS) in which 0 represented no anxiety and 10 represented the highest level of anxiety  during walking under 100%, 75%, and 50% BW conditions.
Data collection using the motion analysis system
All data collections were performed on the OA side in the OA group and on the dominant leg in the control group. The dominant side in the control group was defined according to which leg participants used for kicking. Data were collected using a motion analysis system (H-Gait system, Laboratory of Biomechanical Design, Hokkaido University, Sapporo, Japan) where wearable sensors analyzed 3-D gait kinematics [8, 12]. Briefly, seven wearable sensor units (TSDN121, ATR-Promotions, Inc., Kyoto, Japan), which consisted of tri-axial acceleration sensors and tri-axial gyrosensors, were placed on seven lower-limb body segments (pelvis, right and left thigh, right and left shank, and right and left feet) as shown in Fig. 2. Acceleration and angular velocity data were collected simultaneously during gait via wireless connection (Bluetooth) in real-time. Sensor specifications were the same as those mentioned in the previous studies [8, 12].
According to a previous study , a calibration test for each participant was performed to measure the acceleration data of the sensors in the upright and inclined positions to calculate the initial inclination of each sensor with respect to the gravity. Before each trial, an initial static phase was acquired in the upright position. When participants started walking, subsequent 3-D orientations from the initial one were estimated by integrating the angular velocity with the drift removal using MATLAB (Mathworks, Natick, MA, USA) software . The 3-D angular displacement from the initial upright position was calculated in a quaternion according to a previous study . From these data, spatiotemporal gait parameters and flexion-extension angles of the hip, knee, and ankle joints during walking under each unweighting condition were evaluated for each participant. This H-Gait system divides a 30 s walking into gait cycles and calculates the joint angles of each joint for each gait cycle. A median of a gait cycle during the 30s walking under each unweighting condition was used for analyses. For the gait cycle, one gait cycle from heel contact to the next heel contact was normalized to 100%. The heel contact was defined on the peak angular velocity of the shank in a forward direction . In regards to the validity and reliability of the gait analysis system, Tadano et al. analyzed the kinematics of lower limbs in walking using the H-Gait system and compared them with that of a camera-based motion analysis system . The correlation coefficient was 0.98 for the hip flexion angle, 0.97 for knee flexion angle, and 0.78 for the ankle dorsiflexion angle, respectively.
Comparisons of demographic characteristics and walking speed between the groups were performed using independent Student’s t-tests. One-way ANOVAs with post hoc Bonferroni tests were used to investigate differences in NRS scores during walking under 100%, 75%, and 50% BW conditions for the hip OA group. Two-way repeated ANOVAs (3 BW conditions × 2 groups) were performed to assess the main effect of BW condition (100% BW, 75% BW, 50% BW) and group (control, OA) on spatiotemporal gait parameters and peak angles of each joint. When interactions were non-significant, main effects were assessed. If the main effect of the BW condition was statistically significant, post-hoc Bonferroni tests were performed to evaluate significant differences among BW conditions on spatiotemporal gait parameters and peak angles of each joint. The significance level was set at 0.05. Statistical analyses were performed using IBM SPSS version 17 software (SPSS Inc., Chicago, IL, USA).