Fifty-two female patients with secondary hip OA (age: 47.8 ± 10.7 years) were consecutively recruited for this study. The patients were enrolled from among patients who attended the Department of Orthopedic surgery of a university hospital continuously from April 2013 to March 2015. The inclusion criteria were as follows: patients aged 20–65 years, who had secondary hip OA, and were able to walk without any assistive device in daily life. The exclusion criteria were as follows: patients with a history of previous hip surgery (e.g., osteotomy and arthroplasty), and neurological, vascular, or other conditions that affected gait. The distribution of the patients among the hip OA stages  was as follows: pre-OA (n = 15, 28.8%), early-OA (n = 25, 48.1%), and advanced-OA stages (n = 12, 23.1%). The side with more severe radiographic OA change was used in the analysis. Written informed consent was obtained from all the patients, and this study was approved by the institutional review board.
The participants wore body-fitting T-shirts and short spats. Twenty-six reflective markers were placed at various points on their body by a single experienced examiner. Each body segment was composed of the following marker sets: the trunk, consisting of the seventh cervical spinous process, the tenth thoracic spinous process, the jugular notch, the xiphoid process, and the bilateral acromioclavicular joints; the pelvis, consisting of the bilateral anterior and posterior superior iliac spine; the thigh, consisting of the superior aspect of the greater trochanter, and the medial and lateral femoral condyles; the shank, consisting of the medial and lateral femoral condyles, and the medial and lateral malleoli; and the foot, consisting of the heel, the head of the first and fifth metatarsal, and the medial and lateral malleoli. The marker position (200 Hz) and the ground reaction forces (1000 Hz) were collected using an 8-camera Vicon motion system (Vicon Motion Systems Ltd., Oxford, England) and force plates (Kistler Japan Co., Ltd., Tokyo, Japan). The marker position data and ground reaction force data were filtered using a fourth-order Butterworth low-pass filter at 6 and 20 Hz, respectively. Gait speed, stride length, cadence, and external hip joint moments were computed using Vicon Nexus and BodyBuilder (Vicon Motion Systems Ltd., Oxford, England) . The external hip moment peak in each of the 3 planes and the hip moment impulse (timed integral of the hip joint moment) in each of the 3 planes were calculated as indexes of hip joint loading. The hip joint moment was normalized to body weight and height (Nm/kgm). Stride length was expressed as a percentage of leg length (distance between the anterior superior iliac spine and medial malleolus). The mean values of the gait variables from the 3 trials were used in the analysis.
All the participants practiced normal and fast gaits several times to familiarize themselves with the experimental environment prior to data recording. At least 3 trials were recorded for each of the barefoot gaits at self-selected normal (normal gait) and fast speeds (fast gait). By adjusting the start position, gait trials in which the feet properly contacted the force plates, without making the participants aware of the position of the force plates, were secured.
The types of strategies for increasing gait speed were classified on the basis of the average stride length and cadence of 3 trials each for normal and fast gait, as described in a previous study . First, the rates of increases in stride length and cadence were calculated, and then the ratio of the rate of increase in cadence to the rate of increase in stride length was computed. For the ratio, a value < 0.75 was defined as type S (i.e., increase mainly stride length), a value ≥ 1.55 as type C (i.e., increase mainly cadence), and a value ≥ 0.75 but < 1.55 as type SC (i.e., increase both stride length and cadence; Figure 1). Participants with < 5% increase in gait speed during fast gait as compared to normal gait were excluded from the analysis.
Assessment of hip pain and physical function
The average pain intensity at the hip joint during daily life in the last 3 months was assessed on a 100-mm visual analog scale. Physical function was assessed using the physical component summary of the Japanese version of the Medical Outcomes Study 36-Item Short-Form Health Survey (SF-36) version 2.0. The SF-36 is the tool most commonly used to assess health status in the general population, and has been used for patients with OA with high reliability and validity [12,13].
Assessment of joint space narrowing and hip impairments
A digital supine anteroposterior radiograph of the pelvis was obtained in a standardized manner by skilled radiology technicians. To assess the degree of cartilage degeneration and severity of hip OA, the minimum joint space width (mJSW) was measured digitally on the radiograph by a single examiner, using Centricity Enterprise Web, version 3.0 (GE Healthcare, Little Chalfont, England). The mJSW had the highest level of intra- and inter-rater reliabilities and good applicability as a parameter for hip OA diagnosis . mJSW was measured at the vertex and medial and lateral sides of the weight-bearing surface, and if a minimum distance was present at a position other than those 3 locations, it was also measured as a fourth measurement . The minimum value for 3 or 4 locations was defined as the mJSW [15,16]. The intra-rater reliability (ICC 1,1) of the mJSW measurement was 0.99 .
Hip range-of-motion (ROM) and muscle strength were assessed by a single experienced examiner as previously reported [17,18]. The passive ROM of the hip joint was measured at flexion, extension, and abduction, using a standard two-arm goniometer (Sakai Medical Co., Ltd, Tokyo, Japan). The intra-rater reliability (ICC 1,1) of the ROM measurements ranged from 0.82 to 0.98 . The maximal isometric muscle strengths on hip flexion, extension, and abduction were measured using a handheld dynamometer (μTAS F-1; Anima Co., Ltd, Tokyo, Japan). Muscle strength was measured twice, and the mean of the measurements from the 2 trials was used in the analysis. The intra-rater reliability (ICC 1,1) for the muscle strength measurements ranged from 0.93 to 0.96 . Muscle strength was normalized to the body weight (Nm/kg).
Differences in hip pain severity and physical function status, main outcome measures, demographic characteristics, mJSW, hip ROM, and hip muscle strength were tested using an unpaired t-test with Holm correction. Furthermore, as hip pain and physical function could be influenced by aging and OA severity [19−21], comparisons of these variables were also performed with adjustment for age and mJSW using a general linear model. A sensitivity analysis was also performed to evaluate the robustness of the type classification and the results of comparison of hip pain and physical function among types. Changes in gait speed, stride length, cadence, and hip joint moment were tested using the analysis of variance for split-pot factorial design (type × speed). We also calculated effect size in terms of Cohen’s d and f using GPower 3.1.7 (Heinrich-Heine-Universität Düsseldorf, Düsseldorf, Germany). Cohen’s d values of 0.20, 0.50, and 0.80, and Cohen’s f values of 0.10, 0.25, and 0.40 indicate small, moderate, and large effects, respectively . SPSS version 26.0 (IBM Japan Ltd., Tokyo, Japan) was used for the statistical analysis. The significance level was set at P < 0.05.