Reporting of the current pilot study was guided by the recommendations of the CONSORT extension to pilot and feasibility trials (22). (Additional file 1)
Trial design and participants
This was a non-randomized cross-sectional pilot study involving three groups of participants. Patients with PS- and MP-TKA were recruited from the division of Joint Replacement Surgery from a local Hospital. Potential TKA participants were recruited if they met the inclusion criteria: 60 years or older, independent walking for at least 10 minutes indoors without using a walking aid, no contraindication to exercises. Participants were excluded if they had a diagnosis of neurological or vestibular impairment, uncontrolled cardiopulmonary disorders, severe diabetes mellitus, rheumatic arthritis, a body mass index (BMI) ≥ 40kg/m2, a knee flexion contracture ≥ 10º, or Kellgren and Lawrence grade ≥3 knee osteoarthritis on the non-operated knee indicating absence of moderate or severe arthritis, a history of lower extremity fracture or surgery other than the primary unilateral TKA, and recent lower extremity musculoskeletal injuries that precluded an individual from participating in the study. These patients were excluded because their medical conditions might affect physical performance and confound our findings. Furthermore, patients were excluded if they had major postoperative complications such as superficial or deep infections, deep venous thrombosis, pulmonary embolism or wound healing problems.
TKAs in both groups were implanted by two senior orthopaedics surgeons (one is the co-author CHY) who have been performing both designs (MP & PS prostheses) for more than 5 years. Standard surgical techniques included a longitudinal skin incision, medial parapatellar approach, subperiosteal dissection over the medial tibial plateau, and resurfacing of the patella depending on the intraoperative wear pattern. Surgery was undertaken under spinal or combined spinal and epidural anesthesia and lasted for 90 to 120 minutes. Patients received standard, weight-based doses of preoperative antibiotics and intravenous tranexamic acid, followed by 2 weeks of venous thromboembolism prophylaxis with aspirin. Inpatient physiotherapy was commenced on postoperative day 0 under the management of physiotherapists in order to achieve full weight bearing on the operated leg. Active, active-assisted, and passive knee mobilization exercises were progressively prescribed. Patients were trained to walk with a frame or a quadripod depending on their progress. Additionally, they were given an exercise pamphlet regarding lower limb stretching and mobilization exercises to perform at home following discharge.
Age-matched asymptomatic controls comprised community senior center attendees. Inclusion criteria for asymptomatic controls included: no signs of stiffness or pain in the knees in the last year; and able to walk unaided both indoors and outdoors. Exclusion criteria for asymptomatic controls followed those of participants with TKA. The study was approved by the Human Subjects Ethics Sub-committee of the University (HSEARS20161110003). All participants provided informed consent before participation.
Since the objective of this study was to assess the feasibility of using body-worn sensors to access balance performance of TKA patients with different prothesis design, a formal sample size calculation was not conducted.
Self-reported pain, stiffness and physical functioning
Pain, stiffness and physical functioning of all participants were measured with the Chinese version of the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) (23). This activity-based self-administered questionnaire comprises 24 questions related to knee pain (5 items), knee stiffness (2 items) and physical functions (17 items). WOMAC has demonstrated good test-retest reliability for evaluating the TKA population (intra-class correlation coefficient = 0.82, 0.88, 0.84 for pain, stiffness and function respectively) (23). A higher score in each subscale indicates that the respondent has more knee pain, more knee stiffness or poorer physical function.
Mobility was assessed with two performance-based tests: the timed-up-and-go test (TUG) and the six-minute walk test (6MWT). These tests are commonly used to evaluate the functional recovery after TKA (24, 25). The TUG has an excellent reported inter-rater reliability with an intraclass correlation coefficient of 0.99, and is a functional test of strength, agility and dynamic balance (26). Participants were instructed to rise from an armless chair (seat height of 46cm), walk unaided at a self-selected comfortable pace along a line on the floor for 3 meters, and then turn and walk back to the chair and sit down (26). Participants performed a practice trial, followed by two experimental trials with the faster trial time used for analysis. The 6MWT assesses walking endurance/tolerance and has demonstrated excellent inter-rater reliability (intraclass correlation coefficient of 0.91) (27). The test was demonstrated and then participants were asked to walk as quickly as possible back and forth along a 20m hallway for 6 minutes without running or jogging (28). Participants were allowed to slow down, to stop, or to rest, if necessary. A standard set of recommended instructions and encouraging statements were used (28). The total distance covered was recorded for analysis.
Standing balance and gait stability assessments
Balance parameters during a standing task and during the 6MWT were assessed with two synchronized inertial sensors (Figure 1). Each sensor contained a tri-axial accelerometer and a tri-axial gyroscope (Opal, APDM Inc, Portland, OR, USA; sampling frequency 128 Hz). The lumbar sensor was firmly strapped onto the participant with a belt approximately at the L5 level (near the body center of mass). The head sensor was attached to a plastic helmet that was secured at the vertex of the participant’s head. The three-dimensional angular velocity and acceleration data from each sensor were collected during both static and dynamic tasks and processed using a custom-written program (MATLAB, Natick, MA, USA).
Postural sway was assessed using the near-tandem stance test with eyes open. Participants were instructed to fold their arms across the chest and to stand with the heel of front foot (dominant foot: determined by self-reported leg dominance using the question of which leg will be used to shoot a ball on a target regardless of any pain on lower limbs) 2.5cm anterior and 2.5cm lateral (marked by a 2.5 cm x 2.5 cm cardboard template) to the great toe of the rear foot on a hard surface for 30 seconds. The head and lumbar static balance parameters were assessed over the middle 25 seconds to prevent any movements at the beginning or end of the test affecting the results. Static balance parameters included the 95% range of sway in degrees (ϴ) in the anteroposterior (AP) and mediolateral (ML) directions called pitch and roll, respectively (Equations 1 & 2). The root mean square (RMS) of angular velocity was calculated by combining movements about all axes into one parameter (Figure 2), where fA is the acceleration of the sensor after low-pass filtering with a 4th order bidirectional Butterworth filter at a cut-off frequency of one hertz. (see Equations 1 and 2 in the Supplementary Files)
Dynamic balance was assessed during the periods of straight line walking in the 6MWT from multiple 20-meter laps (Figure 3). Turns were identified and excluded from the analysis by the gyroscope threshold of 30º/s about the vertical axis. To prevent false identification of transient gait rotations instead of turns the gyroscope data were low-pass filtered using a 4th order bidirectional Butterworth filter with a 0.5 hertz cut-off frequency and a total turn rotation of at least 90º was required (Figure 4). For each gait parameter the robust mean was calculated (mean after excluding the best lap and worst lap). Steps during each straight line lap were identified by heel strike acceleration peaks (29). Cadence was calculated as the number steps per minute. Stride time variability was the standard deviation of consecutive strides (1 stride = 2 steps); greater variability has been associated with increased fall risk (30). Step time asymmetry was calculated as the absolute difference between left and right step times as a percentage (29). Relative displacements of the head and lumbar were reported as the RMS along the AP, ML and VT and calculated from the sensor data using validated methods (31); reduced VT displacements correlate with less vigorous gait while increase transverse plane displacements are associated with reduced gait stability (19). Finally, harmonic ratios (HR) were used as a measure of rhythm/smoothness of head and lumbar accelerations during walking. Lower HRs indicate reduced dynamic balance and are associated with increased fall risk (32, 33).
All statistical analyses were performed using SPSS software (Version 22, IBM Corp., Armonk, NY). Non-parametric tests were used for the analysis as most variables did not meet the requirements for normality as determined by Shapiro-Wilks tests of normality. Demographic variables specific to participants with TKA (e.g., months after operation and the percentage of participants using walking aids outdoor) were compared between the PS-TKA and MP-TKA participants using Mann-Whitney U tests (for continuous variables) or Chi-square test (for nominal variables). For the remainder of the demographic and clinical variables of interest, Kruskal-Wallis tests (for continuous variables) and chi-square tests (for nominal variables) were used to compare the differences among PS-TKA, MP-TKA and asymptomatic controls. The significance level was set at 0.05 (2-tailed) and post-hoc analysis using Bonferroni adjustments were performed. Effect sizes (r) of each observed difference were calculated by dividing the Z value by the square root of the total number of participants in that pair of groups (34). Cohen’s guidelines for r suggest that small, medium and large effect sizes are 0.1, 0.3 and 0.5 respectively (35).