Squatting biomechanics following physiotherapist-led care or hip arthroscopy for femoroacetabular impingement syndrome: a secondary analysis from a randomised controlled trial

Participants

Ninety-nine participants with FAIS were recruited from the clinics of one of eight orthopaedic surgeons across 11 study sites between February 2015 and December 2017. Eligible participants met the following inclusion criteria: hip pain, aged over 16 years, cam (alpha angle >55°) and/or pincer (lateral centre edge angle >40°, and/or positive cross-over sign or other radiographic signs of pincer) morphology, the treating surgeon believing the patient would benefit from arthroscopic hip surgery, no pre-existing hip OA (Tonnis grade >1 or <2 mm joint space width on pelvic radiograph), and no previous significant hip pathology, injury (such as fracture or dislocation), or hip-shape changing surgery (Murphy et al., 2017). Fifty-four participants (55% of those recruited in the trial) from three study sites attended a biomechanical testing session at the baseline time point for inclusion in this exploratory secondary analysis.

Randomisation, allocation, and blinding

Of the participants who underwent motion analysis at baseline, 29 and 25 were randomised to receive PHT or arthroscopic hip surgery, respectively, using a 1:1 ratio created by a computer-generated minimisation sequence (adaptive stratified sampling) (Fig. 1). An external biostatistician held the randomisation codes to preserve allocation concealment. Each participant was assigned a study identification number at randomisation, which was used on all trial documentation. Neither participants, treating surgeons, nor physiotherapists could be blinded to treatment allocation. As such, treating surgeons and physiotherapists were not involved in outcome assessments. Patient-reported outcome data were collected via online surveys and postal questionnaires and then recorded in a database by a blinded research assistant. Operators collecting data for, as well as operators performing, imaging and biomechanical analyses were blinded to treatment allocation.

Interventions

Arthroscopic hip surgery for participants in this subsample was performed by one of three orthopaedic surgeons experienced in arthroscopic hip surgery for FAIS (6-, 8- and 30-years of experience), within 18 weeks following randomisation. During hip arthroscopy, cam morphology (Nötzli et al., 2002) and/or pincer morphology (Ogata et al., 1990) were resected, and labral and cartilage pathologies were treated. Post-operatively, patients were discharged from hospital when they could walk safely without crutches (typically within 24 h). Post-operative rehabilitation was not standardised. Rather, participants received the typical outpatient rehabilitation recommended by their treating surgeon. The physiotherapists providing PHT were different from those providing post-operative rehabilitation to avoid contamination between groups. Operation notes, intraoperative imaging, and postoperative magnetic resonance imaging were used to evaluate surgical intervention adequacy (Griffin et al., 2018).

The PHT protocol was developed in a feasibility study for the UK FASHIoN Trial by an international panel of physiotherapists, physicians, and surgeons (Griffin et al., 2016b) in accordance with a consensus statement on developing and evaluating complex interventions (Craig et al., 2008; Wall et al., 2016). Participants were provided PHT during six-to-ten sessions over 12–24 weeks by physiotherapists trained in its delivery and commenced within 1 month of randomisation. The number of physiotherapy sessions provided to patients was reflective of the number offered in public and private health services in Australia. PHT included core components of patient education, advice regarding pain relief (anti-inflammatory medication, and oral analgesics if required), and a progressive individualised exercise program taught and supervised by the treating PHT physiotherapist and repeated at home (Fig. 2). Exercises were prescribed from a provided set of recommended exercises and progressed as appropriate for each participant. Exercise programs initially targeted pelvic and hip stability and progressed to include stretching hip abductors and external rotators during positions of hip flexion and extension, and strengthening the hip external rotator, abdominal and gluteal musculature, as well as the lower limb in general. Evidence of protocol fidelity and exercise program individualisation, progression, and supervision were documented in case forms, which were assessed by either two members of the panel that developed the PHT protocol, or two investigators from the Australian FASHIoN trial (Hunter et al., 2021).

Data collection

Participants underwent biomechanical testing at the University of Melbourne’s Centre for Health, Exercise & Sports Medicine (CHESM) for secondary outcome analysis during two testing sessions: after randomisation but before treatment (baseline for this analysis) and 12-months post-randomisation (follow-up). Three-dimensional motion data were acquired using a 12-camera Vicon motion capture system (Vicon, Oxford Metrics, UK) sampling at 120 Hz. Ground reaction forces were acquired using two ground-embedded force plates (Advanced Mechanical Technology Institute, Watertown, MA, USA), sampling at 1,200 Hz. Participants wore the full Griffith University marker set (Savage et al., 2021), and performed five deep squatting trials in standardised footwear (Volley, Melbourne, Australia). Firm foam wedges (Slant by OPTP, Minneapolis, MN) were placed on the back third of each force plate. Participants were instructed to stand with their feet parallel and shoulder width apart, with each heel on a wedge to position the feet in approximately 30° of plantarflexion to encourage deep hip flexion, unrestricted by ankle dorsiflexion range (Lamontagne et al., 2011). During each trial, participants were instructed to squat using their preferred strategy, with their arms straight to the front (i.e., horizontal) for balance, and their weight distributed evenly between both feet. Participants performed each squat at a self-selected controlled pace, descending until the end of self-selected available range, pause for three seconds, and return to standing at the same pace.

Before treatment randomisation, and 12-months post-randomisation (i.e., baseline for this analysis and follow-up, respectively), participants completed a series of patient-reported outcomes, including the International Hip Outcome Tool (iHOT-33) (Mohtadi et al., 2012), and underwent medical imaging (including MRI and plain radiographs) of their study hip to characterise FAIS pathoanatomy, including cam and pincer morphologies. The imaging sequences and analyses have been described elsewhere (Hunter et al., 2021; Murphy et al., 2017).

Data processing

Motion data were labelled and cleaned in Nexus version 2.9.3 (Oxford Metrics, UK), and subsequently processed using a motion data elaboration toolbox for neuromusculoskeletal modelling (MOtoNMS) (Mantoan et al., 2015), implemented in MATLAB version 2022a (Mathworks, Natick, MA, USA). Marker trajectories and ground reaction forces were filtered using a 2^nd order zero-lag Butterworth low-pass filter (nominal 6 Hz cut-off). Hip joint centres were defined using the Harrington regression equation (Harrington et al., 2007). Knee and ankle joint centres were estimated using the midpoints of lateral and medial femoral condyles and malleoli markers, with inferior offset of 2.7% of shank length (Bruening, Crewe & Buczek, 2008), respectively. A musculoskeletal model (Catelli et al., 2019b) containing 80 lower-limb Hill-type muscle-tendon units and 37° of freedom was implemented in OpenSim version 3.3 (Delp et al., 2007). For each participant, the generic model was linearly scaled based on static anthropometric dimensions obtained during a trial of quiet upright stance, to match their mass, segmental dimensions, and moment of inertia (Kainz et al., 2017). Marker trajectories and ground reaction forces were imported into OpenSim, and inverse kinematics and inverse dynamics tools were used to calculate joint angles and moments for each degree of freedom, respectively. Trunk kinematics were expressed relative to the global coordinate system. Centre of mass velocity of the overall model was calculated using the body kinematics tool. Vertical knee joint centre linear velocity was calculated as the derivative of virtual knee joint centre marker position with respect to time. OpenSim analyses were performed using the OpenSim application programming interface, implemented in Matlab.

Each squat trial was considered in three phases: descent, hold, and ascent. The start and end of squat descent and ascent phases were identified using a previously published automated event detection algorithm (Stevens et al., 2018), which utilised a relative threshold of 10% of peak linear sagittal plane knee velocity. Each trial was visually inspected to ensure that events were identified correctly by the algorithm. Spatiotemporal variables included squat descent and ascent velocity (m∙s⁻¹), and maximum squat depth relative to starting height (change in vertical position of the midpoint of the two sacral markers, expressed as a percentage of limb length, %). Trunk, pelvis, knee and ankle (sagittal plane), as well as hip (sagittal, frontal, and transverse planes) kinematics, and net hip (sagittal, frontal and transverse planes), knee and ankle (sagittal plane) external moments were time normalised to 200 points, consisting of: descent (start to end of descent, 80 points), hold (end of descent to start of ascent, 40 points), and ascent (start to end of ascent, 80 points). Joint moments were normalised to body weight multiplied by height and expressed as a percentage (N∙m/BW∙HT[%]) (Moisio et al., 2003). Peak-to-peak excursion, defined as the difference between maximum and minimum joint angles over the squat cycle, was calculated for each plane at the trunk, pelvis, hip, knee, and ankle.

Statistical analyses

Representative mean trials were calculated for each participant at baseline and follow-up by averaging spatiotemporal, angle, and moment data across four trials at each testing session. Descent and ascent phases were extracted from each representative trial and analysed separately. For each participant, 12-month changes in spatiotemporal parameters, peak-to-peak excursion, and time-series angles and moments during descent and ascent were calculated (mean follow-up trial minus mean baseline trial) for between-group comparisons. Due to the exploratory nature of this secondary analysis, no pre-defined hypotheses were tested. Adjustments for multiple comparisons were not applied to discrete variables based on recommendations from Bender & Lange (2001), and to time-series data as Statistical Parametric Mapping (SPM) is grounded in random field theory, which mitigates the multiple testing problem (Adler & Taylor, 2007). An a priori power calculation was not completed as these analyses were novel, and no data were available. Between-group differences in patient-reported outcomes have been examined for the Australian FASHIoN trial previously (Hunter et al., 2021), and the minimal clinically important difference (MCID) for iHOT-33 reported was 6.0 units. The same MCID has been used in this exploratory secondary analysis.

Statistical analyses were performed in MATLAB and significance was set at p < 0.05. Time-series angles and moments during squat descent and ascent were compared via SPM or Statistical nonParametric Mapping (SnPM), as required, using open-source code from spm1d (www.spm1d.org) implemented in Matlab version 2022a (Mathworks, Natick, MASS, USA) (Pataky, Robinson & Vanrenterghem, 2013). All data (including 12-month changes) were assessed for normality: participant characteristics, spatiotemporal variables, and peak-to-peak excursion were assessed for normality using a Shapiro-Wilk test; time-series angles and moments were assessed using a D’Agostino-Pearson K2 test. Differences in participant characteristics between baseline and follow-up for each treatment group were explored using paired t-tests and Wilcoxon Signed-Rank tests, as required. Differences in patient-reported outcomes between baseline and follow-up were assessed using Wilcoxon Signed-Rank tests. Between-group differences in changes across squat descent and ascent were explored for PHT versus arthroscopy using independent samples t-tests, or Mann-Whitney U tests (for discrete variables), as required. Within-group differences between baseline and follow-up for each treatment group were examined using paired t-tests, or Wilcoxon Signed-Rank tests (for discrete variables), as required. Additionally, a general linear model (GLM) was used to explore 12-month changes in kinematics and moments during squat descent and ascent independent of the effects of speed. For between- and within-group comparisons, the GLM included change in velocity (i.e., follow-up minus baseline) and baseline and follow-up speeds, respectively. Corrected effect sizes (Hedges’ g) were calculated for each of the points in descent and ascent phases (Lawrenson et al., 2021). Mean differences with 95% confidence intervals (CI) (lower-bound, upper-bound), and corrected effect sizes (Hedges’ g) were used to describe parametric data. Median differences and interquartile range (25^th and 75^th percentiles) were used to describe non-parametric data. For significantly different windows identified via SPM, peak mean difference and respective 95% CI (lower-bound, upper-bound) are reported. Thresholds of 0.2, 0.5, and 0.8 were used to identify small, medium, and large effect sizes, respectively (Cohen, 1988).

Kinematics

No significant between-group differences in 12-month changes in trunk, pelvis, hip, knee, or ankle kinematics (Figs. 3–5) were detected during squat descent or ascent, or in peak-to-peak excursion (Table 2). Compared to baseline, both groups exhibited decreased anterior pelvic tilt at follow-up during squat descent (PHT: MD −8.30° (95%CI [0.21– 16.39]), ES = 0.82, p = 0.02; Arthroscopy: MD −10.95° (95%CI [−5.54 to 16.34]), ES = 1.24, p < 0.001), and ascent (PHT: MD −7.98° (95%CI [−0.38 to 16.35]), ES = 0.76, p = 0.02; arthroscopy: MD −10.82° (95%CI [3.82–17.81]), ES = 1.17, p = 0.002). Compared to baseline, both groups exhibited decreased hip flexion at follow-up during squat descent (PHT: MD −11.86° (95%CI [1.67–22.05]), ES = 0.83, p = 0.02; arthroscopy: MD −16.78° (95%CI [8.55–22.01]), ES = 1.43, p < 0.001), and ascent (PHT: MD −12.86° (95%CI [1.30–24.42]), ES = 0.82, p = 0.013; arthroscopy: MD −16.53° (95%CI [6.72–26.35]), ES = 1.41, p < 0.001). Compared to baseline, reduced knee flexion was exhibited at follow-up during squat descent in the PHT group (MD −6.62° (95%CI [0.56–12.67]), ES = 0.54, p = 0.02), and in both groups during ascent (PHT: MD −8.24° (95%CI [2.38– 14.10]), ES = 0.84, p = 0.003; arthroscopy: MD −8.00° (95%CI [−0.02 to 16.03]), ES = 0.79, p = 0.02). Compared to baseline, the PHT group exhibited increased plantarflexion during squat ascent at follow-up (MD −3.58 (95%CI [−0.12 to 7.29]), ES = 0.66, p = 0.04). No significant differences were detected between baseline and follow-up in peak-to-peak excursion for either group.

After adjusting for descent/ascent speed, significant differences in trunk kinematics were detected between baseline and follow-up for PHT and arthroscopy groups that were not identified in unadjusted analyses (Fig. 3). Compared to baseline, increased trunk flexion was exhibited at follow-up during descent in the PHT group (MD 7.50° (95%CI [−14.02 to −0.98]), ES = 0.87, p = 0.03) and in both groups during ascent (PHT: MD 7.29° (95%CI [−14.69 to 0.12]), ES = 0.79, p = 0.02; arthroscopy: MD 16.32° (95%CI [−32.95 to 0.30]), ES = 0.76, p = 0.04). Significant differences identified in unadjusted analyses were not significant after adjusting for speed.

Kinetics

No significant between-group differences in 12-month changes in sagittal, frontal, or transverse plane hip moments, or in sagittal plane knee or ankle moments during squat descent or ascent (Figs. 4 and 5). Compared to baseline, both groups squatted with lower hip flexion moments during squat descent (PHT: MD −0.55 N∙m/BW∙HT[%] (95%CI [0.05 to 1.05]), ES = 0.67, p = 0.001; arthroscopy: MD −0.84 N∙m/BW∙HT[%] (95%CI [0.06–1.61]), ES = 0.82, p < 0.001) and ascent (PHT: MD −0.464 N∙m/BW∙HT[%] (95%CI [−0.002 to 0.930]), ES = 0.570, p = 0.01; arthroscopy: MD −0.90 N∙m/BW∙HT[%] (95%CI [0.13–1.67]), ES = 0.78, p < 0.001).