Arthroscopic hip surgery compared with personalised hip therapy in people over 16 years old with femoroacetabular impingement syndrome: UK FASHIoN RCT

Define eligibility criteria

The eligibility criteria were initially designed in collaboration with the Multicenter Arthroscopy of the Hip Outcomes Research Network (MAHORN), which is an academic group of highly experienced hip arthroscopists within the International Society for Hip Arthroscopy (Moffat, UK) (URL: www.isha.net). These criteria were then discussed with a further sample of 14 UK specialist hip surgeons with experience of treating patients with FAI syndrome. In individual interviews, a variety of clinical scenarios were presented and the surgeons were asked to describe decision-making for treatment. Minor modifications were made to the eligibility criteria. These criteria were then tested during the recruitment of real patients during the pilot RCT and were found to be easy to apply, with little disagreement among clinicians.

Define a protocol for hip arthroscopy for FAI syndrome

A draft protocol for arthroscopic treatment of FAI was developed in a consensus conference with MAHORN members. This draft was circulated among the sample of 14 UK hip surgeons for feedback. After editing, it was recirculated and approved by all. The protocol was then tested in 21 participants randomised to surgery in the pilot trial. We also developed a method to measure fidelity by intraoperative photographs and postoperative magnetic resonance imaging (MRI), which was assessed by a panel of independent international experts. We showed that this approach was acceptable to surgeons and demonstrated complete adherence to protocol in six of seven operations at the first panel conference.

Define a protocol for best conservative care (comparator)

We performed a systematic review of non-operative care for FAI. 20 This revealed little evidence of a standard for best conservative care, even though many NHS commissioners describe the failure of conservative care as a prerequisite for surgery. 21 There was some evidence that physiotherapy-led non-operative care is most frequently used. 20 This is complemented by established theory and evidence supporting treatment effects for physiotherapy in other painful musculoskeletal conditions, including osteoarthritis and back pain. 22,23

We used a combination of consensus methods (e.g. Delphi and nominal group techniques) among physiotherapists to agree a protocol for ‘best conservative care’. 5 We advertised to relevant networks of the Chartered Society of Physiotherapy (CSP) (London, UK) through their interactive communication system (interactiveCSP) and in the Frontline magazine24 (a twice-monthly magazine posted to 52,000 CSP members in the UK). These advertisements invited physiotherapists to help develop a consensus for a best conservative care treatment protocol for FAI syndrome. Electronic invitations were also sent to physiotherapists in the USA and Australia who were known to us through previous collaborative work on FAI syndrome. To encourage a process of ‘snowball sampling’ within the international community, these therapists were encouraged to invite colleagues with experience and interest in managing FAI syndrome to join in the consensus process.

We developed a physiotherapy-led four-component protocol to be delivered over 12 weeks, with a minimum of six one-to-one treatment sessions. 5 The protocol included (1) a detailed patient assessment; (2) education and advice about FAI syndrome; (3) help with pain relief, including hip joint steroid injections if required; and (4) an exercise programme that has the key features of individualisation, supervision and progression. We used a patient focus group to choose the most acceptable name for this protocol of best conservative care. The group made it clear that we should express that this was a coherent and valid alternative to surgery and different from the physiotherapy likely to have been received already, and recommended the name personalised hip therapy (PHT).

In the development of PHT, we struck a balance between the need for a meaningful comparator for hip arthroscopy, the need to ensure that PHT is different from previous physiotherapy that FAI syndrome patients’ may have experienced and the need for PHT to be deliverable in the NHS outside a trial. UK physiotherapists and patients felt that PHT was ‘best’ in that not all patients currently receive such a comprehensive package, but ‘conventional’ in that all its elements are widely used and the package is deliverable within usual constraints in the NHS. We tested the protocol and a logbook approach to assessing fidelity in 21 participants randomised to PHT in the pilot trial. The protocol was acceptable to patients and physiotherapists, and we demonstrated complete adherence in seven of the first eight participants.

Define willingness of centres and patients to be recruited to a randomised controlled trial

We performed a survey of all orthopaedic surgery departments in NHS hospital trusts in the UK. Clinical directors of those departments reported that 120 consultant surgeons were treating FAI syndrome. We contacted all 120 surgeons who reported having performed 2399 operations for FAI in 2011/12. A total of 1908 operations were performed by hip arthroscopy and 491 operations were open surgery. 3 Thirty-four hospital trusts had a workload of 20 or more hip arthroscopies for FAI syndrome in 1 year. We interviewed 18 of the highest-volume surgeons to explore their views about a trial comparing hip arthroscopy with best conservative care in patients with FAI syndrome. One surgeon felt that he could not participate in a trial because he was certain that surgery worked, five surgeons had a bias towards surgery but recognised the need for a trial and were prepared to randomise patients, and 12 surgeons expressed equipoise and were keen to take part in a trial.

We purposively sampled 18 patients who had been treated for FAI syndrome. Fourteen of these patients had received arthroscopic surgery and five had received physical therapy and steroid injections (one patient had both). These patients had a semistructured interview with a qualitative researcher who had not been involved in their care to explore their experiences of diagnosis and treatment, and their views on the proposed trial. The majority of the patients were young and physically active. Symptoms of FAI syndrome had affected their work, recreation and day-to-day activities, and many reported a great sense of relief when a diagnosis was made. Patients said that both surgical and conservative care would be acceptable. The majority of patients saw surgery as the solution for a condition that they perceived as mainly caused by abnormal bone shapes. On the other hand, non-operative care was perceived as attractive if it might be successful and could avoid the risks of surgery. Some patients commented that they had not been offered a non-operative option and saw this as a positive addition to available treatments. Patients were enthusiastic about research in this field, and about being involved, but had reservations about some of the language involved, for instance ‘trial’, ‘random’ and ‘50 : 50 chance’ implied a lack of personalised care. All of these patients said that they would have been prepared to take part in a RCT as long as the treatment options and uncertainty around them had been fully explained, the treatment they received had been personalised for them and they were assured that their care would be continued whatever happened in the research.

Our findings in these in-depth interviews were broadly consistent with Palmer et al. ’s25 questionnaire survey of 30 surgeons who performed FAI surgery and 31 patients with a diagnosis of FAI syndrome. In Palmer et al. ’s25 study, 71% of surgeons and 90% of patients felt that a trial of this question was appropriate.

We concluded that surgeons in most centres in the UK that perform hip arthroscopy for FAI syndrome, and their patients, would be willing to be included in a RCT.

Understand and optimise recruitment

An important objective of the pilot trial was to explore likely issues in recruitment and to develop optimum procedures for the full trial.

We interviewed all principal investigators (PIs) and research associates (RAs) during the pilot trial to ensure that the study was being described and recruitment procedures were being followed in accordance with the study protocol, and to identify when they were not. We developed training packages to correct common problems. We identified structural features associated with successful recruitment, such as running targeted clinics, having a dedicated RA in attendance and ensuring that referred patients arrived with expectations of receiving treatment for FAI syndrome, rather than being told they had been referred for surgery. This learning was shared across all sites.

We recorded and analysed 87 diagnostic and recruitment consultations with 60 new patients during the pilot trial. We identified where improvements could be made in presenting trial information and in engaging patients to consider participation, guided by our previous work. 26,27 The analysis was targeted at the recruitment levels at specific sites, with individual confidential feedback for recruiters on good practice and areas for improvement, and with anonymised findings being fed back to all sites.

Common difficulties with recruitment that were identified included poorly balanced presentations of treatment options (where surgery was presented at greater length and more favourably than PHT), graphic descriptions of surgery that may have put patients off that option or discouraged participation, presenting trial information in an order that was confusing for patients and surgeons going beyond their protocol brief to explain the trial, rather than referring patients to the trial recruiter for this information. Analysis of the consultations led to the development of a six-step model for the presentation of trial information to optimise recruitment. 3,28

Estimate recruitment rate

Ten clinical centres participated in the pilot trial and nine opened to recruitment within 6 months. At one site, local research and development (R&D) approval was delayed until just before the end of the pilot and so no one was recruited.

Of the 144 potentially eligible patients with hip problems identified at the pre-clinic screening of referral letters, 60 met the inclusion criteria after assessment and were approached for randomisation. The most frequent reasons for exclusion were a diagnosis other than FAI (53/84) and a judgement that the patient would not benefit from arthroscopic surgery (21/84). Forty-two patients (70% of those eligible) consented to take part in the pilot RCT. Among those who declined (n = 18), the most common reasons were a preference for surgery (n = 11) and a preference not to have surgery (n = 3). The mean duration and recruitment rate across all sites was 4.5 months and one patient per centre per month, respectively. The lead site recruited for the longest period (9.3 months) and recruited the largest number of patients (2.1 patients per month).

Selection of appropriate outcome measures

A variety of outcome measures have been used to study patients with FAI syndrome. Some, such as the Western Ontario and McMaster Universities Arthritis Index (WOMAC^®) and the Harris Hip Score, were intended for older patients with symptoms of severe arthritis and are most suitable to measure the effect of hip replacement surgery. 29,30 These measures tend to exhibit ceiling effects and are not sensitive to change after treatment in patients with FAI syndrome. 30,31

The Non-Arthritic Hip Score is a self-administered instrument to measure hip-related pain and function in younger patients without arthritis. The score is valid compared with other measures of hip performance, is internally consistent and is reproducible. 32 However, it is not patient derived, raising concern that it may not measure what is most important to patients.

The International Hip Outcome Tool-33 (iHOT-33) is a patient-derived hip-specific patient-reported instrument that measures health-related quality of life in young, active patients with hip disorders. It was developed by a large international collaboration of patients and clinicians led by MAHORN over 5 years. It comprises 33 items, each measured on a visual analogue scale (VAS), to assess functional limitations, sports activities, and job-related and emotional concerns. Importantly, these items were generated and refined by patients, reflecting their most important concerns. The instrument generates a single score in the range of 0 to 100. People with no hip complaints usually score ≥ 95. A diverse international population of younger adults with a variety of hip pathologies had a mean score of 66, with a standard deviation (SD) of 19.3 (Damian R Griffin, University of Warwick, 2021, personal communication).

The iHOT-33 has been validated for use in patients with FAI syndrome and is sensitive to change after treatment. The minimum clinically important difference (MCID) has been determined using an anchor and distribution-based approach in a group of 27 young active patients who were independent of the development population. Clinical change was determined using a global rating scale that asked patients whether their hip condition had improved, had deteriorated or had not changed since the previous assessment, using a single VAS. The MCID was 6.1 points. 33,34

The iHOT-33 and EQ-5D-5L have been adopted as the principal outcome measures by the UK Non-Arthroplasty Hip Registry. This registry is led by the British Hip Society (London, UK) and its use in all patients having arthroscopic FAI surgery is required by the National Institute for Health and Care Excellence (NICE). 15

In our pilot study, we tested the Non-Arthritic Hip Score and iHOT-33 as potential primary outcome measures, and found both to be easy to use and acceptable to patients. The extensive patient involvement in item generation, the availability of an independently determined MCID and the use of iHOT-33 as the principal outcome measure for the UK Non-Arthroplasty Hip Registry led us to choose iHOT-33 as the most appropriate primary outcome measure for a full trial.

Develop and test trial procedures

Protocols, eligibility criteria, patient information material and case report forms (CRFs) were designed for the pilot RCT and were available for the full trial. We interviewed 18 patients who had been treated for FAI syndrome to develop patient information sheets. These patient information sheets were scrutinised by a panel of expert patients with FAI syndrome who helped to improve the content and presentation so that they addressed patients’ key concerns and information needs, and provided explanations with appropriate language and detail. Twenty-eight clinicians, including surgeons, physicians and physiotherapists, also contributed to developing these procedures and documents.

Research Ethics Committee and national R&D approvals were granted for the pilot trial promptly and without any significant concerns. The majority of the recruitment sites were then able to complete local approval within 1 month of our site initiation visits. Typical causes for a delay to approval were identified within the first few sites, allowing these to be addressed in subsequent sites at a much earlier stage. This may help considerably to ensure that further sites in a full trial can obtain local R&D approval more quickly.

Conclusion of the feasibility study and pilot trial

We showed that a robust RCT of hip arthroscopy compared with best conservative care for patients with FAI syndrome was feasible, that patients and clinicians were willing to participate, that we were able to obtain ethics and R&D approval at multiple sites, and that the trial procedures we developed worked well. The pilot trial recruited successfully (70% recruitment rate) to the protocol that will be used for the full trial and these patients were, therefore, included in the full trial analysis.

Inclusion criteria

• Age ≥ 16 years (with no upper age limit).

• Symptoms of hip pain (including clicking, catching or giving way).

• Radiographic evidence of pincer- and/or cam-type FAI morphology on plain radiographs and cross-sectional imaging, defined as:

  • cam morphology – an alpha angle > 55°36

  • pincer morphology – a lateral centre-edge angle of > 40° or a crossover sign on the anteroposterior radiograph of the pelvis. 37

• The treating surgeon believes the patient would benefit from arthroscopic FAI surgery.

• The patient is able to give written informed consent and to participate fully in the interventions and follow-up procedures.

Exclusion criteria

• Evidence of pre-existing osteoarthritis, defined as Tönnis grade > 138 or a > 2-mm loss of superior joint space width on an anteroposterior pelvic radiograph. 39

• Previous significant hip pathology, such as Perthes’ disease, slipped upper femoral epiphysis or avascular necrosis.

• Previous hip injury, such as acetabular fracture, hip dislocation or femoral neck fracture.

• Previous shape-changing surgery (open or arthroscopic) in the hip being considered for treatment.

Preoperative protocol

Patients underwent routine preoperative care, which included an assessment of their general health and suitability for a general anaesthetic.

Perioperative protocol

Arthroscopic hip surgery was performed under general anaesthesia with the patient in a lateral or supine position. Arthroscopic portals were established in the central and peripheral compartment under radiographic guidance and in accordance with the surgeon’s usual practice. Shape abnormalities and consequent labral and cartilage pathology was treated. Bony resections at the acetabular rim and the head–neck junction were assessed by intraoperative image intensifier radiograph and/or satisfactory impingement free range of movement of the hip.

Postoperative protocol

Patients were allowed home when they could walk safely with crutches (usually within 24 hours). On discharge, all patients were referred to outpatient physiotherapy services for a course of rehabilitation, as per usual care for that surgeon. We did not specify a protocol for this postoperative physiotherapy, but recorded it using a treatment log. Postoperative physiotherapists were distinct from those providing PHT to avoid contamination between groups. Patients also had a postoperative MRI, which included a proton density volume acquisition sequence (for MRI protocol see Appendix 1).

Fidelity assessment

To ensure the fidelity of the surgery and to identify participants for a secondary analysis, a panel of international experts reviewed operation notes, intraoperative images and postoperative MRI scan to assess whether or not adequate surgery was undertaken (see Appendix 2). This panel included Mark Philippon (USA; then chairperson of the Research Committee of the International Society for Hip Arthroscopy), Martin Beck (Switzerland; one of the investigators credited with developing the early understanding of FAI), John O’Donnell (Australia; past president of the International Society of Hip Arthroscopy) and Professor Charles Hutchinson (UK; an expert in musculoskeletal radiology). The fidelity assessment process was tested in the pilot trial. The panel rated each surgical case as satisfactory, borderline satisfactory and unsatisfactory (see Appendix 1).

Training physiotherapists

Personalised hip therapy physiotherapists were trained in a FASHIoN PHT workshop and supported by the physiotherapy lead and research facilitator (NF and JS).

We developed and tested the 1-day workshop during our pilot trial. Following the initial PHT workshops during the feasibility study, the remaining workshops were delivered through the recruitment period from November 2014 to March 2016. The workshops included lectures, presentations, discussion of real cases and working through PHT progressions. A PHT manual and exercise sheets (for patients) were provided to all the PHT physiotherapists (see Appendix 1). Ongoing training and support was provided by the physiotherapy research facilitator and this included an initial site visit and monitoring visits. The purpose of the initial visit was to ensure that the treating physiotherapist fully understood the detail of PHT. The first visit was scheduled to occur after the first patients were randomised to PHT and before they had started treatment. Monitoring visits provided opportunities for further training and to conduct a source verification audit (see Fidelity assessment). Although PHT offered a framework to deliver best conservative care, the treatment was not a fully standardised regime. Physiotherapists were trained and encouraged to tailor their treatment to each patient, focusing on deficiencies identified in their assessment and based on the patients’ progression.

Pre treatment

Participants received a PHT information pack (see Appendix 1) that described what to expect during the course of their treatment. The first core component of PHT was an assessment of pain, function and range of hip motion.

Treatment

Personalised hip therapy had three further core components: (1) an exercise programme that had the key features of individualisation, progression and supervision; (2) education; and (3) help with pain relief (which may have included one X-ray or ultrasound-guided intra-articular steroid injection if pain prevented performance of the exercise programme). The intervention was delivered over a minimum of 12 weeks, with a minimum of six patient contacts. Some of the patient contacts were permissible using either telephone/e-mail for whom geographical distance prevented all contacts being carried out face to face. The number and frequency of the treatment sessions was at the discretion of the physiotherapist and was informed by the patients’ deficiencies and progression.

Post treatment

Typically, PHT was delivered over a minimum of 12 weeks. However, in situations in which the patient needed additional review, support or guidance, further sessions with the physiotherapists were permitted up to a maximum of 10 sessions over 6 months.

Fidelity assessment

To assess the accuracy of the PHT CRFs a source verification audit was undertaken to compare the physiotherapists’ hospital notes and the PHT CRF. Source verification was undertaken at each site and with 10% of cases sampled. The CRFs were graded as either a satisfactory or unsatisfactory reflection of the hospital notes. The source verification was undertaken by the physiotherapy research facilitator (JS). The findings of the source verification audit were fed back to the fidelity assessment panel.

The PHT CRFs were assessed to determine the fidelity of each intervention and to identify participants for a secondary analysis. This assessment was completed by the panel that developed the protocol for PHT, including Nadine Foster (Senior Academic Research Physiotherapist), Ivor Hughes and David Robinson (UK; Extended Scope Musculoskeletal Physiotherapists) and Peter DH Wall (Academic Orthopaedic Surgeon).

Treatment was rated as satisfactory, borderline or unsatisfactory. The panel assessed whether or not a sufficient number of treatments had occurred (at least six sessions in 12 weeks, but fewer than 10 sessions in 6 months), whether or not the treatment included all four core components of PHT and whether or not the exercise programme was individualised, supervised and progressive.

Primary outcome

The primary outcome was hip pain, function and hip-related quality of life measured using the iHOT-33 at 12 months following randomisation. The iHOT-33 is a validated hip-specific patient-reported outcome tool that measures health-related quality of life in young, active patients with hip disorders. 31 The iHOT-33 consists of the following domains: symptoms and functional limitations, sports and recreational activities, job-related concerns and social, emotional and lifestyle concerns.

We chose it following our feasibility and pilot study, as it is more sensitive to change than other hip outcome tools, it does not show evidence of floor or ceiling effects in patients undergoing hip arthroscopy and patients were involved extensively in item generation and, therefore, we can be confident that it measures what is most important to patients. The iHOT-33 has an independently determined MCID. The iHOT-33 is also used as the principal outcome measure for the UK Non-Arthroplasty Hip Registry, which is mandated for arthroscopic FAI surgery by NICE. 15,31

Secondary outcome measures

Arthroscopic surgery

Hip arthroscopy requires a general anaesthetic. The risk of complications from hip arthroscopy is about 1–2% and these include the following:

• Infection, which is thought to occur in less than 1 in 1000 patients.

• Bleeding, possibly causing bruising or a local haematoma.

• Traction-related complications. (To perform hip arthroscopy, traction is required to separate the hip joint surfaces. Sometimes after the procedure, the pressure from the traction can cause some numbness in the leg, but the numbness usually resolves within a few hours or days.)

• Osteonecrosis. (During surgery, the blood supply to the hip joint could be damaged; however, there are no reported cases of osteonecrosis following arthroscopic FAI surgery.)

• Femoral neck fractures. (This is also a very rare complication and would require a further procedure to fix the fracture.)

Personalised hip therapy

There are some small risks with pain medications and joint injection. However, the main risk is muscle soreness and transient increases in pain from the exercises that were undertaken.

Overview

A prospective within-trial cost–utility analysis was conducted to estimate the cost-effectiveness of arthroscopic surgery compared with PHT as treatment options for FAI syndrome. Costs were expressed in GBP (2016 price year) and health outcomes in quality-adjusted life-years (QALYs). The base-case analysis was based on the intention-to-treat population and conducted from the perspective of UK NHS and Personal Social Services. The time horizon covered the period from randomisation to end of follow-up at 12 months post randomisation. Costs and outcomes were not discounted because of the short 1-year time horizon adopted for this within-trial evaluation. Sensitivity and subgroup analyses were conducted to investigate the likely impact of alternative data inputs and assumptions on cost-effectiveness, and identify subgroups most likely to benefit from treatment. Findings are reported in accordance with the CHEERS (Consolidated Health Economic Evaluation Reporting Standards) guidelines. 51

Measurement of resource use and costs

Data were collected on (1) resource use and costs associated with delivery of the interventions, (2) health and social care service use during the 12 months of follow-up and (3) broader societal resource use and costs (e.g. private medical costs and lost productivity costs, such as lost income over the 12 months of follow-up).

Measurement of outcomes

The health-related quality of life of trial participants was assessed at baseline and at 6 and 12 months post randomisation using the EuroQol-5 Dimensions, three-level version (EQ-5D-3L) in the feasibility study, the EQ-5D-5L in the main trial and the SF-12 in both feasibility and main trial samples. 59–61 Responses to each health dimension were categorised as optimal or suboptimal with respect to function, with optimal level of function indicating no impairment (e.g. ‘no problem’ on the EQ-5D-3L dimensions) and suboptimal indicating any functional impairment. Between-group differences in optimal and suboptimal level of function for each health dimension were compared for each outcome measure using chi-squared tests.

The responses to each health-related quality-of-life instrument were converted into health-related quality-of-life weights (also referred to as utility weights) using established algorithms for each instrument. Utility values were generated using the UK value set for the EQ-5D-3L, the interim crosswalk value set for mapping from the EQ-5D-5L to the EQ-5D-3L, the newly published EQ-5D-5L tariffs for the EQ-5D-5L and the Short Form questionnaire-6 Dimensions (SF-6D) tariff based on SF-12 responses. 46,62–64

Quality-adjusted life-years were generated for each patient using the area under the baseline-adjusted utility curve, assuming linear interpolation between the three utility measurements. QALYs were generated for patients in the feasibility sample using utilities derived from EQ-5D-3L and SF-6D tariffs and for those in the main study sample using the EQ-5D-5L crosswalk tariff, the new UK EQ-5D-5L tariff and the SF-6D tarrif. 46,62–64 Health utility values and QALYs accrued over the 12-month follow-up were summarised by treatment group and assessment point and presented as means and associated SEs. Between-group differences were compared using the two-sample t-test, similar to the summary analyses of resource inputs and costs.

Cost-effectiveness analysis methods

Randomised controlled trial set up

We identified system features associated with successful recruitment, including:

• having a dedicated RA in attendance at clinics

• dividing recruitment information between the recruitment surgeon and the RA

• ensuring that referred patients arrived with expectations of receiving treatment for FAI, not necessarily having surgery

• running specific clinics for the target population.

During site initiation visits, the TMG and site research teams discussed how feasible these activities were at their site. Staff at most sites organised clinics that allowed RAs to be present and share the screening and approaching tasks with the recruiting surgeon. Referral letters and information about their diagnosis were modified to manage patient expectations. However, not many sites were able to host specific clinics for the target population.

Recruiter training

Analysis of pilot recruitment consultations pioneered in previous work72–76 led to the development of a six-step model (Figure 1). The objectives of the model were to improve informed consent and encourage participation in the FASHIoN trial. 3 This model guided the development of a recruiter training programme, which is explained in Action plans to promote informed consent and improve recruitment.

FIGURE 1.

Six-step model for recruitment to the FASHIoN trial.

After sites started recruitment, the QRI team used various methods to describe recruitment as it happened. The QRI team contributed to monitoring and supporting recruitment by providing feedback on their findings to the chief investigator and TMG, which provided the basis for a plan of action to improve it. 71 These methods were:

• mapping eligibility and recruitment pathways

• interviewing clinicians and RAs responsible for recruitment

• analysing audio-recorded recruitment appointments.

The QRI methods were not necessarily employed sequentially. The ethnographic nature of the QRI meant that the research was moulded to fit the needs of the project and was completed when theoretical saturation was reached (i.e. new data collection did not materially add to the findings). Observation of investigator meetings was carried out informally as one member of the QRI team (AR) worked closely with the TMG.

Qualitative research data were collected at 20 participating sites that recruited patients for the RCT. Two sites were excluded from this report, as they had only started recruitment 2 months before the end of the trial. In the following sections, we will report the data sets, analytic approach and results of each of the three QRI components.

Findings

Patient pathways from eligibility to recruitment varied across participating sites, which was reflected in different numbers of patients screened, approached and consented to be trial participants. Differences between sites are expected in a pragmatic trial; however, these differences were relevant to trial recruitment success. Therefore, we decided to compare regular sites and sites that struggled to recruit. Struggling sites were defined as sites that had a recruitment rate (i.e. the number of eligible patients recruited per centre per month) smaller than the target rate of 0.5 patients recruited per month. Table 4 highlights contrasting aspects of the recruitment pathway between the two site categories.

Patients were referred from other orthopaedic units and musculoskeletal triage systems. Most patients had a multitude of contacts with different professionals and very rarely came directly from general practitioners (GPs). In fact, struggling sites reported receiving a number of ‘complex cases’. These cases were generally referrals from other orthopaedic surgeons who had advised patients to have hip arthroscopy. This type of referral concerned recruiting surgeons, in particular, because patients had waited a long time to receive surgery, for example one PI said:

The patient’s mind has already been made up by people they’ve seen before me and generally by the nature of the condition, they’ve often seen a lot of people before they get to my clinic.

PI site 21

Adding to long waiting times, often suspected FAI patients did not have the correct imaging at the first appointment with the consultant. It was not possible for orthopaedic surgeons to confirm patients’ eligibility to participate in the FASHIoN trial and usually further tests had to be ordered. This meant that approaching patients about the trial had to be postponed until their diagnosis was confirmed. In addition, some patients had to be treated or, at least, examined for concurrent symptoms, as explained by one RA:

We have had a few [patients with] FAI but they have got trochanteric bursitis, or we have got ones with the back pain, so they first have to investigate that, and then they come back.

RA site 8

Although potential participants had been identified during screening and recorded on the screening log, some of these prospective participants were not then approached at follow-up clinics. The TMG published a list of patients waiting for FAI to be confirmed and asked RAs to find out their final diagnosis. This step in the recruitment pathway helped to identify a few eligible patients, among many others, who did not fulfil the eligibility criteria.

Other misunderstandings and errors in screening logs were made in the first 6 months of recruitment. For example, although a high-volume centre was registering every patient attending an orthopaedic clinic, independently of their diagnosis, other sites reported only patients who agreed to become FASHIoN trial participants. To avoid further errors, the screening logs were redesigned to facilitate a standard registration of eligible and approached patients.

In relation to the decision about who to approach, some regular sites’ research teams collaborated in the decision process, whereas struggling sites relied on the recruiting surgeon’s decision alone. The following quotations exemplified this contrast. The first quotation is from a site that recruited well, the second from one that was less successful:

So prior to the clinic on that Monday, I go through all the notes of the patients and have a look at their referral letters to see, because some of the referral letters, the GP or whoever else has referred them, may suggest that they suspect that there’s an impingement. So if there are any of those, I flag that up to [the surgeon] and I see if they’ve had an MRI done.

RA site 17

[The surgeon] is quite happy to discuss all the patients at the beginning of the morning to make sure we’ve found the right ones, you know, [the surgeon] checks them all through.

RA site 21

Research associates in regular sites had an active role in detecting potential participants and confirming their eligibility with recruiting surgeons. During the induction, RAs were trained to identify potential patients based on the referral letters by the trial clinical research fellow (EJD). Surgeons and RAs would look at their clinic list together and decide who to approach. Teams took ownership of the trial and often talked in collective terms about their activities, as exemplified in the quotations above. These teams created a reliable recruitment system that was repeated consistently.

In contrast, recruiting surgeons at struggling sites usually had the responsibility to check patient eligibility and introduce the study to potential participants. Surgeons would then instruct RAs to contact the potential trial participant at a later date. RAs often contacted patients by telephone. Occasionally, RAs would be contacted on the same day and would be able to talk to the patient. Sites often did not have a reliable recruitment system or a settled recruiting team. The following quotations illustrate these difficulties:

What tends to happen is [recruiting surgeon] sees them in clinic, he has a chat with them. He then e-mails me and says ‘I’ve seen Mr Smith, can you contact him? He is appropriate for the study’.

RA site 3

So he’ll e-mail me . . . He’ll introduce the study to the patients and he’s got information sheets that he passes on then I then approach the patient by phone call or letter.

RA site 14

Research associates who become proficient and confident at identifying potential participants had a pivotal role in organising recruitment. Their actions reduced the impact of contacts with other professionals (e.g. registrars) who interfered in the recruitment path:

The registrars will discuss [potential participants] with [consultant surgeon] first and I always make sure they have. I never take that for granted, because [consultant surgeon] is who has to sign the actual form.

RA site 10

Some recruiting surgeons would ask RAs to be present while they talked to patients about the trial, whereas other surgeons would spend time talking to patients alone first and then ask if they would like to talk to the RA. We recorded reluctance from surgeons to share recruitment responsibilities, as illustrated in this quotation:

I’m struggling with this trial in that you have to stop at a point to hand over to the research nurse.

PI site 11

We observed that recruiting surgeons contributed to promote patient equipoise when they worked with RAs to have a separate consultation with patients. A RA from a site that recruited consistently said:

The surgeon presents the [interventions] both as equally beneficial to the patient.

RA, site 12

Finally, often RAs at regular sites embraced the use of the six-step model enthusiastically, whereas in struggling sites the model was perceived as less useful. This is a quotation from an RA in relation to the model:

And I don’t read [the model] out word for word but it kind of makes sure you’re doing things in the right order and with the right stresses and, and going the right way. So yes, I find [the model] quite useful.

RA, site 2

Findings

Data interpretation focused on learning about recruitment difficulties that sites experienced, with the purpose of finding solutions in collaboration with the TMG. Reported difficulties coincided with those reported in past research,40 which was used to classify findings in three axial themes: (1) logistic difficulties, (2) recruitment experience and skills and (3) engagement with the trial.

Views on equipoise

Most recruiting surgeons expressed their agreement about the lack of strong scientific evidence in favour or against hip arthroscopy. Surgeons supported the conduct of a RCT in this area.

Being surgeons, most recruiters preferred surgical intervention for FAI syndrome, but reported that they were able to ‘suspend’ their clinical judgement. Some recruiters seemed rather comfortable offering the trial to patients:

I feel [FAI] patients don’t get enough physiotherapy.

PI site 20

I’ve got no qualms about offering [patients] [PHT] at the start of the study.

PI site 21

Other recruiters (two PIs and eight RAs) expressed concerns because they believed that those in the surgical arm were worse off:

If the physio[therapy] didn’t work you’ve always got the option of the arthroscopy, but then if you opt for the arthroscopy, OK they can say there’s still the physio afterwards but then you’ve put yourself through all that trauma.

RA site 11

Finally, other recruiters assumed a laissez-faire attitude to patient trial participation decisions:

I mean basically I have got my own electronic [list], which has got their phone numbers and address and everything on it and then as I work through they either become consented, which one lady has, or they drop out as not interested.

RA site 3

Different levels of equipoise were observed at regular and struggling sites. Most revealing was the lack of surgeon engagement in struggling sites:

I feel the consultants probably need to be more on board with it and sell it a bit more. So that’s been challenging because we’re new to research . . . I’m not usually a very pushy person but I feel that I have to do that to the patient as well as a consultant, in a nice way, you know.

RA site 14

Procedure

Participating sites were asked to produce as many recordings as possible of two pivotal appointments in the recruitment to this trial. The first appointment to be recorded, named the ‘diagnostic consultation’, was the appointment when patients were told about their diagnosis. We advised recruiting surgeons to present the trial using the first four steps of the recruitment model (i.e. explain the condition, reassure about receiving treatment, introduce uncertainty and explain study purpose), to answer patient questions and to invite patients to talk to RAs about the trial.

The second appointment was the ‘recruitment consultation’, defined as the appointment at which randomisation was discussed. The training of RAs emphasised using the six-step model to structure information given to patients about the trial during the recruitment appointment.

The TMG explained and promoted the QRI at site initiation visits. All materials for recording consultations (i.e. patient information sheets, consent forms for audio-recording, digital audio recorders and instructions for the operation of the recorder and the naming and transferring of data to the QRI team) were given to a main contact within the site team, generally the RA assigned to the trial.

Prior to their consultation with the surgeon, RAs gave patients who could potentially become trial participants at clinics the patient information sheet for the audio-recording of consultations to read. If patients agreed to become trial participants, they were asked to sign a consent form. RAs then proceeded to record the following appointments as per protocol. The recordings were then anonymised and sent securely to the QRI researcher for analysis.

A coding manual was developed and applied directly to the audio-recordings (not transcripts) using NVivo software. After the analysis, the QRI researchers decided what confidential feedback would be given to the recruiters. Teams were assured that the feedback to them was going to be confidential and positive (not critical). Other issues were fed back to the RCT chief investigator/TMG, or were used anonymously in training programmes.

Analytic approach

The sample was purposively selected to include patients who agreed and did not agree to take part in the FASHIoN trial. Good-quality recordings of recruitment consultations (and diagnostic consultations, when available) were analysed with a modified version of the Quanti-Qualitative Appointment Timing method that has been pioneered in previous studies. 75

Using the Quanti-Qualitative Appointment Timing approach, we determined the percentage of time recruiters spent in each of the six steps of the recruitment model and the sequence in which they were used by recording when a recruiter used a step for the first time. We then evaluated appointment content for unbalances and untoward emphases that may upset patient equipoise using focused conversation analysis. In addition, patient questions, concerns and preferences were coded and frequencies tabulated.

Two coders tested the reliability of the coding system (i.e. inter-rater reliability). Consultations of six individual patients were coded and Cohen’s kappas calculated. After reliability checks were satisfactory, we applied the coding to the remaining consultations.

Findings

We received 153 audio-recordings corresponding to 129 individual patients from 11 sites between September 2014 and March 2016. Table 5 contains the characteristics of the audio-recordings, patients and clinicians in this sample. Patient and recruiter characteristics did not differ from those of the trial participants as a whole.

Recordings selected for analysis comprised a combination of patients who agreed to take part in the trial (n = 44) and those who declined (n = 29). The average length of recruitment consultations was 16 minutes 50 seconds and the average length of diagnostic consultation was 7 minutes 57 seconds. The minimum number of recordings per site was two and the maximum 22. Other recordings were excluded because they were incomplete (n = 17), patients in the recordings were not eligible (n = 12), patient diagnosis was uncertain (n = 8) or the recording included more than one recruiter (n = 9).

Coding reliability

The QRI researcher and an independent qualitative researcher achieved substantial agreement (κ = 0.67) in the coding of a subsample of six consultations. Both the QRI researcher and the independent qualitative reached almost perfect agreement when identifying statements explaining details of hip arthroscopy, PHT and randomisation (κ > 0.8). Coders disagreed more often (κ < 0.4) when identifying statements of reassurance (step 2) and uncertainty (step 3), in part, because reassurance was seldom used and uncertainty often appeared paired with study purpose (step 4).

Diagnostic consultations

We received 34 diagnostic consultations and in 22 of these consultations patients became trial participants. The remaining 12 participants declined participation. We observed two different approaches to diagnostic consultations. Some surgeons spent time going through trial information with patients, whereas others made a brief intervention and handed over the potential participant to the RAs.

Figure 3 shows the distribution of the time spent on different elements of the consultation, comparing consultations in which the patient agreed to participate with those in which they did not. Recruiting surgeons spent more time framing the decision using the first four steps of the model in trial participant consultations than in non-participant consultations. This was also the case in relation to answering patient questions and providing information about PHT, hip arthroscopy and RCT procedures. These recruiters spent less time dealing with patient concerns and preferences than their colleagues in non-participant consultations.

FIGURE 3.

Percentage of time dedicated to specific recruitment tasks in diagnostic consultations.

In relation to the sequence in which information was given, we observed that surgeons used the first four steps before explaining study procedures in 16 consultations and in 10 of these of consultations patients became trial participants. When the sequence was reversed, four out of five patients took part in the study. Surgeons did not mention randomisation in 13 consultations and in eight of these consultations patients became trial participants.

These results suggested that surgeons managed to convert patients into participants more often when they framed the decision of trial participation, answered patient questions about their medical condition and treatment alternatives, and did not go into much detail about study procedures or randomisation.

Recruitment consultations

Our results indicate that the majority of recruiters used the model to structure their consultations, as per training. On average, 72.8% (SD 16.5%) of the consultation time was spent introducing information based on the six-step model. The remaining time was spent taking medical history, giving instructions and chit-chat.

Figure 4 shows time distributions of recruitment consultations. More time was spent talking about study procedures and the non-surgical arm in trial participant consultations than in non-participant consultations. Furthermore, patients appeared to have fewer questions, concerns and preferences in the former than in the latter.

FIGURE 4.

Percentage of time dedicated to specific recruitment tasks in recruitment consultations.

Most recruiters (n = 47, 64%) used the information delivery sequence suggested by the model. Recruiters framed patient decisions using the first four steps of the model and then proceeded to introduce information about randomisation. Recruiters did not use this framework in 20 cases (27%). However, independent of the chosen sequence, 50% of approached patients became trial participants.

In relation to the content of recruitment consultations, we observed little variation on topics because recruiters were provided with extensive training. The content of consultations became highly standardised in a script available to recruiters in cue cards (see Training site teams on recruitment strategies and trial-specific information).

Our observations and recruiters’ reports indicated that recruitment to this trial depended on two pivotal aspects of recruitment: (1) recruiters needed to make an argument in favour of PHT and (2) recruiters needed to describe randomisation in a way that was acceptable to potential participants.

Making an argument in favour of PHT was especially challenging when patients have had previous hip-specific physiotherapy. Recruiters succeeded at promoting patient equipoise when they knew details about PHT and were equipped to answer patient questions about this treatment arm. Recruiters used their knowledge to dispel misconceptions or calibrate treatment expectations by given details about how a treatment may affect the person at an individual level.

The following extract is an example of how this strategy worked. The recruiter offers an interpretation of patient preference and confronts this assumption based on her own clinical experience, which the patient accepts:

Yeah and it’s funny maybe I don’t get that when I – and I find like picking up my son if I go like that hhh I don’t have any power, I don’t have any – is that sort of thing

[Inaudible]

But I think . . .

The thing is, if you just must think, because if you think about the surgery as mechanical . . . just taking the bump away and then it’s away . . . but I also work in the clinic follow-up of people post surgery

Hmmm

And I did find that some of the high-level sports people it’s taking us a year to a year and a half to actually get them – and some people you don’t, some people are always off, so you do get worse off so ehm . . . after surgery it’s not all just straightforward

No, no, I understand.

Contrast between previous physiotherapy and the new PHT arm could not always be achieved, either because recruiters used statements that were too general or irrelevant to patient needs, or simply because they did not know how different these approaches really were. The following extract exemplifies this type of exchange. The patient declares that he has already had physiotherapy and the recruiter attempts an open question but changes it for a closed one. The patient rejects the assumption and emphasises the difference. The recruiter moves to explain general aspects of PHT, but the patient rejects this. The recruiter persists but gives up at the end:

I[ve already said] I had months of physio already though

Ahh yeah What kind of phys[io] is that through your GP?

No it was through the army

OK

Private physio

Right ahh . . . So what the physio is doing ahh is . . . the physio is

a personalised hip therapy programme so you’d come to the hospital and see Mr [surgeon]’s specialist physio who deals with just hips, nothing else ehm which is where you get the personalised structure from ahh and what she would do with you is work with you to build up your muscles ehm and your hip [flexes]

I’ve already been doing [that]

And everything around . . . yeah ehm look at how ehm where your pelvis is positioned? So if you tilt your pelvis slightly forward, slightly back to get you in a good postural positioning [laughs] I’m preaching to a convert here, ain’t I?

Describing randomisation in a way that was acceptable to patients often challenged recruiters’ equipoise. At times, conversations became awkward, hindering recruitment. The following extract illustrates this effect:

I’d still do your study but I go the surgical route

That’s a – that’s another proble(h)m ehmm . . . to make it a fair test ahh . . . we . . . randomly allocated . . . your . . . option . . . so I don’t get a say in what you get, you don’t get a say and Mr [surgeon] doesn’t get a say

[Sniggers]

So leave it to chance of which option you’d be given

I can’t do that because I want to go back

I know

I’m in [timescale]

Have they said what sort of, what sort of

6 months to a year from when I get my discharge papers

That’s your window of getting back in?

No that’s how long I’ve gotta wait 6 months to a year

OK

Before I can reapply

OK

So I want to get my fitness back.

Laughter in the patient response can be interpreted as responding to the awkward discussion about the surgeon not having a say, which the recruiter framed as ‘another problem’. The patient continues to be concerned about the timescale of their treatment and uses this reason to opt out of the trial.

Finally, we noticed that step 2 of the model (i.e. reassurance about receiving treatment) was seldom used by recruiters. Only 23 of the 73 recordings contained reassuring-type statements and most of the patients from these recordings became participants (74%). Recruiters expressed concerns about using reassuring-type statements, as they may have been construed as undue influence or coercion. Nevertheless, it can be argued these statements were ethically valid because the FASHIoN trial was a superiority trial and, therefore, it was assumed that both treatments were effective. An example of a reassuring statement is as follows:

[Surgeons] know that both [treatments] work, treatment, trial or whatever, you’re going to get treatment anyway, so physio and surgery, they both work.

RA site 4

Improving screening of eligible patients

A simplified screening form was introduced to correct screening misunderstandings and to obtain accurate estimates of patient recruitment. RAs received detailed instructions on how to fill out the new screening logs.

Research associates were asked to submit screening data and upcoming recruitment clinic dates on monthly basis, including whether or not a RA would be available to attend clinics. This information was reported during the monthly TMG meetings.

Reports to the TMG included an approach percentage and a conversion percentage for each site. The approach percentage was the percentage of eligible patients who had been asked to take part in the trial. The conversion percentage was based on the number of approached patients who became trial participants. These two percentages allowed the trial team to identify where potential problems in the flow of patients may be. For example, two sites recruited 0.6 patients per month; however, the sites differed on their approach and conversion percentages. One site had a low approach percentage because patients travelled from Ireland to receive treatment and, therefore, these patients were ineligible to participate in the trial. The other site had a nearly 100% approach, but it did not convert patients to participants. There was not much that could be done with the composition of clinics in the first site, but the second team was offered further training on talking to patients about the trial.

The majority of participating site teams responded positively to these changes and RAs developed their own systems to report their recruitment numbers regularly. Screened patients almost doubled between September and January (n = 352) and the following 4 months after these changes were implemented (n = 681). The number of approaches and recruitment conversions increased slightly from 27 to 30 at the same sites.

Facilitating recruiting sites’ participation and engagement

Information collected through interviews and screening logs for each site was organised into cases. Each case detailed team structure and number of patients screened, eligible, approached and recruited, as well as a description of any particular difficulties at logistic, engagement or skill level. Cases were discussed with the chief investigator and TMG in weekly meetings. A summary was presented and strategies to address particular issues discussed during the TMG monthly meetings.

Training site teams on recruitment strategies and trial-specific information

A comprehensive training programme was designed to support site teams in optimising their recruitment opportunities. Different activities were organised from the start to the end of the recruitment period. These activities responded to observations of common barriers and good practice across sites. A list of activities in chronological order is presented below.

Recruitment rates overview

The FASHIoN RCT aimed to recruit 344 participants. Recruitment was planned to take place across 25 centres over a 20-month period. Centres would be open and recruiting 6 months after recruitment started. Forty-two patients had already been recruited in the internal pilot. The main RCT, therefore, needed to recruit an additional 302 patients.

Recruitment targets were set, assuming centres would comply with this schedule. Figure 5 shows the original recruitment target line. However, there were considerable delays in opening centres, which had an impact on recruitment targets. An amended recruitment target line was estimated based on the actual numbers of centres open per month for the main trial. The recruitment target of 302 patients was achieved in 22 months and included 23 centres.

FIGURE 5.

Target vs. actual cumulative recruitment: the FASHIoN trial.

Recruitment figures roughly followed the amended target line until March 2015, after which the actual recruitment figures started to exceed the amended target. Recruitment rates steadily remained above the amended target for the remainder of the recruitment period.

Research associates’ survey on recruiter training activities

We conducted an online survey of FASHIoN trial RAs at the end of the recruitment period. The main aim of this survey was to gather views of a diverse group of recruiters about the strengths and weakness of recruitment training and trial conduct. We also sought advice about how to improve the design of future orthopaedic trials.

Recruitment training activities

We asked recruiters how useful they found each of the seven training activities described in Phase II. We obtained many ‘not applicable’ responses recorded in activities that happened once. The results are as follows.

Study procedures

In relation to how difficult or easy the study procedures were, the survey covered all aspects, from setting up the trial to follow-up procedures. Procedures catalogued as easy or neutral included communicating with the TMG, working with the local PI, completing CRFs and obtaining clinical data about participants. A small number of RAs reported difficulty in setting up the study, attending clinics, identifying eligible patients, approaching eligible patients and obtaining informed consent.

Two aspects of the study procedures that were rated as difficult by RAs were (1) having a separate recruitment consultation after the diagnostic consultation (n = 7, 28%) and (2) recording consultations (n = 7, 28%). Notably, the RAs said that these difficulties differed from other trials in which they had been involved. However, overall, RAs were satisfied with the study procedures.

Research associates were very satisfied with the trial management support, specifically in relation to dealing with treatment-specific queries, general trial queries and telephone and e-mail contact with the trial team. A RA made the following comment:

The trial management team were excellent in this study. Easily accessible and very helpful and professional.

Study documentation

Research associates’ ratings of study documentation (including information about FAI and treatments, the trial site file, the patient information sheet, screening logs, CRFs, the trial manual and the newsletter) ranged from good to excellent in all questions. A comment from one RA reads:

Having worked on many studies now, I appreciate how hard is to design and produce such high quality of documentation.

When we asked RAs to compare FASHIoN trial procedures and documentation with that of other trials they had been working on, ratings ranged from good to excellent. We noticed a few not applicable responses from RAs who had not worked in other trials.

Overall, RAs reported satisfaction with study procedures and documentation, as well as with the range of recruitment training activities. The trial was well resourced in terms of staff and budget, and this allowed consistent communication between the trial and site teams and resulted in highly professional documentation.

Emphasis on RAs talking to patients in a separate recruitment appointment and the use of a recruiter training programme made the FASHIoN trial somewhat different from other orthopaedic trials. It is not surprising that RAs found these extra activities more challenging than other common trial procedures. For example, there were suggestions to remove recordings from future trials, as RAs found recordings awkward and feedback was inconsistent and may have felt critical at times. However, RAs appreciated the note cards, which could not have been created without data from these recordings. Another strength of our approach was having a variety of activities for recruiting staff who usually have a high turnover. This meant that recruiters were well supported, as evidenced in their feedback.

Recruitment

The number of patients approached and recruited by site is shown in Table 6. Overall, 94% of eligible participants were approached, 61% of whom consented to participate in the trial. In total, 351 participants were recruited. There were three participants who were randomised; however, post randomisation, these participants were found to not meet the inclusion criteria and were excluded and withdrawn from the study (referred to as post-randomisation exclusions). For the remainder of the report, data are presented for the 348 participants who were not post-randomisation exclusions. In total, 171 patients were allocated to hip arthroscopy and 177 patients were allocated to PHT.

Arthroscopic surgery

Personalised hip therapy

Primary outcome

The iHOT-33 score increased between baseline and 12 months in both groups (Table 16 and Figures 8 and 9). iHOT-33 scores improved from a mean of 36 points at baseline to 50 points at 12 months in the PHT group and from a mean of 39 points at baseline to 59 points at 12 months in the hip arthroscopy group. In the primary intention-to-treat analysis at 12 months, the adjusted estimate of treatment effect measured with iHOT-33 was 6.8 (95% CI 1.7 to 12.0) in favour of the hip arthroscopy group compared with the PHT group. The p-value of 0.009 indicates that there is evidence for a statistically significant difference (the 95% CI includes the prespecified MCID and does not include zero).

FIGURE 8.

Box plots of iHOT-33 baseline and follow-up scores.

FIGURE 9.

Overall trend in iHOT-33 unadjusted mean scores and 95% CIs.

Secondary outcomes

There were no were significant differences between treatment groups at 6 or 12 months in the EuroQol-5 Dimensions (EQ-5D) or SF-12 (Figures 10–13 and Table 17).

FIGURE 10.

Box plots of EQ-5D-5L baseline and follow-up scores.

FIGURE 11.

Box plots of EQ-5D VAS baseline and follow-up scores.

FIGURE 12.

Box plots of SF-12 (physical component score) baseline and follow-up scores.

FIGURE 13.

Box plots of SF-12 (mental component score) baseline and follow-up scores.

Per-protocol analyses

Two per-protocol analyses were conducted. The first per-protocol analysis was for patients who received surgery only (n = 144) or PHT only (n = 154), irrespective of the quality of the treatment they received (Table 18). The adjusted estimate of treatment effect measured with iHOT-33 was 8.2 (95% CI 2.8 to 13.6) in favour of hip arthroscopy.

A second exploratory per-protocol analysis was conducted that included only those participants who received their allocated intervention and whose treatment was deemed to be of a satisfactory quality by the respective review panels. In this analysis, 95 surgical cases and 106 PHT participants were included (Table 19). In this exploratory per-protocol analysis, the adjusted estimate of treatment effect measured with the iHOT-33 was favour of surgery (Table 20).

Missing outcome analyses

There was a low level of missing item-level data in all patient-reported outcome measures at all time points. After imputation for missing data, the adjusted estimate of treatment effect was similar at 6.6 (95% CI 1.7 to 11.4) points in favour of hip arthroscopy (Table 21).

Subgroup analyses

Planned subgroup analyses were conducted for those participants with cam, pincer and mixed-type FAI syndrome and those aged < 40 years and ≥ 40 years.

There was no evidence of a subgroup effect, with the adjusted estimate of treatment effect measured by the iHOT-33 being as follows: 5.0 (95% CI –1.2 to 11.3) in participants aged < 40 years, 10.9 (95% CI 1.7 to 20.1) in participants aged ≥ 40 years, 8.3 (95% CI 2.5 to 14.2) in participants with cam morphology, 1.1 (95% CI –11.5 to 13.7) in participants with mixed cam and pincer morphology and 4.0 (95% CI –14.6 to 22.7) in participants with pincer morphology, in favour of surgery (Table 22).

Sensitivity analyses

In a post hoc analysis, there was no significant difference in iHOT-33 score at 12 months for patients in the hip arthroscopy group who were treated within 6 months of randomisation compared with those treated ≥ 6 months after randomisation (0.9, 95% CI −10.7 to 8.8).

Complications at the 6-week assessment

A total of 138 and 146 participants following surgery and PHT, respectively, returned a 6-week complication log. The most commonly reported complication in both groups was muscle soreness (Table 23). Eight (6%) patients reported a superficial wound infection following surgery, but antibiotics were prescribed in only four (3%) cases. Thirty-five (24%) participants reported groin, leg or foot numbness after surgery.

Complications reported at the 6- and 12-month follow-up time points

Complications at 12 months post randomisation are reported in Table 24.

At 12 months, seven SAEs had been reported, six of which were in patients who received surgery (Table 25). Two patients allocated to surgery required hospitalisation and four required further medical intervention. One patient who received PHT required hospitalisation.

Three-year follow-up: further procedures or physiotherapy

In the 24-month period from 12 to 36 months post randomisation, about one-third of all patients had a further procedure and about half had further physiotherapy sessions (Tables 26 and 27). A (further) one-quarter of PHT patients chose to have a hip arthroscopy and 12% of hip arthroscopy patients had a further hip arthroscopy. Two per cent and 7% of the patients in surgery and PHT groups, respectively, had a hip replacement.

Cost of personalised hip therapy

Table 29 summarises PHT attendance by type of consultation, impingement classification, missed appointments and recruitment site. A total of 1219 physiotherapy appointments were offered to 166 (94%) of the 177 patients in the PHT group. Of these patients, 909 (75%) were face-to-face consultations, 38 (3%) were telephone consultations, seven (0.6%) were e-mail contacts and 256 (21.2%) were recorded as unknown or missed appointments. A total of 166 (94%) of the 177 PHT patients had at least one physiotherapy contact, 105 (59%) had six or more contacts (as recommended in the treatment protocol for PHT) and 11 (6.2%) did not receive the intervention at all. Excluding missed appointments, the mean number of physiotherapy contacts per patient was 5.6 (range 1–11) and the mean duration of contact for all sessions attended was 178.2 (range 30–375) minutes per patient. Among those who had the recommended six or more physiotherapy contacts, the mean number of contacts per patient was 7.7 (range 6–11) and the mean total duration across all sessions attended was 222.6 (range 140–375) minutes per patient.

A total of 225 out of the 1219 PHT appointments offered were missed by 94 patients (53% of the PHT sample), giving an overall non-attendance rate of 18.5%. Among the 94 patients who missed an appointment, the mean number of appointments missed was 2.3 per patient, the mean total duration of all appointments missed was 75.4 minutes and the mean total cost of missed physiotherapy time was £65.60 per patient. The mean cost of PHT was £192.16 per patient, including missed appointments, and £155.01 per patient, excluding missed appointments. Among those attending six or more consultations, the mean cost of PHT was £196.56 per patient, including missed appointments, and £193.65 per patient, excluding missed appointments.

Resource use and cost of arthroscopic surgery for FAI

Resource use data associated with the delivery of arthroscopic surgery for FAI were collected for 47 (27%) of the 173 patients in the surgery group and one patient in the PHT group across 18 (82%) of the 22 orthopaedic centres participating in the study. (Note that the one patient in the PHT group was randomised to PHT, but had surgery immediately after the 12-month follow-up period had ended.) Six (13%) of 46 patients were excluded because data on duration of surgery and/or post-surgical inpatient length of stay were missing. The mean age among the remaining 40 patients included in the micro-costing of the surgery sample was 34 (range 18–54) years, 60% were male and 68% had the cam impingement type. Compared with participants in the surgery arm of the trial and who were not included in the micro-costing sample (Table 30), those patients included were, on average, 2 years younger (mean age 34 vs. 35 years; p = 0.405), as likely to be male (60% vs. 58% male; p = 0.997) and had similar presentation with respect to type of impingement (68% vs. 78% cam; p = 0.118).

Estimates of resource use associated with the surgery and sources of unit cost data for resource inputs are presented in Appendix 3, Table 34 and Table 48, respectively. Resource categories included operating theatre use/time, clinical staff, anaesthetic drugs, disposable surgical equipment and implants, and inpatient length of stay for post-surgical recuperation. The mean duration of surgery was 2.12 (range 1–3) hours and the mean length of inpatient stay was 1.6 (range 1–3) days (see Appendix 3, Table 42). The composition of the surgical team/staff remained broadly similar across centres and consisted of two surgeons (a consultant and an assistant or registrar), one anaesthetist, a radiographer, one or two nurses, two operating department practitioners and a health-care assistant.

Unit costs of clinical staff time were obtained from the PSSRU Unit Costs of Health and Social Care 201652 services and ranged from £28 per hour for a health-care assistant to £137 per hour for consultant surgeon (including qualifications and overheads). The running cost of an operating theatre was estimated based on data published by the Information Services Scotland. 54 The Information Services Scotland data showed that a total of £157,150,194.90 was allocated to operating theatres across Scotland for the 2015–16 financial year. The total number of theatre hour use recorded for the same period was 526,145.14 hours, generating a running cost of £298.68 per hour (see Appendix 3, Table 42). Allocated costs included non-clinical staff costs, property and equipment maintenance, domestics and cleaning, heating, lighting and power, and capital charges (e.g. the purchase of new equipment). 55 Equivalent data on the running costs of operating theatres in England and Wales were not publicly available and so only the Scottish data were used in our cost calculations. An inpatient stay was assumed to cost £332.77 per day, which is the excess bed-day cost for elective orthopaedics procedures in the 2016 reference costs schedules. 77 The unit costs of anaesthetic drugs were obtained from prescription cost analysis database,53 electronic searches of the BNF58 and searches of the literature, when necessary. Unit cost of syringes and needles and other medical consumables were obtained from online sources when more direct NHS sources were unavailable (see Appendix 3, Table 42).

The quantity and unit costs of disposal equipment and implants such as blades, shavers, burrs and sutures used during surgery are presented in Appendix 3, Table 44. The cost of implements ranged from £35 per item for a Smith & Nephew-supplied banana blade (Smith & Nephew, London, UK) to £349.32 per item for a TAC-S radio-frequency probe (Smith & Nephew), and were primarily extracted from the NHS supply chain catalogue held within hospital finance and procurement departments. 56

Cost estimates for the surgery resource use are summarised in Table 31 by resource category and recruitment centre. The mean cost for the surgery ranged from £2286 per patient at North Bristol NHS Trust to £4076 per patient at the Royal National Orthopaedic Hospital. Across all the centres, the overall mean cost was £3042, 35.3% of which was staffing costs, 23.5% disposal surgical equipment and implants, 19.4% theatre running costs, 17.8% inpatient costs and 4% represent the cost of anaesthesia, including drugs, syringes and needles. These figures represent estimates associated with delivery of hip arthroscopy and do not account for pre- and post-surgery consultations, diagnostic scans and post-surgical rehabilitation costs. These additional costs were, however, taken into account in the cost-effectiveness analysis as follow-up costs if they were reported by trial participants at the 6- and 12-month assessment points.

Follow-up resource use and costs

Descriptive summaries of self-reported health and social care service use are presented in Appendix 3, Table 45, by assessment point, resource category and treatment group. The proportion of missing data across all categories of resource inputs ranged from 1.2% at baseline to 17.5% at 12 months post randomisation in the surgery group and from 0.6% at baseline to 26.6% at 12 months post randomisation in the PHT group. Across the majority of resource categories, the baseline values displayed in Appendix 3, Table 45, suggest no more than 8.2% and 10.7% of the patients in surgery and PHT groups, respectively, reported use of health and social care services in the 3-month period prior to randomisation. The only exceptions were GP attendance, medication use and orthopaedic outpatient attendance in the same period. Slightly over one-third of patients (36% in the surgery group and 41% in the PHT group) reported attending at least one GP consultation, with a similar proportion of patients reporting being prescribed medication and 57% attending an orthopaedic outpatient appointment. Overall, the treatment groups were generally balanced with respect to reported resource use across all categories of health and social care services at baseline.

Rates of service use over 12 months of follow-up were also broadly similar between the two treatment groups for most resource categories. However, at the 6-month assessment point, the surgery group reported, on average, 0.035 (95% CI 0.011 to 0.118; p = 0.038) more inpatient hospitalisation days for the category ‘other inpatient admissions’, 0.362 (95% CI 0.116 to 1.032; p = 0.016) more orthopaedic outpatient consultations, 0.499 (95% CI 0.224 to 1.588; p = 0.002) more outpatient physiotherapy attendance for the category ‘your hip/leg’, and significant use of prescribed medication, walking aids and adaptations, than the PHT group. At the 12-month assessment point, the surgery group also reported, on average, 0.596 (95% CI 0.089 to 2.544; p = 0.002) more inpatient days, 0.041 (95% CI 0.007 to 0.129; p = 0.030) more day case admissions and 1.789 (95% CI 1.379 to 4.554; p < 0.001) more outpatient physiotherapy attendances than the PHT group. Compared with the PHT group, attendance rates at outpatient orthopaedic and physiotherapy clinics were also higher among the surgery group at 12 months post randomisation, as were the use of clutches, dressing aids and number of prescriptions received.

Unit cost of resource inputs together with corresponding sources of unit costs are presented in Appendix 3, Table 46. Appendix 3, Table 47, presents health and social service costs generated by assigning the unit cost data to resource inputs summarised by resource category, treatment group and assessment point. The mean total cost across all categories of resource use at baseline, covering the 3-month period prior to randomisation, was £502.12 in the surgery group and £508.53 in the PHT group, generating an unadjusted mean total cost difference of –£6.41 (95% CI –£235.01 to £246.59; p = 0.880). Total cost across all resource categories based on the 6- and 12-month data was generally higher in the surgery group than in the PHT group, but the unadjusted between-group difference in costs was only statistically significant at the 6-month assessment point. Over the 12 months of follow-up, the mean total cost was £1640.91 in the surgery group and £941.02 in the PHT group, generating a statistically significant unadjusted cost difference of £699.88 (95% CI £274.36 to £1121.23; p < 0.001).

Private health expenditure, lost income and other costs

Private health-care service use and related out-of-pocket expenses by patients, friends and family members in the 3-month period prior to randomisation and the 12-month post-randomisation period are summarised in Appendix 3, Tables 48 and 49, respectively. The proportion of missing data across all categories of resource inputs ranged from 1.2% at baseline to 18.1% and 17.5% at 6 and 12 months post randomisation, respectively, in the surgery group and from 0.6% at baseline to 26.6% at 12 months post randomisation in the PHT group. Private health-care services included private physiotherapy consultations, purchase of over-the-counter medication and purchase of walking aids, such as crutches, sticks and specialised shoes. There was no statistically significant difference between the two treatment groups across all categories of private health-care utilisation at baseline and over the 12 months of follow-up (see Appendix 3, Table 48). The mean cost of private health utilisation in the 3-month period prior to randomisation (baseline figures) was £10.89 in the surgery group and £14.95 in the PHT group, generating an unadjusted cost difference of –£4.06 (95% CI –£25.07 to £11.16; p = 0.760) (see Appendix 3, Table 49). Over the 12 months of follow-up, private health-care and out-of-pocket expenditure increased from the baseline figures to about £32.67, on average, in the surgery group and £27.44 in the PHT group, generating a mean cost difference of £5.23 (95% CI –£23.55 to £32.41; p = 0.704).

Appendix 3, Table 46, presents a summary of additional (indirect or non-health service) costs, for example the value of lost income incurred by patients and their families for the 3-month period prior to randomisation (baseline figures) and over the 12-month period of follow-up. Only costs incurred as a result of inability to work or perform tasks because of their hip pain were included in these additional (indirect) cost calculations. These costs included lost earnings, help with house work, child care and expenditure on specialised equipment. The mean additional cost reported by patients for the 3-month period prior to randomisation was £164.15 in the surgery group and £111.49 in the PHT group, generating a mean additional cost difference of £52.66 (95% CI –£189.36 to £470.01; p = 0.748). Over the 12 months of follow-up, the corresponding costs reported increased to £1143.20 in the surgery group and to £184.81 in the PHT group, generating a cost difference of £958.39 (95% CI £219.74 to £2001.32; p = 0.004). Lost income accounted for approximately 92% of the additional costs in the surgery group, but only 60% of additional costs in the PHT group at baseline. Over the 12 months of follow-up, lost income accounted for approximately 83% and 76% of the additional costs in the surgery and PHT groups, respectively.

Total economic costs

Table 32 summarises the total costs during the 12-month follow-up period by treatment group, cost category and cost perspective. Among the complete cases, the mean total cost from a NHS and Personal Social Service Perspective was £3712.77 in the surgery group and £1282.93 in the PHT group, generating an unadjusted cost difference of £2429.84 (95% CI £1865.92 to £2911.70; p < 0.001). Surgery costs accounted for approximately 72% of total costs in the surgery group, whereas the treatment costs (including surgery costs for PHT patients who had surgery) accounted for only 28% of the overall costs in the PHT group. The corresponding mean total cost estimated from a societal perspective was £4993.84 in the surgery group, of which 53.5% was surgery costs, and £1481.96 in the PHT group, of which 24% was accounted for by treatment costs, generating an unadjusted cost difference of £3511.88 (95% CI £2490.57 to £4784.53; p < 0.001).

Benefit payments

Patient self-reports of benefit payments, pensions and statutory sick pay are presented in Appendix 3, Table 51. The proportion of patients with missing data on benefit payments ranged from 2.3% at baseline to 17% and 14% at the 6- and 12-month assessment points, respectively, in the surgery group and from 0.6% at baseline to 20.3% and 22% at 6 and 12 months post randomisation in the PHT group. Examples of benefit payments that were reported during follow-up include Attendance Allowance, Income Support, Housing Benefit, Carer’s Allowance, Child Tax Credit, Council Tax Reduction, Employment and Support Allowance, and Disability Living Allowance. The proportion of patients in receipt of at least one of these payments increased from 10.6% at baseline to 17.6% over the 12 months of follow-up in the surgery group, but decreased in the PHT group from 6.2% at baseline to 5.7% over the 12 months of follow-up. The average reported payment per patient at baseline was £17.62 in the surgery group and £11.90 in the PHT group, generating a between-group difference of £5.72 (95% CI –£20.89 to £36.78; p = 0.512) in the 3-month period prior to randomisation. Over the 12 months of follow-up, receipts increased from baseline values to £57.93 in the surgery group and £12.94 in the PHT group, generating an unadjusted mean difference of £44.99 (95% CI £10.88 to £97.42; p < 0.001).

Base-case analysis results

Table 34 presents estimates of the cost-effectiveness of hip arthroscopy compared with PHT for FAI. In the base-case analysis, surgery was associated with an adjusted mean additional cost of £2483 (95% CI £1533 to £3432) and an adjusted mean additional QALY of –0.018 (95% CI –0.051 to 0.015) per patient compared with PHT over the 12 months of follow-up. On average, surgery was more costly and marginally less effective than PHT in the adjusted analysis during the first year of follow-up. The mean base-case ICER was –£140,361 per QALY gained for surgery compared with PHT. Figure 14 shows the uncertainty around this central estimate of the ICER. Figure 14a displays 1000 simulated replicates of the ICER on a cost-effectiveness plane and Figure 14b displays the probability that surgery is cost-effective compared with PHT for a range of cost-effectiveness thresholds. Almost all simulated replicates of the ICER fell on the left-hand side of the £30,000 and £50,000 per QALY cost-effectiveness threshold lines, with the central estimate (indicated by the black diamond) falling in the north-west quadrant. This suggests that surgery is unlikely to be cost-effective at the £20,000–£30,000 per QALY threshold range (see Figure 14b) that NICE currently uses to determine the cost-effectiveness of health technologies. 67 Figure 14b shows that the probability that surgery is cost-effective compared with PHT is close to zero for threshold values < £100, 0000 per QALY.

FIGURE 14.

Base-case analysis comparing the cost-effectiveness of arthroscopic surgery with PHT for FAI. The analysis accounted for missing data using multiple imputation and adjusting for age, sex and baseline health-related quality of life (effectiveness regression). (a) Incremental costs; and (b) probability that surgery is cost-effectiveness. WTP, willingness to pay.

Mean incremental net (monetary) benefit was also generated from the 1-year data for different cost-effectiveness thresholds (Table 35). The mean incremental net benefit was –£2713 if the cost-effectiveness threshold was £13,000 per QALY gained, –£3013 if the cost-effectiveness threshold was £30,000 per QALY gained and –£3367 if the cost-effectiveness threshold was £50,000 per QALY gained. These values (and the associated upper and lower 95% CI confidence limits) are all negative and suggest a net loss to the health and social care service based on the 1-year data if surgery is adopted (or a net gain by the same amount, on average, if PHT is adopted).

Sensitivity analyses results

Only the unadjusted analysis generated a difference in mean QALYs of 0.001 in favour of the surgery. The probability that surgery is cost-effective was 0.005 at £30,000 per QALY and no more than 0.05 at £50,000 per QALY (see Table 34). All other sensitivity analyses adjusted for baseline characteristics, such as age, sex, impingement type, study site, health-care service use prior to randomisation and health-related quality of life. In the adjusted sensitivity analyses, surgery was significantly more expensive (adjusted mean difference in costs ranged from £2047 to £5628) and generated fewer QALYs (adjusted mean difference in QALYs ranged from –0.018 to –0.003), on average, than PHT over 12 months of follow-up (see Table 34). Cost-effectiveness acceptability curves and plots of simulated ICERs, displayed in Appendix 3, Figures 15–21, for the different scenarios evaluated through the sensitivity analyses, show that surgery is unlikely to be cost-effective even at willingness-to-pay threshold values as high as £100,000 per QALY.

Subgroup analyses results

The subgroup analyses results are also summarised in Tables 34 and 35 and displayed graphically in Appendix 3, Figures 22–28. There was substantial uncertainty around the central estimates of incremental costs and incremental QALYs because of the reduced sample size in each subgroup, but the direction of relative cost-effectiveness of the interventions remained mostly the same as in the base-case analysis. Across all subgroups of patients, surgery mostly generated fewer QALYs (adjusted mean difference in QALYs ranged from –0.019 to –0.006) and was significantly more expensive (adjusted mean difference in costs ranged from £2058 to £3260), on average, than PHT.

Long-term modelling

The study protocol had allowed for trial participants to be followed up beyond the initial 12-month period for up to 3 years and outcome data collected at the end of the second and third year post randomisation. Given that the 12-month within-trial economic analysis did not show evidence of cost-effectiveness in favour of the surgery, it is doubtful if long-term economic modelling would be meaningful without this additional data. In addition, 11 patients representing 7.3% of the PHT group had crossed over and received surgery during the 12-month follow-up period. The net effect of patients crossing over to surgery may increase costs in the PHT arm and decrease the incremental costs between surgery and PHT. If more and more PHT patients continue to cross over to surgery in subsequent years, then they will be picked up at the second- and third-year assessments. Therefore, an assessment of the utility of a long-term economic model should be delayed until the second- and third-year data become available. This would provide a more accurate assessment of outcomes over a longer follow-up period and determine whether or not modelling is needed to capture the long-term (i.e. lifetime) costs and consequences of treatment.

Follow-up case report form: 6 and 12 months

(PDF download)

Follow-up case report form: 2 and 3 years – further procedures

(PDF download)

Surgical case report forms

(PDF download)