Facet joint injections for people with persistent non-specific low back pain (Facet Injection Study): a feasibility study for a randomised controlled trial

A global problem

Low back pain (LBP) is ranked as highest in the Global Burden of Disease in terms of years lived with disability. 1 The most recent UK ‘cost-of-illness’ study was in 1998. 2 At that time, the direct health-care costs of LBP were £1067M for the UK NHS and £565M for private health care, resulting in a cost of £28 per head of population. A 1998 US study3 estimated direct health-care costs as US$90,601M and a cost of US$335 per person. Much has changed since then. In the USA, there has been a 2.4-fold increase in spinal fusions and a massive increase in facet joint interventions, both of which are still increasing. 4,5 In response to this problem, the National Institute for Health and Care Excellence (NICE) commissioned guidelines for the management of non-specific LBP lasting between 6 weeks and 1 year. 6 The excluded treatment approaches included the injection of therapeutic substances into the back. Deep divisions in the scientific and clinical communities have become clear since the publication of the 2009 NICE guidelines and the 2009 American Pain Society guidelines for LBP,7 which also indicated that there was insufficient evidence to support the injection of therapeutic substances into the back. 8 Although the methodological approach used by the American Pain Society has been challenged, the American Society of Interventional Pain Physicians also concluded that the evidence for therapeutic intra-articular facet joint injections (FJIs) was limited. 9,10 A new NICE guideline for LBP and sciatica,11 published in 2016, did not support the use of intra-articular FJIs (www.nice.org.uk/guidance/conditions-and-diseases/musculoskeletal-conditions/low-back-pain). It does, however, in contrast to the 2009 guidance, support the use of radiofrequency denervation in people who have had a positive diagnostic medial nerve branch block.

There is a variety of different interpretations of the available evidence from randomised controlled trials (RCTs) and observational studies on the effectiveness of intra-articular FJIs. 12–18 Controversy surrounding this issue remains in the USA, and in the UK there has been new research evaluating lumbar facet joint interventions. 9,19,20 However, the outcome of all this work was that, at the outset of this study, the data were not a robust evidence base on which to inform decisions about the use of therapeutic intra-articular FJIs. Notwithstanding current NICE guidance and insufficient evidence to justify the use of FJIs, Hospital Episode Statistics21 record that, in 2014/15, 81,963 FJI procedures were performed in England for the NHS, an increase from 62,671 in 2012/13. Although there may be some inaccuracies in the coding of different facet joint-related procedures, this still represents a 30% increase in activity over 2 years.

Thus, there was a clear need for a trial to test the effectiveness of adding intra-articular FJIs to usual care for the treatment of persistent LBP when usual care as recommended by NICE or the American Pain Society has been ineffective. It is important that the proposed trial provides data that all parties can agree on. If the trial has positive results, then reinvestment in this treatment will be justified. If the trial has negative results, its conclusions need to be sufficiently robust that all parties to the debate on current guidance are satisfied that the evidence does not support the use of therapeutic intra-articular FJIs.

The UK National Institute for Health Research, via a specific call, funded two feasibility studies in preparation for trials of therapeutic intra-articular FJIs. We were funded to carry out one of these studies. Our proposed main study will test the addition of a therapeutic intra-articular FJI to best usual care (BUC) [Health Technology Assessment (HTA) programme reference number 11/31/01]. 22 A different team is funded to test the feasibility of a more explanatory trial comparing active intra-articular injection with a sham control in people with a positive diagnostic medial branch nerve block (HTA 11/31/02). 23 These two studies will produce complementary data that will inform decisions on the merit of offering therapeutic intra-articular FJIs to selected people with LBP.

Diagnosis of low back pain

Low back pain is pain or discomfort felt in the back between the bottom of the rib cage and the buttock creases. Once medically serious causes of LBP have been excluded (infection, fracture, malignancy, inflammatory disorders such as ankylosing spondylitis), LBP is labelled as being non-specific. None of the early treatment options recommended in the 2009 NICE guidance requires any further diagnostic work-up.

Unless there is a concern that there is a serious medical cause for the pain, imaging of the lumbar spine is not indicated. This approach recognises the difficulty of reliably identifying subgroups of people who may respond better to any particular treatment approach. 24,25 Notwithstanding this overall approach to the diagnosis of LBP, it is possible that there are subgroups of individuals who would benefit from one or more specific approaches to the treatment of their LBP. One subgroup who may benefit from a more specific back treatment is patients whose pain wholly, or partially, arises from the facet joints. This notion is supported by the 2016 NICE back pain guideline, which recommends the use of diagnostic medial branch nerve blocks before considering radiofrequency denervation in people with suspected facet joint pain. The clinical assessment of suspected facet joint pain suggested in NICE 2016 guidance is based on work described later in this report. 26

Facet joint pain

The facet joints are paired structures between the superior and inferior articular processes of adjacent vertebrae that, in the lumbar spine, allow flexion and a degree of rotation of the spine. They are synovial joints whose capsule is richly innervated. With increasing age, progressively more people develop radiological osteoarthritis of the facet joints. A systematic review estimated the prevalence of facet joint degeneration increased from 4% for those in their twenties through 32% for those in their fifties to 83% for those in their eighties. 27 There is not, however, an association observed between radiological change in facet joints and the presence of back pain. 28 If pain arising from the facet joint is suspected clinically, it may be abolished temporarily by the injection of local anaesthetic, which may be used as a diagnostic test. A depot preparation of steroid added to the injection may prolong the analgesic benefit. The relief or reduction of pain may facilitate compliance with a programme of exercise designed to improve lumbar range of movement and muscular stability.

Conclusion of initial scoping literature review

In our own 2016 review,60 we identified six relevant randomised placebo/sham-controlled trials of intra-articular FJIs. 12,52,55,58,61,62 Two studies (Lilius et al. ,55 n = 109; Carette et al. ,52 n = 97) used placebo injections into the facet joints as the control treatment, one study (Ribeiro et al. ,62 n = 60) used intramuscular corticosteroid as the control treatment, two studies (Mayer et al. ,58 n = 70; Kawu et al. ,61 n = 18) used exercise as the control treatment and one (Celik et al. ,12 n = 80) used bed rest plus analgesia and non-steroidal anti-inflammatory drugs as the control treatment. Four studies made a clinical diagnosis of facet joint pain,12,55,61 one study62 used clinical and radiographic features to make the diagnosis and two studies52,58 used diagnostic blocks. We considered the study populations, and the comparisons made, to be too heterogeneous for any robust conclusions to be drawn. 60

The quality of reporting of these trials of FJIs is generally poor and it is not a robust evidence base to inform decisions about the use of FJIs.

Where might facet joint injections fit in the care pathway?

For people with acute or subacute LBP or early persistent LBP, the prognosis is generally very good and there is little need for invasive treatments. Low-intensity, low-risk and therapist-delivered interventions, as recommended by NICE, are sufficient. Although there are some differences between the 2009 NICE guidance6 and the 2016 guidance,11 both support the use of therapist-delivered interventions as the first treatment approach after advice and analgesics. Those with substantial problems persisting after such interventions (typically ≥ 6 months from onset) require more intensive treatment. Persistent LBP is a complex biopsychosocial phenomenon. In those people with pain persisting beyond 6 months, in spite of good conservative care, a syndrome of chronic pain and disability will already be present. For this reason, one would not expect FJIs, on their own, to resolve the problem. Rather, pain relief obtained from a FJI may give the person with back pain the confidence and a window of opportunity to engage more fully with a rehabilitation programme.

There is a clear need for a trial to test the effectiveness of adding FJIs to usual care as recommended by NICE for the treatment of persistent LBP. It is important for this trial that it provides conclusive results. If the trial has positive results then this will be a justification for investment in this area. On the other hand, if the trial is negative then its conclusions need to be sufficiently robust that all parties to the debate on the role of therapeutic intra-articular FJIs are satisfied that the evidence does not support their use. There are methodological challenges to setting up and running such a trial; principally, these centre on the identification of people with LBP that is, at least in part, from facet joints. Our feasibility study addressed these methodological issues, tested trial processes and recruitment in an external pilot, conducted an interim analysis and identified sites to conduct the main study.

For the NHS to consider reinstating FJIs for people with otherwise non-specific LBP, a package of care including FJIs, for selected patients, needs to be an effective and cost-effective addition to BUC. This is true regardless of the result of any placebo-controlled trial of the efficacy of FJIs. The components of any overall effect (positive or negative) of FJIs will be the non-specific effects of attending for the injection (including any advice from the treating clinician), any local effects from injecting fluid into the facet joint and the specific effects of the drug/s injected. A positive efficacy study in a tightly controlled population will not necessarily transfer to a treatment that is effective in real life. Conversely, failure to show a positive effect in an efficacy study using a placebo or sham injection would not necessarily exclude the possibility that the overall package of care is effective. Furthermore, such an efficacy study will not be able to answer a question on the cost-effectiveness of adding the intervention to usual care.

The situation here is perhaps analogous to interpreting the evidence on the use of acupuncture, a treatment with a much weaker theoretical base than FJIs. There is a substantial body of evidence that acupuncture is superior to usual care for a range of common disorders, with meaningful effect sizes. The evidence that verum acupuncture is superior to a sham control is, however, much weaker, with very small apparent effect sizes. Nevertheless, acupuncture was recommended by NICE for LBP in 2009,6 although not in 2016 guidance,11 and for headaches in 2012. 71 For these reasons we propose a two-arm study testing the effect of adding FJIs to a BUC package.

Patient and public involvement

We are not able to report on patient and public involvement (PPI) in the framing of the original commissioning brief from the HTA programme. Throughout the development of this proposal, in response to the HTA brief, and subsequent conduct of the study there has been input from PPI representatives and groups. All of the project’s development work and interpretation of the results have had input from PPI representatives.

• Two very active co-applicant PPI representatives on the studies trial management group played an important part in the development of study material and intervention development at the consensus conference, and one provided a reflection of their experience working on the trial (see Appendix 1).

• One PPI representative on the trial steering committee who gave clear guidance throughout.

• Six lay representatives participated in our consensus meeting. 74,76

We also acknowledge the help and support provided by University/User Teaching and Research Action Partnership (UNTRAP) at the University of Warwick, which provides training and support to ensure effective PPI.

Before the conference

The three stages of the study are presented in Figure 1: (1) scoping review and identification of key design considerations, (2) evidence reviews and (3) consensus conference.

FIGURE 1.

Facet Injection Study protocol development process. Adapted with permission from Mars et al. 26

The consensus conference

Potential conference participants were invited through relevant professional and lay organisations (Box 1). We sought participation from experts from across the UK. By expert, we mean that participants were professionals or lay people with an interest in, or experience of, back pain, its treatment and, in particular, its treatment with therapeutic intra-articular FJIs. The invitation was to a 1-day conference with no attendance charge and travel expenses were reimbursed. This was held at the University of Warwick on 27 June 2014.

Approximately 1 week before the consensus conference, a document consisting of the design considerations and related evidence was sent to all those registered to attend.

We used nominal group technique to gain consensus. This allows for discussion, while avoiding individuals or groups dominating the consensus process, and allows participants to draw on available evidence and expertise. 79 We started the conference with a brief reminder of the key design considerations and evidence. We then held 15 small group consensus sessions, each lasting 1 hour, with five groups meeting in parallel at any one time (Figure 2). Small group results were fed back to a plenary session, in which final consensus was reached. With participant consent, all sessions were audio-recorded for reference during analysis. Participants were randomly assigned to small group sessions stratified by profession (approximately 10–12 per group) with each participant discussing three different design considerations. Each small group had a trained facilitator, a scribe and a subject expert from our team. The subject expert did not participate in the discussions but answered questions about technical issues when invited to by the facilitator. Discussions centred on the particular design consideration, with the suggested ‘protocol’ as a starting point when appropriate. Nominal group technique was adapted to the design consideration under discussion as described in Figure 2. Each participant confidentially ranked the acceptable approaches identified by the group. Results were collated by the scribes.

FIGURE 2.

A diagrammatic representation of the consensus process. Adapted with permission from Mars et al. 26 NGT, nominal group technique.

In the sections following, we outline the five design considerations and how the discussions on these were conducted during the consensus conference day. More detailed information is provided in three documents:

1. the evidence reviews76

2. the consensus final report74

3. a publication based on the consensus day. 26

Post conference

All results were checked and verified from all small group sessions and the plenary session. A small number of errors were found in the collation of rankings. The team therefore contacted participants with relevant expertise via e-mail to clarify and reach a consensus on these items.

Evidence reviews

A full evidence document was produced for each delegate, which was distributed electronically before the day and provided in hard copy on the day. 76 This included tabulated results of the searches, brief summaries and, in several cases, suggested ‘protocols’.

Consensus conference

We present a brief summary of the results from the consensus conference for each of the five design considerations. Full data are available. 74 Results related to the first three of these design considerations (diagnosis, injection technique and BUC) have been published elsewhere and data and a number of tables have been reproduced here with the full permission of the editors.

Post conference

Clarifications were needed for three of the design considerations (diagnosis, the process of FJI and the BUC package).

Meta-analysis of data from large trials of physiotherapist-delivered interventions for chronic low back pain

To gain an indication of the likely magnitude of MID unit differences that may be expected in a positive study of an intervention to treat LBP, we conducted a meta-analysis using published data from large trials with which we were already familiar. We selected studies that we had previously included in a database of individual patient data from RCTs of physiotherapist-delivered interventions for back pain. 25 Studies were included if they included data on ≥ 300 participants and had used the RMDQ. 25 Our original intention was to also include a meta-analysis of Pain-NRSs; however, none of the studies in our sampling frame included both a RMDQ and useable Pain-NRS data. The purpose of this analysis was to obtain illustrative data for the consensus meeting rather than to systematically report all studies meeting these criteria. For this reason, we made the pragmatic decision to include only those studies that we had assessed as part of this previous project.

Full-text versions of the included studies were retrieved. Two reviewers extracted study-level information from the included articles: a standardised data extraction list included study-specific information (authors and trial population) and outcome-specific information [primary and secondary outcomes, mean (SD) between-group differences in scores at baseline and follow-up and MIC if calculated].

The mean between-group differences in change scores for the RMDQ and across treatment groups and at different time points were reported for each trial. Each value was compared with the known MIC for the RMDQ. 87

For all included studies, MID units were calculated using published MIC guidance. MID units were calculated per trial and as an overall value (all trials combined) at 3 and 12 months. As two MIC values are recommended for the RMDQ, two forms of MID unit were calculated, reflecting the MIC score change and MIC 30% improvement from baseline. 87

Calculation of the between-group differences of change scores and the minimally important difference unit from a large UK trial (UK Back pain Exercise And Manipulation trial)

We carried out a further analysis of data from a large UK trial of therapist-delivered interventions (n = 1169) (the UK BEAM trial86). Using individual patient data, we were able to obtain Pain-NRS data for pain today as a single item extracted from the Modified Von Korff pain grade scale. 86,99 For this analysis, all three active treatments in UK BEAM trial were pooled for comparison with the control intervention. Scores for the RMDQ and Pain-NRS were adjusted for sex, age and scores at baseline. Mean between-group differences of change in RMDQ and Pain-NRS scores and MID units (30% and score change) at 4 weeks, 3 months and 12 months were calculated.

Consensus meeting: score interpretation

Finally, a consensus meeting was held. 76 The results from the two previous stages were used to inform a 1-day consensus meeting of clinical and academic experts and lay representatives who sought to determine and make recommendations for between-group score interpretation. All participants received an evidence synthesis in advance of the meeting. A nominal group technique was adopted to gain consensus. Following a brief reminder of key evaluation considerations and evidence, delegates were randomly assigned to small group sessions that were stratified by profession (approximately 10–12 per group). Discussions lasted up to 1 hour. Each group had a trained facilitator, a scribe and a subject expert. Participants were invited to consider the following overall question: what is the difference in magnitude of response between treatment and control groups that should be considered large enough to establish the scientific or therapeutic importance of the results?

Specific subquestions, pertaining to the RMDQ and/or Pain-NRS, included:

• At 3 months should we be seeking a mean between-group difference in change scores that is smaller than/the same/or larger than that observed for the trials of physiotherapy?

• Informed by the MID units calculated for the trials of physiotherapy, at 3 months should we be seeking a small (< 0.5), medium (0.5–1.0) or large (> 1.0) MID unit as proof of important difference?

The results from all small group sessions were collated and presented during the final plenary session, in which there was considerable discussion on this topic. When there were inconsistencies between small groups, these were discussed and further ranking undertaken. Final consensus was sought for all questions.

Meta-analysis of data from large trials of physiotherapist-delivered interventions for chronic low back pain

Following application of our inclusion criteria, three out of five shortlisted large trials were included in the analysis (Table 5).

Consensus meeting: score interpretation

These data were presented at the consensus meeting. The outcome of the meeting was that any trial of FJIs should show a larger effect size than that typically seen in trials of physical interventions and that a medium MID unit difference of 0.5–1.0 points should be sought. The largest effect size seen on the RMDQ at 3 months in the example trials was 1.87 points (UK BEAM trial combined treatment). This is a MID unit difference of 0.37 using an absolute difference of 5 points, or a MID unit difference of 0.69 based on a 30% reduction from baseline. Delegates were also keen that results should be presented as responder criteria and that they would expect any trial of FJIs to have a greater number of responders compared with usual care than a trial of physiotherapy.

Primary objective

The primary objective of this trial was to explore the feasibility of running a RCT to test the hypothesis that for people with suspected facet joint pain contributing to persistent LBP, adding the option of FJIs to best usual non-invasive care available from the NHS is clinically effective and cost-effective.

Secondary objectives

• To evaluate a standardised control treatment deliverable in the NHS and congruent with NICE guidance (BUC).

• To develop and test systems for collecting short-term and long-term pain outcomes, including measures required for economic evaluation.

• To demonstrate that recruitment to the main trial is feasible.

• To collect the recruitment and outcome data required to inform sample size and number of sites needed for the main trial.

• To conduct a between-group analysis to inform the decision on the need for a full trial.

• To undertake a process evaluation of patient experience within the trial data from exploratory work, data from the pilot trial and data from both exploratory work and the pilot trial.

Study design

The FIS was a mixed-methods randomised multicentre feasibility trial to compare FJIs with BUC.

Setting

The feasibility RCT was conducted in five NHS acute trusts in the centres based in England. These were:

1. University Hospitals Coventry and Warwickshire NHS Trust

2. Kidderminster Hospital and Treatment Centre, Worcestershire Acute Hospital NHS Trust

3. King’s Mill Hospital, Sherwood Forest Hospitals NHS Foundation Trust

4. The James Cook University Hospital, South Tees Hospitals NHS Foundation Trust

5. Warwick Hospital, South Warwickshire NHS Foundation Trust.

In each hospital we were working with the clinicians providing pain services for patients. One further trust, Hull and East Yorkshire Hospitals NHS Trust, was preparing to join the study at the time recruitment stopped.

Participants

Patients referred to the trust for treatment of LBP present for at least 6 months, after failure of conservative treatment, were considered as potential participants. We aimed to recruit 150 patients. We expected up to 40 participants to be recruited at each participating centre, with up to 20 participants randomised to each group at each site. Recruitment was planned to be primarily from pain clinic services, which was an approach that was congruent with the commissioning brief.

Inclusion criteria

1. Was able and willing to comply with the trial procedures and signed and dated informed consent was obtained.

2. Was aged ≥ 18 years with at least moderately troublesome LBP present for at least 6 months. 103

3. Had LBP as their predominant musculoskeletal pain.

4. Had undergone registered health professional delivered treatment for LBP in the 2 years prior to study entry.

5. Met clinical criteria for possible facet joint pain when there is no radicular symptoms (defined as pain radiating below the knee) and no sacroiliac joint pain elicited using a pain provocation test and increased pain unilaterally, bilaterally on lumbar paraspinal palpation and increased LBP on one or more of the following: extension (more than flexion), rotation, extension/side flexion or extension/rotation.

6. Was able to manage text messaging, or an alternative means of daily data collection (paper-based diary).

7. Was fluent in written and spoken English.

Exclusion criteria

1. Was unable to attend for randomised treatment, other circumstances that would significantly decrease the chance of obtaining reliable data, achieving trial objectives or completing the trial and follow-up assessments or was considered unsuitable to participate in the trial by an investigator.

2. Was unable/unwilling to undergo injections.

3. Had used oral corticosteroids or had a corticosteroid injection in the preceding 3 months.

4. Had an underlying serious psychiatric or psychological disorder that precludes participation in either intervention.

5. Had previously undergone spinal injections.

6. Had previously undergone spinal surgery.

7. Had a contraindication to FJIs, for example a serious comorbidity (e.g. severe chronic obstructive pulmonary disease, poorly controlled diabetes), such as malignancy, infection, inflammatory disorder or fracture, or was taking anticoagulant medications.

8. Had a known allergy to the constituents of the planned injections.

9. Pregnancy or suspected pregnancy.

10. Was previously randomised in this trial.

11. Was currently participating in another clinical trial (with an unregistered medicinal product) or < 90 days had passed since completing participation in such a trial.

Facet joint injection with ‘best usual care’

Participants received detailed information (verbal and written) about the procedure prior to and during the diagnostic physiotherapy assessment. When the participant attended for injection, the clinician (consultant) responsible for injection made a clinical assessment to ensure FJI remained appropriate. Briefly, the clinical assessment, as per normal clinical practice, included a review of the medical history and an assessment of the participant’s suitability for FJIs. There was a particular emphasis on ensuring that the participant had no important comorbidities such as malignancy, inflammatory disease, severe chronic obstructive pulmonary disease, poorly controlled diabetes, fracture or infection (trial exclusions). The clinician could also ‘postpone’ a treatment session if the patient presented with a short-term ailment (e.g. influenza). Following normal local trust clinical practice, the clinician obtained informed consent for the injection procedure from participants prior to injecting the facet joints. If during the assessment the clinician identified significant pathologies that had not been identified previously that excluded a potential participant, they ‘flagged’ these in the patient notes or to a local investigator site clinician or a GP, whichever was the most appropriate. Participants were fully informed that they may not receive the injection at this consultation. If it was determined that the injection was not appropriate, the participant would continue to receive the ‘BUC’ physiotherapy package.

Up to six facet joints (L3/L4, L4/L5 and L5/S1) bilaterally in each participant were injected. However, when, on clinical assessment, there was unilateral pain or involvement of only some levels, the operator may have chosen to carry out a unilateral injection or to be selective on levels injected. This pragmatic approach reflects current clinical practice within the NHS and was informed by the earlier consensus meeting.

This approach is consistent with that used in trials of other complex interventions for LBP, for example a manual therapy or a cognitive–behavioural approach, in which practitioners choose from a limited range of options based on their clinical assessment of the patient. 13,14

Control treatment: ‘best usual-care’ package (standardised treatment deliverable in the NHS and congruent with 2009 National Institute for Health and Care Excellence guidance)

A study ‘physiotherapist and cognitive–behavioural treatment manual’ was prepared to ensure that all sites adhere to the same procedures when delivering the ‘BUC’ package. Prior to commencement of the trial, physiotherapists within the investigator site team were trained in the delivery of the package and its processes. Below we outline the principles and theoretical underpinnings of our ‘BUC’ package. This builds on previous work showing that physiotherapists can deliver an effective cognitive intervention but has an additional emphasis on a physical approach to treating facet joint dysfunction. 104 Informed by the Medical Research Council framework for complex interventions, we developed a best practice intervention that is congruent with NICE. 105 Consensus agreement ensured that the ‘BUC’ package was deliverable within the NHS at a reasonable cost, that is, for much less than 100 hours of contact time recommended by NICE at the time this study took place. 106

Fundamentally, treatment followed guidance from NICE but tailored to individual patients. Participants underwent a thorough initial physical assessment (60 minutes) based on the principles of Maitland manual therapy. 107 The assessment included discussion of patient expectations, fear avoidance and perceived self-efficacy.

Participants randomised to the injection arm had their FJIs between their first and second physiotherapy treatment sessions and within no more than 3 weeks. They then had four further sessions, of 30 minutes each, which incorporated manual therapy, active and motor control exercises, soft tissue stretches and a cognitive–behavioural approach. This included cognitive restructuring, challenging unhelpful thoughts associated with physical activity and chronic LBP, pacing and goal-setting. The sessions were delivered in a participant-centred manner, with each session tailored to the participant’s needs.

The ‘BUC’ intervention was piloted with five participants recruited from participating trusts before the trial went into the randomised phase.

Procedures for screening

A member of the local principal investigator (PI)’s site trial team actively identified referrals to secondary care for patients with LBP. We expected that most of these patients, having been referred after failure of conservative treatment, would already have a diagnosis of non-specific LBP. The referring GP would have referred the patient based on their knowledge and the patient’s history of chronic back pain and previous treatments. Those patients in whom the GP suspected a specific cause for their back pain would typically have been referred in through a different pathway, such as the 2-week wait pathway for suspected cancer. Any patients for whom the referral from the GP queried a specific cause of back pain were not approached for the study. As the entry criteria for this study required participants to have chronic pain, to have been assessed by their GP and to have already had a course of conservative treatment, it is extremely unlikely that any patients with malignancy, infection or inflammatory disorders were approached. We were, however, aware that the 2009 NICE guidance6 for the management of non-specific LBP recommends keeping the diagnosis under review. As originally planned, we recruited from secondary care pain services. It was here that we perceived the question on the effectiveness of intra-articular FJIs to be of greatest importance. Recruitment in this environment was challenging. At the time the study closed because of poor recruitment, we were working towards being able to recruit from neurosurgery, rheumatology and orthopaedic clinics. We were also setting up a GP and community physiotherapy referral system into a secondary care physiotherapist-led research clinic. Posters were placed in GP and community clinics. Interested patients could contact local research nurses in their participating trusts.

A member of the investigator’s trial team based at the local site (e.g. a research nurse or physiotherapist) sent a brief screening information sheet, a screening questionnaire and an expression of interest form to potential participants, for whom there were no concerns about a specific cause of LBP, to assess preliminarily their eligibility for enrolment into the trial. At this time, the brief screening information sheet was sent to potential participants to provide them with sufficient information to make an informed judgement about whether or not they might be interested in finding out more about the study. At some sites, patients with LBP who were attending clinics were approached by their treating clinician (e.g. the local PI or team) to assess their possible participation in the trial. In these cases, the treating clinician gave the patient a trial information pack containing a brief screening information sheet, a screening questionnaire and an expression of interest form. The potential participant was invited to return the completed screening questionnaire and expression of interest form to the PI’s trial team at the local site using an envelope provided. From review of the questionnaire, if the potential participant appeared eligible and interested, a member of the local site trial team sent out the full participant information sheet regarding the trial. Specifically, the full participant information sheet informed the patient about the eligibility assessment they were being invited to attend, which helped to determine their possible inclusion in the trial, and clearly explained that it was only at this assessment that their eligibility would be confirmed. They were invited to attend an appointment for their clinical/physical (diagnostic) assessment at a nominated research clinic. Each site was required to maintain an anonymised trial screening log (see Appendix 2) monitoring the number of screening packs sent out, returned and recruited, noting reasons for ineligibility (Figure 3).

FIGURE 3.

Participant pathway and procedures. EQ-5D, EuroQol-5 Dimensions.

Clinical diagnosis of suspected facet joint pain (visit A)

A study-trained physiotherapist met with potential participants who had agreed to attend the assessment appointment. On arrival the physiotherapist carried out the FIS diagnostic assessment.

Based on the agreement from the consensus conference, suspected facet joint pain was considered to be present when there was:

1. no radicular symptoms (defined as pain radiating below the knee)

2. no sacroiliac joint pain elicited using a pain provocation test

3. increased pain, unilaterally or bilaterally, on lumbar paraspinal palpation

4. increased LBP on one or more of the following:

   • extension (more than flexion)

   • rotation

   • extension/side flexion

   • extension/rotation.

Note that both extension/side flexion and extension/rotation were representative of regular compression patterns. 49

A manual was developed for the diagnostic assessment to aid the physiotherapist. 72

After it had been confirmed that the patient met all of the inclusion criteria and none of the exclusion criteria, and was deemed ‘eligible’ following the diagnostic assessment, their written informed consent to participate in the trial was obtained by suitably trained physiotherapists or research nurses, or by the PI as a delegated responsibility.

If potential participants were considered not to have facet joint pain, the physiotherapist informed them that they would not be included in the trial and that their standard treatment/management care plan would continue as referred.

The outcome of the eligibility assessment for suspected facet joint pain was documented for all potential participants. Those potential participants who were considered not eligible based on eligibility assessment were asked to consent to the use of the data generated from their assessment. This was documented in each person’s clinical record.

After giving written informed consent, the participant was scheduled to attend a 1-hour preliminary ‘BUC’ physiotherapy session delivered by a specially trained physiotherapist.

First ‘best usual-care’ session (visit B1)

Participants initially underwent a thorough physical assessment based on the principles of Maitland manual therapy assessment and clinical reasoning,107 in which symptomatic levels are identified, and the severity and nature of the symptoms recorded and used to direct treatment.

Assessment included discussions of acceptance, goal-setting and pacing as well as general discussion of patient expectations, fear avoidance and self-efficacy to assess any perceived challenges and barriers that patients felt may be preventing them from engaging in self-management of chronic pain and to allow subsequent treatment sessions to be tailored to individual need. A manual was developed for the ‘BUC’ intervention to aid the physiotherapist. 73

Once the first trial physiotherapy session was completed, the participant was randomised to receive FJI with BUC package or BUC package only. The treating physiotherapist, within 1 working day, informed the investigator trial team that the first treatment session has been completed. The investigator trial team then completed the randomisation process and informed the participant whether or not they needed to attend for an injection.

This approached ensured that all participants had received the introductory physiotherapy session and that we were testing the addition of intra-articular FJIs to BUC physiotherapy package rather than comparing FJIs with physiotherapy. Our concern here was that those allocated to injections might have elected not to attend for physiotherapy assessment if they expected a substantial benefit from injections.

At this visit we also collected an immediate pre-treatment pain score that would be important for any analysis of short-term benefits, or harms, from treatment. This is pertinent because there would inevitably be a delay between consent and arranging treatment sessions.

If a patient was randomised to receive FJI, the injection was scheduled between the preliminary and secondary physiotherapy treatment session. This took place within 3 weeks of randomisation.

Facet joint injections (visit C)

For those participants who were randomised to undergo FJI, the injection took place between the first and second BUC sessions. The treating clinician at this time made their own assessment of the participant’s suitability for FJI and obtained consent for the procedure following normal practice in each participating trust. The treating clinician could postpone injection in the presence of short-term illness (e.g. influenza). If significant comorbidity was identified at this time that contraindicated the injection, this was ‘flagged’ in the patient notes or to the local investigator, site clinician or GP, whoever was most appropriate. The participant would continue to receive the control intervention. A manual was developed for the clinician, which gave clear instructions on how the injection(s) should be carried out. 75

‘Best usual-care package’ (visits B2–B6)

It was expected that all 150 participants would have up to a total of six ‘BUC’ physiotherapy sessions (including the first treatment session immediately prior to randomisation). The five remaining sessions were to last approximately 30 minutes. All participants (those in the intervention and those in the control arms) were encouraged to attend all of these sessions. The package was a series of one-to-one sessions with a study physiotherapist who would use the BUC manual informed by consensus to help them tailor the treatment to the participant’s needs. 26 Sessions were a bespoke package of physical and behavioural rehabilitation. All treatment sessions were to be completed within 12 weeks of randomisation.

Follow-up

Follow-up was conducted at 3 months after randomisation using self-administered questionnaires. The majority of questionnaires were completed in postal format. Response was tracked carefully by the trial office and a reminder questionnaire was sent after 2 weeks if a follow-up questionnaire had not been returned. If the questionnaire had still not been returned after a further 2 weeks, participants were telephoned to check that they were receiving the questionnaires and to arrange for another to be sent if needed. A core data set was requested via the telephone, at a time convenient to the participant, if a participant had still not responded.

The case report form (CRF) was stamped with the date and initialled on receipt at the WCTU office. It was checked for correctness and completeness and coded for data entry. Any queries were checked with the statistician. Missing data were clarified with participants when possible.

The baseline and follow-up data were entered into the study database and a random sample check of 10% was undertaken to ensure accuracy.

Demography and baseline assessment

Appendix 3 includes an example of the CRF listing the demographic and clinical data that were collected at pre-enrolment and pre-randomisation stages. This includes the baseline assessments that are carried out during visit ‘A’ at the time of enrolment.

Clinical outcomes

For the randomised pilot trial we used a package of outcome measures consistent with consensus recommendations for outcome assessment in back pain trials. These included pain measurement, physical function, emotional function, back-related function, generic well-being, disability (social role), satisfaction with care, patient rating of improvement and satisfaction with treatment, symptoms (pain), adverse events, participant disposition and a modified form of the Patient Generated Index (PGI). 67,108 In this feasibility trial we assessed performance of these measures and with a view to reducing the questionnaire burden in the main trial.

Table 8 summarises the outcome measures and their time of completion by participants. The main questionnaire packages are completed at baseline (at trial entry assessment) and follow-up (3 and 6 months post randomisation). A pain severity (today) score was recorded daily for 35 days from 7 days before first treatment session and following this weekly until 3-month follow-up. Health utility (EQ-5D-5L) was recorded weekly from 1 week prior to first treatment session until the night before injection appointment when we asked participants to record it daily for 8 days and then weekly until the 3-month follow-up. In addition, intervention participants record a pain severity 45–60 minutes before and after receiving an injection. Clinical data were collected and recorded by physiotherapists and clinicians, covering patients’ assessments, injections and involvement in the physiotherapy intervention. Copies of questionnaires are provided in Appendices 4 and 5.

Primary outcome

The primary outcome for this feasibility trial was a numerical rating scale for pain collected via text messaging over 3 months following randomisation. 84

A method for frequent (daily) collection of pain-related outcome data using a text messaging system was developed and piloted within the feasibility trial. To test the feasibility of using an electronic diary, a review of studies using electronic diary-related pain or pain-related disability measurement was conducted, as well as the identification or development of an application or text system with candidate diary prompts and choice of prompts and frequency. 117 Research participants find electronic symptom diaries acceptable and can generate valid symptom data. 118 During the exploratory phase, initial feasibility testing, we refined the diary recording for the needs of the trial.

For those participants unable or unwilling to use a text messaging system, we used a paper-based system. We collected data on pain severity using an 11-point numerical rating scale (Pain-NRS). 84 Data on pain were collected daily for 35 days starting 1 week prior to randomisation and then weekly for until 3 months after randomisation. It was felt that any benefit or harm in the immediate post-injection period was likely to change on a daily basis, and, therefore, less frequent data collection should identify any between-group differences.

Our second primary outcome focused on back pain related disability. We used the RMDQ at baseline and 3 months (follow-up collected using a postal questionnaire).

It was planned that the between-group analysis would be looking for a ‘signal’ that the intervention package might be effective. If we failed to find an indication that there may be a worthwhile benefit, there would be no need to proceed to a full trial. We were not, here, interested in the point estimate for any between-group differences. Rather, we were interested in whether or not it was plausible that FJIs would achieve a benefit of sufficient magnitude to make them a worthwhile intervention. Thus, a decision not to proceed to a main trial would have been based on it being implausible that the intervention would achieve a worthwhile clinical effect. The size of effect below which the clinical and academic community accept that the use of FJIs was not worth considering was decided at the consensus development conference. However, because of the early closure of recruitment, no between-group analyses were undertaken (see Analyses).

Sponsor and governance arrangements

The University Hospitals Coventry and Warwickshire NHS Trust and University of Warwick acted as cosponsors for this trial.

Ethics

Originally we submitted an ethics application for approval with a full protocol and all of the relevant paperwork (e.g. patient information sheets, consent forms) on 7 April 2014. In this application we noted that we were seeking approval for the general methodology of the study but that it was possible that some things, such as the drugs used for the intervention, would change after our developmental conference. This application was rejected and we were instructed to reapply after the consensus conference. After resubmission we received approval on 11 August 2014 (Research Ethics Committee reference 14/YH/0161). This initial refusal prevented us from being able to set up sites for the trial when originally planned. There were a number of amendments, which are outlined below along with their dates of approval:

• approval for new documents (24 October 2014):

  • a cover letter to supplement the follow-up questionnaires

  • a reminder letter for non-return of follow-up questionnaires

  • a daily/weekly pain diary

  • a weekly health diary

  • a daily health diary

• approval to add a new site/PI (20 November 2014):

  • added Castle Hill Hospital, Cottingham, Mr Marek Karpinski

• approval to add a new site/PI and to remove a site/PI (as pain service discontinued at this time) (13 February 2015):

  • added The James Cook University Hospital, Middlesbrough, Dr Sam Eldabe

  • removed South Warwickshire NHS Foundation Trust, Warwick Hospital, Warwick, Dr James (Hugh) Antrobus

• approval to add a new site/PI (29 June 2015) (reinstating site after pain management service restored):

  • added South Warwickshire NHS Foundation Trust, Warwick Hospital, Warwick, Dr James (Hugh) Antrobus

• approval to make changes to the protocol and information sheets and gain approval for a new clinic poster (14 September 2015)

• approval for the new scheduled diagnostic assessment and screening questionnaire response letters (16 December 2015).

Regulatory authorities

The FIS comes under the definition of a Clinical Trial of an Investigational Medicinal Product under the European Union Clinical Trials Directive 2001/20/EC121 and, therefore, required submission to the Medicines and Healthcare products Regulatory Agency (MHRA). The trial did not start before the necessary approvals from Research Ethics Committee and the MHRA were obtained in accordance with the International Conference on Harmonisation of Technical Requirements for Registration of Pharmaceuticals for Human Use E6-Good Clinical Practice. 122 Before commencing recruitment, each PI was required to obtain local NHS permission from within their trust.

Approval was granted on 18 November 2014. A copy of all regulatory/ethics/NHS permission approval documentation was provided to WCTU prior to commencing recruitment of participants into the trial.

Adverse events

The monitoring and reporting of potential adverse events was carried out in line with WCTU standard operating procedures.

The following were expected adverse events and, if found, were recorded in the CRF:

• pain, bleeding, discomfort and minor bruising at the injection site (transient)

• numbness in the buttocks and legs from local anaesthetic (transient)

• infection of injection site; inadvertent intravenous injection (uncommon)

• musculoskeletal injuries requiring medical attention including serious sprains, joint dislocation, falls or other injuries occurring as a direct consequence of the intervention (i.e. while participating in the ‘BUC’ physiotherapy treatment intervention in real time).

Procedures in case of pregnancy

Pregnancy was an exclusion criterion for entry into the study; however, we planned to follow up and document the outcome of all pregnancies (spontaneous miscarriage, elective termination, normal birth or congenital abnormality) even if the patient had been discontinued from the trial.

All female participants were advised to notify the investigator immediately if they became pregnant during the trial. The investigator would then report any pregnancy to the WCTU. As noted, we planned to follow up any pregnancy, and any complications were recorded as an adverse event. All reports of congenital abnormalities/birth defects would be reported and followed up as serious adverse events.

Complete study procedures for adverse reactions, adverse events, serious adverse events and suspected unexpected serious adverse events were included in the study protocol.

Data collection and management

All data for an individual participant were collected by a member of the investigator’s research team and recorded in the CRF. Participant identification in the CRF was through their initials and unique trial number allocated at the time of enrolment. Data were collected from the time the patient was considered for entry into the trial through to the completion of the intervention.

All data were entered into the CRF; a copy of each form was returned to the WCTU and a duplicate was retained at the participating site. The unique trial number and other contact details of all participants were supplied to the WCTU to allow follow-up telephone or postal questionnaires to be administered to the participant at 3 months and 6 months after randomisation.

Data collection was restricted to variables required to define patient characteristics at enrolment, to monitor the treatment received, to monitor adverse effects and to determine quality of life and the use of health-care resources.

Database

The database was developed and managed by the programming team at the WCTU, and all specifications (i.e. database variables, validation checks and screens) were agreed between members of the project team, including the programmer and statistician.

Staff training

As noted in Chapter 2, Consensus conference (Diagnosis; Injection technique the process of therapeutic intra-articular facet joint injection; and Best usual care), the research team developed manuals/protocols for diagnosis, BUC and injection technique, informed by best current evidence and consensus. Training days were scheduled for the appropriate staff in these trial procedures (e.g. physiotherapists for diagnosis and BUC, and clinicians for injections). We piloted this training in the pilot study. The manuals/training were updated before the main feasibility trial based on the experiences and observations of their use in the pilot study.

Pilot procedure

We planned to recruit around 10–12 participants to the pilot study from one or more of our participating trusts. The recruitment and assessment processes would be the same as those designed for the main study. At the end of the diagnostic assessment, participants with suspected facet joint pain who gave informed consent were asked to complete the baseline questionnaire. The participants in the pilot study were not randomised.

All of those participants who met the eligibility criteria were offered the whole treatment package (including the injection and ‘BUC’ package). Once treatment was complete, a member of the research team contacted participants by telephone to get their feedback on their experience of the assessment and treatment process. This included asking for comments on the face validity and acceptability of the baseline questionnaire. The research team obtained feedback from the treating physiotherapists about the study processes. The data from participants and therapists were used to refine the assessment process and treatment package prior to the randomised feasibility trial being launched.

Statistical analysis

At the outset it was planned that the main quantitative outputs from this feasibility trial would, first, be process outcomes, including:

• proportion of eligible participants who are randomised and complete follow-up

• proportion who obtain immediate (diagnostic) pain relief

• recruitment rates, that is, the number of referrals per site and proportion of referrals converted to participants. These data are needed to estimate the number of sites needed for, and the duration of, the main trial

• completeness of data from short-term electronic data collection.

Our primary effectiveness analyses was planned to be the overall difference in average pain and back pain related disability between intervention and control groups with a 95% CI. We planned to adjust our models for baseline stratification factors and other baseline covariates and missing values imputed. However, early closure precluded us from taking this approach, and the following plan was developed.

Health economic analyses

Patient Generated Index

The majority of PROMs, such as the RMDQ83 and Von Korff pain grade,99 contain a set number of standardised questions that may or may not include the outcomes that are considered important to patients. Individualised measures, such as the PGI, provide an alternative assessment format that enables respondents to specify those areas of life affected by the nominated condition that they judge to be most important, thus increasing content validity and relevance of assessment to the individual. 134 An international expert panel has recommended use of the PGI, alongside more standardised PROMs specific to LBP assessment, as a method of assessing the individualised nature of LBP and the meaning of recovery. 63

The PGI was originally developed in the UK in the 1990s for application across a range of conditions,135 including LBP. 135,136 Although PGI was widely utilised in prospective cohort studies to complement more traditional PROM-based assessment (e.g. Haywood et al. 137), the complicated formats of earlier versions meant that its application was limited in clinical trial settings. 134,138 However, recent revisions of the PGI have resulted in improved acceptability and completion rates. 137,139 The version of the PGI evaluated in this study retains the three completion stages of the original, but scoring (stage 2) and point spending (stage 3) have been revised. Stage 1 requires the respondents to list up to five important areas of life affected by their LBP. This is facilitated by the provision of a ‘trigger list’ generated for the target condition:134 a 17-item trigger list was created, informed by qualitative evidence of the impact of LBP,63,64,140,141 items included in widely used LBP-specific measures63,65 and recent application of the PGI in a Norwegian LBP population113 (Box 9).

A sixth predefined area relates to all other areas of life affected by their LBP and not listed separately. Stage 2 utilises a shorter, 7-point response scale (each nominated area is scored between 0 and 6, where 0 is as ‘bad as could possibly be’ and 6 is ‘as good as could possibly be’). Finally, in stage 3, a reduced number of points are available; respondents spend a total of 10 points to indicate the areas in which they would most value an improvement. A closed format version of the PGI was adopted for the clinical trial. At follow-up, respondents were presented with the first and the last columns completed using data brought forward from their baseline forms. They were then asked to score each item (stage 2) related to how they felt about it now.

The PGI score is generated by multiplying the area scores (stage 2) by the proportion of points awarded (stage 3) and summing and transforming to a 0–100 score, in which higher scores are representative of a better quality of life. 134 Here, we describe the first application of a revised format of the PGI in a UK LBP patient cohort and the first application of the revised format in a clinical trial setting.

Diagnostic assessment

We obtained diagnostic assessment data on 27 participants (Table 13). These show that our participants typically had very longstanding LBP, had tried multiple previous treatments and had continued with very or extremely troublesome LBP.

Best usual care

Adherence to study treatments was good. All participants, inevitably, attended the first BUC session. Three people (23%) randomised to BUC only subsequently withdrew from BUC, two because their LBP did not improve and one because they had problems travelling to the intervention. One of these participants also withdrew from follow-up. Attendance at BUC was good, with 21 out of 26 (81%) attending at least four treatment sessions (Table 14). Overall, the process data indicate that participants received the essential components of the package. This included acceptance, goal-setting and pacing. Participants were also given specific homework tasks to complete between sessions. In comparison with other studies, adherence for CBT for chronic back pain has shown to be associated with treatment gains in a variety of outcome measures including accomplishment of daily goals. 142 Attendance was good, with 23 participants (88%) attending at least half of the sessions.

Facet joint injections

One person randomised to injection decided against this after their first BUC session. We obtained injection records for 12 out of 12 (100%) of those injected. These showed that most people had had at least one injection at each spinal level and over one-third had had six facet joints injected. There is evidence that clinicians were making a judgement as to which joints were appropriate to inject. There was a wide variation in the reported time of exposure to radiation, ranging from 35 to 600 seconds. We have not been able to resolve queries on the total radiation exposure because of the incompatibility of different approaches to radiation exposure and different machinery. We also have some concerns about the quality of recorded data on the duration of radiation exposure because of the wide difference in times recorded.

We had immediate pre-injection and 60-minute post-injection data on pain for 12 participants (Figure 5). There was an overall reduction in pain scores, falling from 5.9 to 2.6 following injection. Seven participants reported a > 50% reduction in pain, five of whom reported that they were pain free following the injection. One participant reported that their pain was worse, and their pain score increased from 3 to 6 (Table 15). This study was not designed to assess the validity of our clinical assessment of possible facet joint pain against a gold-standard test. 119 Nevertheless, these data suggest that we have identified a population that has some potential to benefit from intra-articular FJIs, with one-third of participants achieving the target level of immediate pain relief (a 50% reduction) suggested by the consensus conference.

FIGURE 5.

Box plot of pre- and post-injection pain scores.

Adverse events

No adverse events resulting from either intervention were recorded.

Area under the curve for pain

Although there were missing data in the text messaging responses for pain, these were largely replaced with the results of the written pain diary (Table 16). A priori we set three assumptions for this, which were as follows:

1. If two text scores were submitted on the same day, then the earliest score that was received after 6 p.m. was used for the analysis. The other score would be ignored (n = 20 occasions).

2. If a written response was received instead of a number, then this was converted into a score (n = 1 occasion on which a participant replied ‘nine’ = pain and the score was calculated as 9 for this day).

3. If a weekly score was received after the specified day, then the score would be used. For example, if a score was requested on day 49 and received on day 50 during the weekly follow-ups, then this observation would count as the day 50 score (n = 8 occasions).

A simple imputation rule allowed us to estimate the AUC when no values were available for any individual day. This has provided good-quality data, allowing us to see the pattern of pain over the initial post-randomisation period. We achieved usable outcome data on 23 out of 26 (88%) of our participants for this, our primary outcome (Table 17). Visual inspection of the graph of the pooled outcome over time shows that there is a modest improvement over the first 35 days of daily measurement, with little further improvement after this time (Figure 6). This is also reflected in the responses given to questions on improvement since baseline in the 3-month questionnaire.

FIGURE 6.

Participants’ average pain score over lifetime of the project (n = 23).

The plot (see Figure 6) represents the mean pain score for all participants on each given day of data collection and shows the overall trajectory of pain scores. The dotted vertical line indicates the proposed date of the intervention at day 7 and the solid line at day 35 represents the transition point from daily scores to weekly scores.

Three-month questionnaire

Seven 3-month questionnaires were not received, resulting in a follow-up rate of 73% for the 3-month questionnaire outcomes. One of these participants withdrew from questionnaire follow-up prior to 3-month follow-up and the remainder did not respond.

The small numbers mean that extreme caution is needed in interpreting the follow-up findings. Nevertheless, all pain-related outcomes show the expected improvement between baseline and follow-up. For example, for the RMDQ the difference is 2.8 points on a 0–24 scale and for the Modified Von Korff pain scale the difference is 10 points on a 0–100 scale. However, for psychological and attitudinal outcomes (SF-12v2 mental component score, WEMWEBS, DAPOS and PSEQ) there does not appear to have been any meaningful change between baseline and follow-up. With the exception of the PGI, all of the measures used are established outcome measures and we will not assess their performance further within this study (Table 18).

Resource use and cost estimates of facet joint injection treatment for low back pain relief

The average resource use and costs estimates associated with the delivery of the FJI procedure for LBP relief are presented in Table 21. Cost categories included use of the treatment room/injection facility (equipped with a radiograph or equipped with C-arm fluoroscopy machine, resuscitation trolley and/or monitoring back-up equipment), staff time, anaesthetic drugs and disposable medical consumables. The mean duration of the FJI procedure was estimated to be 36 minutes (range 15–60 minutes) based on data from four of the trial participating centres. The unit costs for resource use variables were obtained primarily from the Personal Social Services Research Unit (PSSRU) Unit Costs of Health and Social Care 2015 compendium146 and the NHS Prescription Cost Analysis 2015 database. 147 The unit costs of syringes and needles and other medical consumables were obtained from online sources (www.medisave.co.uk, www.farlamedical.co.uk; both accessed 21 November 2015) when more direct NHS sources were unavailable. The cost of the treatment room facility was estimated to be £183.34. This was based on treatment lasting 36 minutes on average and on a treatment room cost of £175 per hour as reported in a report for NICE Technology Appraisal 279 (Stevenson et al. 145) and updated to 2015 prices using the NHS Hospital & Community Health Services inflation index. 146 It was assumed that the FJIs are performed in a similar treatment facility as percutaneous vertebroplasty or percutaneous balloon kyphoplasty for treatment of osteoporotic vertebral fractures, the interventions being appraised in NICE Technology Appraisal 279. The mean total cost of 36 minutes of clinical staff time was £191.60 and the mean total cost of drugs and disposable medical equipment/consumables was £44.28 per patient, respectively. Thus, the mean total cost of performing a FJI was estimated at £419.22 per patient.

Resource use and cost estimates for programme of physiotherapy treatment for low back pain

Table 22 presents the resource use and cost estimates associated with the delivery of the programme of physiotherapy treatment to relieve LBP as specified in the FIS feasibility protocol. The unit costs were obtained from the PSSRU Unit Cost 2015 compendium146 and the published Department of Health National Reference Cost 2014–15 schedules. 148 The mean total cost of the full course of physiotherapy treatment, consisting of one initial consultation (diagnostic assessment) followed by four follow-up treatment consultations, was estimated to be £264 per participant.

The mean total cost of the overall treatment package (comprising one injection and six physiotherapy sessions) was estimated at £683.22 per participant. This is similar to a NHS tariff cost for a course of FJIs of £686.84 (National Reference Costs 2014/15148 – main schedule, currency code AB16Z). What is not included here is cost of the initial consultant appointment (£144.79, National Reference Costs 2014/15148 – main schedule) at which a decision to offer FJIs is made. The minimum overall cost of the overall FJI package is thus £831.63. This does not account for the care costs of patients who have a consultant appointment and do not proceed to FJIs.

Health-care utilisation and costs during trial follow-up

Tables 23–27 present summaries of patient self-reports of health-care resource use and costs for the 3-month period prior to randomisation (medication use only) and the 3-month post-randomisation follow-up period for all resource use categories including pain relief medication. In total, 25 (96%) and 19 (73%) of the 26 participants provided complete data on medication use at baseline and the 3-month follow-up time point, respectively. The total cost of prescribed pain relief tablets, patches and gels was £86.27 (standard error £37.33) per patient and £33.34 (standard error £18.79) per patient in the 3 months prior to randomisation and the 3 months following randomisation, respectively (see Table 23).

Complete primary and secondary care utilisation data were available from 17 (65%) of the 26 study participants for the 3-month post-randomisation period (see Table 24). None of the study participants was admitted for an overnight stay in hospital and one patient was treated as a day case, generating a mean cost of £40.40 per patient. There was no reported use of some outpatient services such as blood tests and ultrasound investigations, as well as the majority of community care services. The mean (standard error) cost per patient for outpatient services ranged from £3.68 (£3.68) for hospital nursing services to £59.62 (£25.01) for consultant pain clinic services. The mean (standard error) cost of primary care services was £36.24 (£17.39) for GP surgery services and £0.93 (£0.63) for practice nursing services, with both lasting 11.7 minutes on average. The mean (standard error) total cost to the health services over the first 3 months post randomisation was £236.60 (£79.65). Private health-care utilisation was entirely self-funded by one patient reporting private health-care use and cost £38.24 per patient on average (see Table 25).

The health-care resource use and out-of-pocket expenses incurred by patients and their relatives are presented in Table 26. The mean cost per patient of over-the-counter pain relief medication, gels and patches increased from £2.92 (standard error £1.30) in the 3-month pre-randomisation period to £18.28 (standard error £5.05) in the 3-month post-randomisation period. The mean cost per patient of health-care-related travel (e.g. bus fares and taxis) was £1.15 (standard error £1.15) for study participants and £2.00 (standard error £2.00) for their partners. Overall, the mean total health-related out-of-pocket expenditure was £26.23 (standard error £7.19) per patient in the 3 months post randomisation. There were no reported costs associated with child care, laundry- and household-related services that could have been associated with the delivery of the study interventions.

The employment- and income-related status of trial participants at the end of the 3-month follow-up period is summarised in Table 27. Nine of the 17 patients with complete data on employment status were in work, six were retired and two were unemployed because of ill health. Twelve patients reported information related to loss of income: one patient reported lost income of £409 because of ill health, translating into an income loss of £34.08 (£34.08) per patient on average. The mean number of work days lost was 2.78 (standard error 2.22).

Health-related quality of life at baseline and 3 months’ post randomisation

Self-reports of the health-related quality of life of patients with LBP were made using the EQ-5D-5L instrument at baseline and at 3-month follow-up; the results are summarised in Tables 28 and 29, respectively. The proportion of missing data among responders was negligible and there was evidence of floor effects in the EQ-5D-5L dimensions related to self-care and anxiety and depression and, to some extent, the mobility, dimension at both baseline and end of follow-up. The mean health utility score on a scale (range –0.28 to 1), generated using recently derived UK tariffs for the EQ-5D-5L,126 changed from 0.50 (standard error 0.10) at baseline to 0.60 (standard error 0.20) by the end of the 3-month follow-up period. Similarly, patients’ self-rating of their health-related quality of life on the EQ-5D visual analogue scale (VAS; 0–100) indicated an improvement from 59.3 (standard error 4.4) at baseline to 65.6 (standard error 17.88) at the end of follow-up.

The plots in Figure 7 compare 3-month health-related quality-of-life profiles of patients generated using (1) self-completed questionnaires at baseline and at the end of follow-up (3-month time point) and (2) daily (during the first week) and then weekly (for the remainder of the 3-month follow-up period) self-assessment of health-related quality of life recorded in daily/weekly health diaries. Not all patients completed both versions of the questionnaire at each time point. Figure 7a shows the mean health-related quality-of-life profiles of all participants who completed the baseline/end of follow-up version of the EQ-5D-5L questionnaire (baseline, n = 25; follow-up, n = 19) and the daily/weekly diary records (n = 15). Figure 7b shows the profiles for a subset of participants (n = 11) completing both approaches to data collection. The diary scores indicate marked fluctuations in health-related quality of life (both EQ-5D-5L index and VAS scores) in the first few weeks post randomisation. In particular, there is a substantial short-lived improvement in both EQ-5D-5L index and VAS scores in the first few days after the initial treatment. The fluctuation in health-related quality of life is less pronounced during the remainder of the follow-up period, almost returning to the baseline rating at the end of the 3-month follow-up period.

FIGURE 7.

Three months’ QALY profile of patients with LBP generated via daily and weekly self-reports of EQ-5D-5L scores and self-completed questionnaire at baseline and at end of follow-up. (a) Scores of all participants who completed the postal questionnaire or diary records; and (b) scores from a subset of participants who provided postal questionnaire and diary EQ-5D data.

The QALY estimates for the 3-month follow-up period were generated based on the EQ-5D index-derived utility values for the 11 patients who reported both baseline/follow-up questionnaire and daily/weekly diary data. The mean QALY estimate was 0.141 (standard error 0.018) based on health-related quality-of-life outcomes measured at baseline and at the end of follow-up, and 0.170 (standard error 0.017) based on the more regular diary records. Over the 3-month follow-up period, the mean QALY difference generated by daily/weekly assessment of health-related quality of life, compared with the normal practice of assessing health-related quality of life at less regular time points (typically at baseline and end of follow-up), was 0.029 (standard error 0.01).

Figure 7a is based on the scores of all participants who completed the postal questionnaire (baseline, n = 24; follow-up, n = 19) or on diary records (n = 15). Figure 7b is based on scores from a subset of participants (n = 11) who provided both postal questionnaire and diary EQ-5D data.

Project timeline of study approvals

Figure 8 gives a brief overview of the study approvals and includes planned and revised timelines. Full ethics approval was obtained in September 2014 in line with our planned timeline, although we had originally planned to have an outline approval in early 2014 with an updated/amended approval post consensus in July/August 2014. The ethics committee rejected our early submission and advised that we should resubmit after we had held our consensus conference, which we did.

FIGURE 8.

Original revised and actual timelines, approvals and site set-up.

Full ethics approval is the driver to start the formal processes of obtaining research and development (R&D) and site approvals. This proved to be a long process. This study sought approvals at the time that an older system of approvals was being changed over to a newer system, operated by the Health Research Authority, and this may explain some of the severe delays we experienced in setting up sites. Our request for global R&D approval was submitted in August 2014 but was not approved until February 2015, which was a far longer period of time than normal and meant that approval was gained far later than planned. Repeated requests for updates from the team, who had provided all of the relevant materials, did not help to resolve this. Once global R&D was approved, each of the sites required site approval that had to be in place before any recruitment took place. This again proved to be a slow process. One site was approved within 1 month, three sites were approved in 4 months, and for the final two sites approval took over 5 months.

These delays, which were largely outside the study team’s control, impacted on the timeline of the study. The planned 6-month recruitment process that was meant to start in November 2014 was delayed for 7 months until June 2015. The study team provided the staff training needed at three of our sites in preparation for the start of recruitment in late 2014.

Study research sites

Reach of the study across the eligible population

It was necessary to know the size and characteristics of the population that had the potential to benefit from the intervention to identify if the recruited population was comparable. Although Table 30 gives an indication of the overall population that our study pain clinics serve, we have little information on the local prevalence of back pain.

Study site 2, which provided a pain management service, collected data on referrals received and treatments provided during a period of 8 months that overlapped with the study period. During this time they received 335 referrals, of which 116 (35%) had back pain. Of the patients with back pain, 34% received physiotherapy, 22% received advice only and 13% were referred to a pain management group. Four patients received FJI, of whom one was recruited to our study and the remainder were seen by the service outside our recruitment phase. It is not possible to draw a conclusion about the reach of our study from this audit but it suggests that pain management services may see only small numbers of patients who may benefit from FJIs.

Screening and recruitment participants

The study plan was to recruit 150 participants in a 6-month recruitment period. It was calculated that each site would recruit around 30 patients. Recruitment was planned to be primarily from pain clinic services, which was an approach that was congruent with the commissioning brief. As noted at the beginning of this chapter, recruitment started on 26 June 2015 and was terminated by the funder on 11 December 2015 when only 320 patients had been approached, which had resulted in the 26 patients who were randomised. In the following sections we look at the recruitment process in more detail.

Delivery of the interventions to recruited patients

Process evaluation strengths and weaknesses

The process evaluation provided us with insights about implementing this complex study within the NHS. Engagement with the clinical teams throughout the study allowed us to collect qualitative data about the study while they were fresh to those involved with the study. Study teams were willing to share positives and negatives about the study.

Our recruitment of participants to interview was disappointing and reflects the low and slow recruitment rate to the study. We have used the interview data from study participants cautiously to avoid overclaiming from this very small data set.

Early pain outcomes

Short-term benefits and harms from interventional procedures are important to capture. The conventional approach of using postal questionnaires for which outcomes are often not collected until 1 or 3 months after randomisation will not capture these. It may be that, even if there is little difference in long-term outcomes, the short-term benefits mean that FJIs are worthwhile. Our original intention had been to collect these using text messaging only. It became clear as the project developed that we also needed to offer a paper alternative. Indeed, the age range of our participants, up to 80 years, suggests that using text messaging alone might have meant that we missed useful information. Although there were missing data in both text and paper versions of daily/weekly pain scores, we obtained sufficient data to have useable outcome data on the pattern of pain over the first 3 months following randomisation of 23 out of 26 (88%) participants. This is an important development.

Patient Generated Index

The revised PGI was correctly completed by all baseline questionnaire respondents, although these constituted a small cohort. This compares favourably with completion rates in another study that applied the revised format in a LBP population (89% completion)113 and with the original format that was completed in a LBP population (68% completion). 136

A Norwegian translation of the revised PGI has recently been completed by LBP participants in a longitudinal postal survey. 113 Participants listed more than 380 areas in stage 1; the most frequently reported were pain (65%), sleep (47%), stiffness (34%), socialising (34%) and housework (33%).

We found that the impact of LBP on an individual’s ability to walk and sleep were the most frequently reported areas of concern, at 52% and 48%, respectively. In the Norwegian study,113 sleep disturbance was identified as the second most frequent area of concern at 48% after pain, at 65%; however, only 24% of our participants identified pain as an important symptom. Fatigue is increasingly recognised as an important symptom across many chronic, long-term and rheumatological conditions. 153,154 We found that 20% identified fatigue as a concern. Neither sleep nor fatigue is part of the RMDQ. 83 The Oswestry Disability Index155 includes sleep but not fatigue. Comparisons between the areas of concern identified within these two aforementioned studies of the PGI and the domains within leading back pain in specific outcome measures indicate discrepancies between what participants consider to be important and the items we are measuring in our trials. A systematic review of qualitative studies has also identified that current outcome measures may not be adequately capturing the social component of the impact of LBP. 156

Although the numbers in our study are limited, people appear to have been able to complete the adapted follow-up PGI score satisfactorily. The results show the expected improvement over time. This was also the experience in the Norwegian study. 113 Overall, our preliminary data indicate that it would be worthwhile exploring further the performance of the PGI as an outcome measure in back pain trials.

Health economic outcomes

There are two components ensuring that we have satisfactory health economic outcomes: costs and outcomes. We have produced detailed micro-costing of both the BUC and the injection treatments that reflect the true cost of delivery.

Our overall micro-costing of £419.22 per injection compares with a NHS tariff cost for FJIs of £686.84 (National Reference Costs 2014/15148 – main schedule, currency code AB16Z). What is not included here is cost of the initial consultant appointment (£144.79, National Reference Costs 2014/15148) at which a decision to offer FJIs is made.

Although the number of completed follow-up questionnaires is limited, we have been able to satisfactorily estimate NHS and personal health-care costs from our participants.

An important concern at the start of the study was how to capture short-term benefits and harms. We described earlier the process for collecting and analysing pain data. We also wanted overall health utility data to assess how these early changes might impact on cost-effectiveness. For these, we sought to collect data weekly throughout follow-up rather than the early data we used for pain. This was to reduce questionnaire burden. We had originally planned to collect these electronically, but the only licensed versions of the EQ-5D were not suitable for use when linked to smartphones, which would have allowed a seamless collection of data on pain and the EQ-5D. For this reason, the pain score was completed as weekly diaries. Participants found these difficult to complete fully and we received only 13 fully completed sets of data. These limited data show a very interesting pattern, with a marked short-term increase in health utility soon after joining the study (see Figure 7). At the end of 3 months, the diary data are similar to the data collected in the final postal questionnaire. There is a clinically important difference in QALYs accruing over this time according to the two collection methods of 0.029 (standard error 0.01). This provides us with a challenge for the main study, as there are potentially important short-term effects that should be captured in the main trial for which we do not have an approach to data collection that yields data on enough participants.

In developing our main trial proposal, we will develop a decision-analytic model to simulate the clinical pathway of patients with LBP within the UK NHS and over the lifetime of the individual. The model parameters for prevalence of LBP in the target population, health outcomes (in QALYs) and health-care utilisation and associated costs will be informed by results of the FIS feasibility study and systematic review of the relevant literature. Bayesian evidence synthesis methods will be used to derive distributions for model parameters based on all available information. The model will be used to estimate the expected value of sample information of a trial of the protocolised intervention and management plan, using algorithms developed by Madan et al. 157 and Strong et al. 158 The expected value of sample information calculations will be performed for alternative study designs to determine the optimal design in terms of factors such as sample size, length of follow-up and study population.

Recommendations for research

A definitive trial of adding intra-articular FJIs to BUC for people with suspected facet joint pain that is not resolving after conservative treatment.

University Hospitals Coventry and Warwickshire NHS Trust

Shyam Balasubramanian (PI and Injector).

Davina Hewitt (Research Nurse).

Nicolas Aldridge (Research Nurse).

Sarah O’Toole (Research Nurse).

Stefanie Gwynne (Administrator).

Janet King (Administrator).

Jo Freelove (Diagnostic Assessor).

Tim Rhodes Jones (Treating Physiotherapist).

Lizzie Fort (Treating Physiotherapist).

Lois Burton (Treating Physiotherapist).

Allan Jackson (Treating Physiotherapist).

Atique Rehman (Treating Physiotherapist).

Mojid Khan (Lead Pharmacist).

Worcestershire Acute Hospitals NHS Trust

Alan Bennett (PI and Injector).

Julie Wollaston (Research Nurse).

Terry Martin (Research Nurse).

Sharon Williams (Research Sister).

Janice Gastinger (Diagnostic Assessor).

Vicky Chapman (Diagnostic Assessor).

Dianne Spalding (Treating Physiotherapist).

Judith Jehring (Treating Physiotherapist).

Ann White (Lead Pharmacist).

Sherwood Forest Hospitals NHS Foundation Trust

David Walsh (PI and Diagnostic Assessor).

Bala Veemaraon (Injector).

Yvette Girvan (Research Nurse).

Deborah Wilson (Research Nurse).

Debbie Bottomley (Research Nurse).

Stephen Bliss (Diagnostic Assessor).

Asha Gudibandi (Treating Physiotherapist).

Amanda Buckley (Treating Physiotherapist).

Linda Otter (Lead Pharmacist).

South Warwickshire NHS Foundation Trust

Hugh Antrobus (PI and Injector).

Dev Talik (Injector).

Cheryl Ritchie (Research Facilitator).

Suzanne Jones (Diagnostic Assessor).

Roger Wedden (Diagnostic Assessor).

Glynn Alderman (Treating Physiotherapist).

Elizabeth Rourke (Treating Physiotherapist).

Julia Jones (Lead Pharmacist).

South Tees Hospitals NHS Foundation Trust

Sam Eldabe (PI and Injector).

Vicky Phelps (Research Nurse).

Morag Brookes (Research Nurse).

Grace Madzinga (Research Nurse).

Karoline Middleton (Administrator).

Nicolas Harland (Diagnostic Assessor).

Amada Hughes (Diagnostic Assessor).

Victoria Robinson (Diagnostic Assessor).

Crystal Rosser (Treating Physiotherapist).

Richard King (Treating Physiotherapist).

Samantha Cook (Treating Physiotherapist).

Jason Wong (Lead Pharmacist).

The team wish to acknowledge all of those who contributed to the FIS consensus conference (see tables below).

The team acknowledge the support of the Revalidation and Continuing Professional Development Team at The Royal College of Anaesthetists for providing continuing professional development points for this event. We also extend thanks to Warwick conferences for the excellent facilities and refreshments.