Posterior cervical foraminotomy versus anterior cervical discectomy for Cervical Brachialgia: the FORVAD RCT

Anatomy

The cervical (neck) spine consists of seven vertebral bones, termed C1–C7, that are connected to each other by ligaments, the intervertebral discs and the facet joints. Vertebrae have a large vertebral body positioned ventrally (anterior) and in the midline with bilateral lateral masses positioned dorsally (posterior). The pedicles connect the vertebral body to the lateral masses and the laminae connect the lateral masses to the midline and dorsal spinous process. The intervertebral disc is positioned between the vertebral bodies, and the facet joints are positioned between the lateral masses. There are multiple ligaments, the most important of which, in the pathophysiology of brachialgia, is the ligamentum flavum, which is elastic and runs between adjacent laminae.

The spinal cord is positioned within the cervical bones in the spinal canal and gives rise to eight pairs of nerves, one pair for every spinal level. The first section of each nerve is called the nerve root, and it exits from the spinal canal via the nerve root foramen (Figure 1). Each nerve exits above its equivalently named vertebra, with the exception of the C8 nerve, which exits below the C7 vertebra and above the first thoracic vertebra. The nerve exits at a 40–45° angle to the coronal plane and at a 10° angle to the axial plane. 1

FIGURE 1.

An oblique view of a model of the left side of the lower cervical spine to demonstrate the margins of the C5/C6 nerve root foramen. The ligamentum flavum is medial to the FJ and not shown. D, C5/6 interarticular disc; FJ, facet joint; N, C6 nerve root; P, C5 pedicle; VB, C5 vertebral body.

The nerve root foramen is bounded ventrally (anteromedial) by the intervertebral disc, cranially (superior) by the pedicle of the vertebra above, dorsally (posterolateral) by the facet joint and ligamentum flavum and caudally (inferior) by the pedicle of the vertebra below. The nerve root foramen is widest at the cranial end and tapers at the caudal end. The nerve root is normally positioned at the widest section towards the cranial end of the foramen.

Degenerative disease

As people age, degenerative disease affects the intervertebral disc, facet joint and ligamentum flavum; consequently, the nerve root foramen may change shape and the nerve root becomes compressed. Most commonly, the intervertebral joint will develop bony growths called osteophytes that cause ventral compression of the nerve root; alternatively, the facet joint or ligamentum flavum may hypertrophy, causing dorsal compression of the nerve root.

The rate at which degenerative disease develops is related to the amount of movement, especially flexion, which occurs at a joint. In the cervical spine, the C5/C6 and C6/C7 joints flex more than any others;2 consequently, these are the most common locations for cervical degenerative disease. The nerve that exits between C5 and C6 is the C6 nerve, and the nerve that exits between C6 and C7 is the C7 nerve; therefore, these are the most commonly affected nerves.

Symptoms and signs of cervical radiculopathy

Cervical nerve root compression at the nerve root foramen causes a syndrome called cervical radiculopathy. Cervical radiculopathy is characterised by the following:

• neck pain

• brachialgia – the symptom of pain coming from the neck and radiating into the upper arm

• symptoms and signs related to nerve dysfunction, which include sensory symptoms (e.g. pain, pins and needles, and sensory loss), or, less commonly, motor symptoms (weakness, muscle loss and diminished reflexes). These symptoms and signs are in the distribution of the nerve affected (Table 1).

Brachialgia incidence

The reported incidence of cervical brachialgia is 1.79 cases per 1000 per year,3 with > 110,000 cases of brachialgia annually in the UK.

Cervical brachialgia typically affects people aged 40–60 years, with up to 15% of patients unable to work owing to the pain. 4 In a large registry study (1809 patients), patients had significantly worse scores than the general population in all eight Short Form questionnaire-36 items quality-of-life dimensions. 5

Natural history and non-surgical management

In most patients, brachialgia is self-resolving with conservative management including analgesia and physiotherapy. 6 Foraminal injections may also be used to provide analgesia. Surgery is not normally indicated in the first 6 weeks of treatment; however, 26% of patients will undergo surgery if their symptoms are persistent or remain debilitating after at least 6 weeks. 6

Role of surgery in the treatment of cervical radiculopathy

Among patients for whom symptoms persist 6 weeks to 3 months after onset, randomised controlled trials (RCTs) have shown that surgery results in a more rapid recovery than further conservative management, with significantly better pain intensity, muscle weakness and sensory loss at 4 months. 7,8 However, by 12 months after the surgery, the difference between the surgical and the non-surgical groups is smaller,9 and some studies have shown no difference. 7 The loss of benefit is due to both ongoing improvement of the conservative group and late recurrence of symptoms in surgical patients. 7,10 Careful patient selection and informed consent are therefore critically important.

Surgical treatment of cervical radiculopathy is targeted at decompressing the nerve as it passes through the nerve root foramen and, in anterior cervical discectomy (ACD), at fusing the joint so that further degenerative disease cannot occur at this point. Operations may be performed using an anterior approach (ACD or anterior cervical arthroplasty) or a posterior approach [posterior cervical foraminotomy (PCF)].

Efficacy

The efficacy of both ACD and PCF in the management of cervical brachialgia is well established.

Matz et al. 25 conducted a systematic review of 13 retrospective and three prospective studies on the surgical outcomes of ACD for cervical brachialgia. They concluded that ACD provides rapid relief (within 3 to 4 months) from arm and neck pain, weakness and/or sensory loss. Improvement in motor function at 12 months was also noted.

Heary et al. 26 conducted a systematic review of 13 studies investigating the surgical outcomes from PCF. Although the studies identified were observational, retrospective and often lacked validated outcome measures, they concluded that PCF is an effective treatment for cervical brachialgia.

Safety

Most authors have reported fewer surgical complications with PCF than with ACD,11,13,35,37,43 although some authors have found comparable complication rates. 12,19,21,38

Korinth et al. 35 found that surgery-related complications (dysphagia, hoarse voice, transient neurological deficit and postoperative haematoma) were observed in 6.5% of ACD patients, compared with 1.8% of PCF patients (cerebrospinal fluid fistula, wound infection and transient neurological deficit) (p < 0.05).

Witiw et al. 11 retrospectively analysed American insurance data; 4851 PCF and 46,147 ACD procedures were included. The reported complications for PCF and ACD were death (0% and 0.01%, respectively), vascular injury (0% and 0.02%, respectively), dysphagia and hoarse voice (0.17% and 1.62% for PCF and ACD, respectively), cerebrospinal fluid (0% and 0.02%, respectively) and deep-vein thrombosis (0.06% and 0.14%, respectively). Wound infections were more common with PCF (2.05% and 0.57%, respectively). The mean length of stay was shorter with PCF [1.47 (SD 1.39) vs. 1.23 (SD 1.06) days].

Degenerative disease occurring at the adjacent spinal level following ACD occurs at a rate of 2.9% per annum; this may be a consequence of the fusion that occurs with this operation. 15 PCF maintains movement,20,39,40 and may therefore reduce the incidence of degenerative disease occurring at adjacent segments.

Health economics

Tumialán et al. 36 found that both ACD and PCF procedures had similar median operating times (151.8 vs. 154.0 minutes), median blood loss (32.6 ml vs. 41.3 ml) and analgesic use, but the mean direct cost of PCF was US$20,094–30,553 lower per case than for ACD.

Alvin et al. 37 performed a retrospective 1-year cost–utility analysis on 45 patients to determine the cost effectiveness of ACD in comparison with PCF for patients with single-level cervical radiculopathy. The authors found PCF to be less costly (US$12,777 vs. US$18,473) and more cost-effective [0.16 vs. 0.14 increase in quality-adjusted life-years (QALYs) over 1 year] than ACD.

Witiw et al. 11 analysed American insurance data for 4851 PCF and 46,147 ACD procedures, and found that the mean PCF cost was US$15,281 whereas the mean ACD cost was US$26,849 (p < 0.001).

Mansfield et al. 41 conducted an American retrospective cohort study and calculated the mean direct cost of PCF and of ACD among 101 patients with cervical radiculopathy. The mean cost of ACD was US$8192; the mean cost of PCF was US$4320.

In our UK institution (Leeds Teaching Hospitals NHS Trust), the average cost of PCF is £4200 and the average cost of ACD is £5380. This is mostly because of the cost of the implant used in an ACD. The average cost of PCF is, therefore, lower, and its widespread implementation could lead to significant savings for the NHS. However, there is no published economic study of these surgical treatments from a UK NHS perspective.

Primary objective

The primary objective was to determine whether or not PCF is superior to ACD in terms of improving clinical outcome as measured by the NDI at 52 weeks post operation.

Secondary objectives

The secondary objectives were to compare PCF and ACD in terms of:

• NDI scores over 52 weeks post operation

• neck and upper-limb pain, including the shoulder, arm and hand, over 52 weeks post operation

• dysphagia (difficulty swallowing) and globus (sensation of a lump in the throat) over 52 weeks post operation, as assessed by the participant

• hoarse voice over 52 weeks post operation, as assessed by the participant, and at 6 weeks post operation for a randomly selected subset of participants, as assessed by expert review

• extent and severity of a patient’s spinal cord functional impairment, including upper-limb nerve root function, at day 1 and at 6 weeks post operation

• incidence of revision surgery over 52 weeks post operation

• incidence of surgical complications up to 6 weeks post operation

• cost effectiveness over 52 weeks post operation.

Exploratory objectives

The exploratory objectives were to explore the impact of variations in the optional surgical components of PCF and ACD on the NDI and EuroQol-5 Dimensions, three-level version (EQ-5D-3L) scores. The types of variation in the optional surgical components are as follows:

• ACD – fusion with a cervical plate in addition to ACD, versus simple ACD without a plate

• PCF – posterior cervical minimal-access technique versus open-access technique.

Internal pilot phase

A 12-month internal pilot phase evaluated the feasibility of recruitment within the planned timelines, based on the number of actively recruiting sites and overall average recruitment rate. The pilot also aimed to assess early safety data, data validity, compliance with trial procedures, compliance and completeness of patient-completed quality-of-life questionnaires, eligibility of the clinical teams and subtype of procedure performed (e.g. minimal-access or open-surgery PCF).

At the end of the internal pilot phase, trial progress was reviewed by the trial funder against predefined progression criteria (see Appendix 1) to determine whether or not the trial could proceed in full.

Patient and public involvement statement

Six members of the public, including four postoperative patients, provided feedback on the proposed trial design at the grant application stage. The feedback provided was used to inform the selection of appropriate patient-centred quality-of-life outcomes, resulted in improvements to the proposed patient information sheet and informed appropriate follow-up procedure for data collection.

A PPI representative was a member of the Trial Management Group (TMG) and a second PPI representative was a member of the independent TSC, to provide insight and experience from the patients’ perspective, ensuring that the trial maintained focus on service users and needs. Their key roles involved input into patient information documentation, including drafting and reviewing information leaflets and other patient resources to ensure that patient information was meaningful and clear.

Summary of protocol changes

Protocol changes made during the trial are summarised in Table 2.

Inclusion criteria

The trial inclusion criteria were as follows:

• age ≥ 18 years

• diagnosis of unilateral cervical brachialgia as confirmed by MRI or CT myelogram taken within the preceding 12 months

• symptoms of cervical brachialgia present for at least 6 weeks

• single-level nerve entrapment

• posterolateral disc and/or foraminal narrowing

• failed conservative management (including, but not limited to, medication, physiotherapy and modification of daily activities)

• able and willing to comply with the terms of the protocol, including quality-of-life questionnaires (which were provided in English only; for this reason patients had to be English-speaking)

• able to provide written informed consent.

Exclusion criteria

The trial exclusion criteria were as follows:

• cervical disc causing cord compression

• cervical myelopathy

• bilateral cervical brachialgia

• previous cervical spine surgery

• professionals for whom a hoarse voice would be exceptionally significant (e.g. singers or speakers)

• skin disease at surgical sites (e.g. eczema)

• pregnancy

• cervical deformity

• not suitable for ACD

• not suitable for PCF.

Trial setting

The FORVAD trial opened to recruitment at NHS hospital trusts, all of which had to fulfil a set of prespecified criteria and complete a registration form that verified that the research site was willing and able to comply with the trial requirements prior to trial participation.

To be eligible to participate in the trial, research sites had to:

• be able to perform both ACD and PCF

• have the capacity to recruit at least 10 participants per year.

Once site eligibility had been confirmed, research sites were required to obtain local management approval, return all essential documentation to the Clinical Trials Research Unit (CTRU) and undertake a site initiation with the CTRU prior to the start of recruitment to the trial.

In addition to site eligibility, participating surgeons were required to have performed a minimum of 10 PCF and 10 ACD operations as the primary surgeon, and to complete a bespoke training package hosted by the e-brain platform (www.ebrain.net).

Pre-operative interventions and preparation

Pre-operative investigations and preparation were as per individual site protocols.

Posterior cervical foraminotomy

Anterior cervical discectomy

Postoperative care

Postoperative care was as per institutional protocol. Postoperatively, the participant could mobilise immediately as pain allowed. If used, the drain was removed the following day. Unless the participant developed a new neurological deficit, postoperative imaging was not mandatory, but surgeons could elect to perform anterior–posterior and lateral cervical spine radiography if they deemed it necessary. It was anticipated that participants would be discharged the day after physiotherapy review.

Registration

Following confirmation of written informed consent and eligibility, patients were registered into the trial by authorised members of staff at the trial sites. Registration was required to take place at least 1 day and no more than 28 days prior to the planned surgery date, to allow trial-specific assessments to take place. Registration was performed centrally using the CTRU automated 24-hour telephone randomisation system. Authorisation codes and personal identification numbers (PINs), provided by the CTRU, were required to access the registration system. At the point of registration, participants were allocated a five-digit trial identification number.

Day of surgery (day 0)

Postoperative follow-up assessments

Participants were reviewed for trial purposes on the ward at day 1 post operation and at 6 weeks in clinic post operation. At both time points, participants underwent the restricted ASIA assessment (described in Restricted American Spinal Injury Association assessment) and completed quality-of-life questionnaire booklets (see Participant-completed questionnaires). In addition, those participants randomly selected at registration to undergo an additional assessment of hoarse voice were asked to provide a recording of their voice at 6 weeks post operation. Participants were not required to indicate whether or not they considered their voice to be ‘normal’ at this time.

Any further clinic visits were according to local standard clinical practice. If participants required any further intervention for brachialgia, as per routine NHS practice, further clinical intervention was permitted. If a participant did receive additional treatment, information on the type of intervention, the details of the treatment received and the reason for the intervention were collected on the trial case report forms.

Primary outcome measure

The primary outcome measure was the NDI44 at 52 weeks post operation. The NDI is a validated45 patient-reported measure consisting of 10 items that assesses different aspects of daily functioning among patients with neck pain. It comprises four items regarding subjective symptoms (pain intensity, headache, concentration and sleeping), four items regarding activities of daily living (lifting, work, driving and recreation) and two items regarding discretionary activities of daily living (personal care and reading). 46 Each item is scored from 0 (best) to 5 (worst) and the total score is expressed either as a raw score (0–50) or rescaled as a percentage score (0–100), with a higher score corresponding to greater disability. In this trial, the primary outcome measure has been expressed as a percentage score. If two or fewer items were missing, the questionnaire was scored using the maximum attainable score as the denominator in the percentage calculation, as per the scoring manual. For example, if 9 out of 10 items were present, the maximum attainable score was 45, instead of 50, but the maximum percentage score remained 100.

Secondary outcome measures

Exploratory outcomes

• Whether participants receiving PCF received minimal- or open-access surgery.

• Whether participants receiving ACD received surgery using a plate or without using a plate.

Analysis

Formal statistical analysis of outcome measures was not possible as a result of the limited sample recruited to the trial. For the analysis of the primary outcome measure, it was intended to utilise a multilevel linear regression model incorporating random effects with respect to centre, and adjusting for baseline (day 0) NDI score and minimisation factors (duration of upper-limb symptoms and smoking status). The statistical analysis plan for the trial was updated to account for early trial closure and subsequent limited sample size prior to the final analysis of the trial. All descriptive statistics were computed using SAS^® version 9.4 (SAS Institute Inc., Cary, NC, USA).

Analysis populations

Primary and secondary outcome measures

Primary and all secondary outcome measures are summarised using descriptive statistics by treatment group. For participant-reported outcomes, item-level summaries and overall scores are provided in Chapter 3 and Appendix 5 by time point and treatment group.

Imputation of missing questionnaire responses was not performed, except as described in Patient-reported outcomes for the PainDETECT score categories.

Exploratory outcome measures

Owing to early trial closure, the analysis of exploratory outcomes is limited to a descriptive summary of the predefined PCF variation (i.e. operation performed using open- or minimal-access technique) or the ACD variation (i.e. operation with or without a plate) conducted in the trial.

Non-COVID-19-related deviations

There are two reported deviations of participants completing the baseline questionnaire on the date of registration instead of the date of randomisation; they were mistakenly not given the envelope to seal the questionnaire booklet in. This occurred once in each treatment group.

COVID-19-related deviations

COVID-19 was explicitly reported as an issue in the conduct of the trial in 10 instances. Six of these relate to postal questionnaires being sent out early, four in the ACD group and two in the PCF group. Two of the issues relate to ASIA assessments that could not be performed because of COVID-19; one of these could not be performed because the participant was assessed over the telephone instead of in clinic. Finally, two registered patients were not randomised as a result of COVID-19, as the pandemic meant that one had their surgery at an external facility and another had their surgery cancelled.

Baseline demographic and clinical data

Summaries of minimisation factors at baseline are reported in Table 5 and data collected at the baseline clinic visit are summarised in Table 6. The mean age was 54.0 years (SD 8.1 years, range 34–70 years). A total of 15 participants (62.5%) were female, and the ethnic background of all trial

participants as reported at baseline was white. In the 6 months prior to baseline, five participants had received physiotherapy (21.7%), four in the PCF group and one in the ACD group, and one participant in the PCF group had been seen by a chiropractor (4.3%). No participants had been treated with cervical nerve root injections.

Baseline data for outcome measures are reported in Table 7.

Comparability of treatment groups at baseline

The baseline characteristics listed in Table 6 appear to be balanced across treatment groups. Notable exceptions to this are the cause of nerve root compression, depression in the previous 6 months, duration of upper-limb symptoms and the arm that participant has worse pain in. A proportionally higher number of participants in the PCF group had depression in the preceding 6 months (28.6%, compared with 11.1% in the ACD group). The duration of upper-limb symptoms is more commonly > 12 months in the ACD group (77.8%, compared with 57.1% in the PCF group), and the arm that participants have worse pain in is more commonly the left arm in the ACD group (66.7%, compared with 42.9% in the PCF group).

For outcome measure data (see Table 7), the median baseline NDI score was 44.0 (IQR 36.0–62.0) in the PCF group and 35.6 (IQR 34.0–44.0) in the ACD group, meaning that, at baseline, the ACD group was observed to have a better (i.e. lower) NDI score; this was influenced by outlying higher scores observed by chance in this small sample.

Primary outcome: Neck Disability Index score at 52 weeks post operation

Summaries of the primary outcome, percentage NDI score at 52 weeks post operation, are presented in Figure 4, alongside baseline (day 0) summaries of percentage NDI score. Mean scores are denoted in figures by circles for PCF and by triangles for ACD. In general, a lower NDI score is better, that is it corresponds to a lower level of neck pain. The minimum clinically important difference for the NDI is 10%. At 52 weeks post operation, the median NDI score was 25.3 (IQR 20.0–42.0) in the PCF group and 45.0 (IQR 20.0–57.0) in the ACD group. Summary tables of all descriptive statistics computed for this outcome measure can be found in Appendix 4.

FIGURE 4.

Percentage NDI scores at baseline (day 0) and week 52, by treatment group.

Secondary outcome: repeated Neck Disability Index score assessed over 52 weeks postoperatively

Summaries of percentage NDI scores at baseline, at day 1 and at 6, 12, 26, 39 and 52 weeks post operation are presented in Figure 5. The median NDI score at day 1 post operation is higher, indicating worse functional disability, than at other time points. Thereafter, there does not seem to be a clear trend across time points, or a clear difference between the treatment groups in the NDI scores.

FIGURE 5.

Percentage NDI scores at each time point, by treatment group.

The summary statistics calculated for each time point, including medians and IQRs, can be found in Appendix 4, Table 25.

Secondary outcome: Numerical Rating Scale scores for neck and arm pain

Summaries of NRS scores (0–10) at baseline, at day 1 and at 6, 12, 26, 39 and 52 weeks post operation are presented in Figure 6 (NRS-NP) and Figure 7 (NRS-AP). In general, a lower NRS score is better, that is it corresponds to a lower level of pain. It appears that, following an initial increase in neck pain score at day 1 post operation [median of 8.5 (IQR 6.0–10.0) in the PCF group and 7.0 (IQR 4.0–9.0) in the ACD group], neck pain then decreases over the postoperative period in both treatment groups. At 52 weeks, the median NRS-NP scores were 4.0 (IQR 2.0–5.0) in the PCF group and 5.0 (IQR 3.0–7.0) in the ACD group.

FIGURE 6.

The NRS-NP scores at each time point, by treatment group.

FIGURE 7.

The NRS-AP scores at each time point, by treatment group.

For upper-limb pain, median arm pain reached its lowest level in both groups at 12 weeks post operation; thereafter, median arm pain increased to week 52 post operation in the ACD group and fluctuated over time in the PCF group, although IQRs overlap at each time point. At 52 weeks, the median NRS-AP scores were 5.0 (IQR 2.0–7.0) in the PCF group and 5.0 (IQR 3.0–6.0) in the ACD group.

The median surgical wound length was 32.0 mm (IQR 25–50 mm) in the PCF group and 39.0 mm (IQR 35–41 mm) in the ACD group. Although the median wound length was smaller in the PCF group, it was also much more variable, as demonstrated by the IQRs.

The summary statistics calculated for each time point, including medians and IQRs, can be found in Appendix 5, Table 27.

In addition, the number of participants who took pain medications in the previous 24 hours and the amount by which participants felt that their pain was reduced (on a NRS from 0–10) is presented in Appendix 5, Table 28. At any given time point, it appears as though the majority of participants [15 (65.2%) at 52 weeks] had taken pain medication in the previous 24 hours (across both treatment groups), and that pain was reduced to a greater degree in the PCF arm [median of 6.0 (IQR 2.0–7.0) in the PCF group and 3.5 (IQR 2.5–4.5) in the ACD group] at 52 weeks (see Appendix 5, Figure 15).

Secondary outcome: the PainDETECT questionnaire

Summaries of PainDETECT category scores at baseline and at 6, 12, 26, 39 and 52 weeks post operation are presented in Table 8. Categories represent different types of pain: a score of 0–12 means that pain is nociceptive [neuropathic pain is unlikely (< 15% probability)], a score of 13–18 is an unclear result and a score of 19–38 represents a high probability of neuropathic pain (> 90% probability). Table 8 summarises the score categories after considering best- and worst-case scenarios for missing responses. If best- and worst-case scores agree (are contained within the same scoring zone), participants are reported in the corresponding score category; otherwise, they are reported as ambiguous. It appears that, across time points, PainDETECT scores fluctuate in the PCF group, but reduce in the ACD group.

Summaries of the PainDETECT total scores can be found in Appendix 5, Table 29.

Secondary outcome: the Eating Assessment Tool-10 items

Summaries of the EAT-10 total scores at baseline, at day 1 and at 6, 12, 26, 39 and 52 weeks post operation are presented in Figure 8. In general, a lower EAT-10 score is better, that is, it corresponds to less severe swallowing problems. Up until week 6 post operation, the ACD group was observed to have worse (higher) EAT-10 scores, after which there was no clear difference between groups. The shaded section of Figure 8 shows the section of the graph where a participant would not be considered to have a swallowing problem, corresponding to an EAT-10 score of ≤ 3. 47

FIGURE 8.

The EAT-10 total score at each time point, by treatment group.

Summary statistics calculated for each time point, including medians and IQRs, can be found in Appendix 5, Table 31. At 52 weeks, the median EAT-10 scores were 0.0 (IQR 0.0–6.0) in the PCF group and 0.0 (IQR 0.0–1.0) in the ACD group.

Secondary outcome: the Glasgow–Edinburgh Throat Scale

Summaries of the GETS scores at baseline, at day 1 and at 6, 12, 26, 39 and 52 weeks post operation are presented in Figure 9. In general, a lower GETS score is better, that is it corresponds to less severe globus symptoms. At all time points post operation, the median GETS scores are lower in the PCF group, although the IQRs overlap at all time points except day 1. There is an observed spike in median GETS score at day 1 post operation for the ACD group.

FIGURE 9.

Total GETS score at each time point, by treatment group.

Summary statistics calculated for each time point, including medians and IQRs, can be found in Appendix 5, Table 33, as can summaries of additional questionnaire items that did not contribute to the overall score. At 52 weeks, the median GETS scores were 0.0 (IQR 0.0–8.0) in the PCF group and 2.5 (IQR 2.0–5.0) in the ACD group.

Secondary outcome: grade, roughness, breathiness, asthenia and strain assessment

Owing to the small sample size and concerns about identifiability in this subgroup, results for this outcome are presented overall, and not by treatment group.

Six participants were selected at registration to provide additional voice recordings. In total, eight voice recordings from five participants were collected during the trial, five at baseline and three at 6 weeks post operation. All collected recordings were considered by the central reviewer to be of adequate quality to review, and all participants performed the sentence production and at least one of the spontaneous speech or alternative reading tasks (see Appendix 2 for details of voice tasks). There was one participant who partially completed the vowel sound task at baseline, and who did not complete the same task at 6 weeks.

Parameters are scored from 0 (‘normal’) to 3 (‘severe’). Higher scores indicate a more pronounced vocal problem for the parameter in question. Across all voice recordings, no participants were judged in any parameter to experience more than a mild degree of issues, and this did not change between baseline and 6 weeks post operation. Summaries of score by parameter are presented in Appendix 5, Table 38.

Secondary outcome: the Voice Handicap Index-10 items

Summaries of the VHI-10 total scores at baseline, at day 1 and at 6, 12, 26, 39 and 52 weeks post operation are presented in Figure 10. In general, a lower VHI-10 score is better, that is it corresponds to less severe vocal problems. At all postoperative time points, the median VHI-10 scores are low or zero across both treatment groups; variability in scores appears to increase over time. At day 1 post operation, the ACD group appears to have a worse (higher) VHI-10 score, although there is no clear difference between groups thereafter.

FIGURE 10.

The VHI-10 total scores at each time point, by treatment group.

Summary statistics calculated for each time point, including medians and IQRs, can be found in Appendix 5, Table 35. At 52 weeks, the median VHI-10 scores were 0.0 (IQR 0.0–16.5) in the PCF group and 0.0 (IQR 0.0–7.0) in the ACD group.

Secondary outcome: Restricted American Spinal Injury Association score

Across all summarised outcomes [the total motor score for the upper limb on the operated side, the total light-touch sensory score for the upper limb on the operated side, the total pinprick sensory score for the upper limb on the operated side, the total motor score (Figure 11), the total light-touch sensory score (Figure 12) and the total pinprick sensory score (Figure 13)], scores remained relatively unchanged and were similar across all time points (i.e. baseline, day 1 and 6 weeks post operation) and both treatment groups. In general, a higher ASIA score for each of these outcomes is better, that is it corresponds to a higher level of motor or sensory function. Additional figures and tables are presented in Appendix 5, Figures 16–18 and Table 37.

FIGURE 11.

Total motor score at each time point, by treatment group.

FIGURE 12.

Total light-touch sensory score at each time point, by treatment group.

FIGURE 13.

Total pinprick sensory score at each time point, by treatment group.

Secondary outcome: incidence of revision surgery within 52 weeks postoperatively

There were no reported reoperations occurring within the trial.

Secondary outcome: impact of variations in the optional surgical components of posterior cervical foraminotomy

All participants in the trial received PCF using an open technique.

Secondary outcome: impact of variations in the optional surgical components of anterior cervical discectomy

The majority of participants receiving ACD in the trial received ACD without a plate (n = 8, 88.9%). The remaining participant received a plate.

Incidence of intraoperative complications

No intraoperative complications were reported during the trial.

Incidence of postoperative complications within 6 weeks of surgery

Five postoperative complications occurring within 6 weeks of surgery were reported in the trial, all in the ACD group. No serious or unexpected serious complications were reported. Line listings of postoperative complications are presented in Table 9. All listed complications presented in different participants.

Deaths

No deaths were reported in the trial.

Revision to economic analysis plan

• Costs and QALYs will no longer be combined into a summary cost-effectiveness ratio measure. Instead, mean quantities of resource use, mean costs and mean EQ-5D-3L utility scores will be reported for each trial arm, including measures of sample variation, but without estimation or statistical testing for differences.

• The number of missing observations for each item of health-care resource use and of the EQ-5D-3L questionnaire will be presented.

• Imputation of missing data will no longer be conducted.

• Regression analysis will no longer be conducted.

• Uncertainty analysis and both deterministic (one way and multiway) and probabilistic (bootstrapping of mean costs and QALYs) sensitivity analyses will no longer be conducted.

Data collection for the health economics component

Health-related quality-of-life data and health resource use questionnaire data were collected as shown in Table 10.

Intervention and comparator surgery and initial hospitalisation costs

The costs of resources used during the surgical procedures were collected for each trial participant in the trial case report forms;61 these included the type of implant used (cage or plate) for ACD; time in theatre from skin incision to closure for ACD and PCF; the grade of surgeons and anaesthetists involved in the operations; and number of days in the inpatient ward, high-dependency unit and/or intensive care unit.

Information on unit costs of resources involving surgical procedures was obtained from published sources, including Hutton. 55 The hourly cost of theatre (Information Services Division Scotland 2012), surgeon56 and anaesthetist salaries,57 and the mean cost per day spent in an inpatient ward and high-dependency and intensive care units58 were also obtained. The cost of theatre is from Scotland, which was the only identified source of data on costs in time units available for the UK. The cost of medications was obtained from the British National Formulary59 for typical regimens prescribed to patients according to the opinion of clinical experts in the research team. Costs were reflated to 2020 prices using published price indices56,60 as needed. Table 11 presents further details.

Subsequent health-care resource use

Health-related quality of life

The EQ-5D measures the patient HRQoL for obtaining health states and deriving preference-based assessments or utility scores for health states. Utility scores are typically used to estimate QALYs. The EQ-5D-3L questionnaire was given to families in paper format and utilities were derived using the UK tariff values. 72

Analysis

We estimated the costs of intervention, health-care resource use and productivity losses using the unit costs presented in Tables 11 and 12. Out-of-pocket costs were presented as reported by participants in the trial.

Average health-care resource use was estimated for each treatment arm. The frequency of contact for each health service was tabulated. The mean and SD and median and IQR for the frequencies and costs of health-care resource use elements, productivity costs and EQ-5D-3L dimensions and scores in each trial arm were calculated at baseline, at day 1 and at 6, 12, 26, 39 and 52 weeks post surgery. For the EQ-5D-3L, the differences in the mean score from baseline at each follow-up point are also presented for each trial arm.

Health-care resource use data were combined with unit costs to calculate health-care service use costs. Recorded use of medications is described, but, in the absence of information on regimens administered, these resources were not costed. The analysis is based on complete-case data. When aggregating the costs of intervention and health care, no adjustment was made for patients who were missing data. Uncertainty around the means of EQ-5D scores and health-care and productivity costs are described using 95% CIs for each treatment arm.

We present per-patient cost estimates from an NHS and Personal Social Services perspective, and, separately, from a societal perspective. The societal perspective includes, in addition to the costs included in the NHS and Personal Social Services (i.e. health-care costs of providing the initial surgical intervention and downstream health-care and Personal Social Services costs post initial hospital discharge), the costs borne and reported by the participant (i.e. out-of-pocket costs) and the productivity costs of time taken off work as a result of their surgery and its complications.

Because all costs and utilities occur within a 12-month period, no discounting is applied to costs. Results are presented in Great British pounds at 2019–20 prices.

Completeness of health economics data

Questionnaire response rates were calculated for each item of health-care resource use and EQ-5D classification system. The number of returned questionnaires was recorded and the response rate percentage was calculated relative to a maximum response rate of 14 participants for the PCF arm and nine participants for the ACD arm, at baseline and each follow-up point.

We present costs for all those participants with available data for each cost item, whereas, when we aggregate into total costs for each category of resource (i.e. intervention, downstream health-care costs, out-of-pocket costs and productivity costs), only participants with complete data for all items in the respective category are included (complete-case analysis). As high attrition after the 6-week follow-up resulted in few observations with complete cost data across the full 52-week period of follow-up, we have presented total costs from the NHS and Personal Social Services perspective and, separately, total costs from the societal perspective up to the 6-week follow-up point, in a complete-case analysis.

Surgical intervention costs

The mean duration of the operation was 61 (SD 24) and 90 (SD 24) minutes in the PCF and ACD arms, respectively (Table 13). The mean duration of anaesthesia administration was 98 and 124 minutes for the PCF and ACD arms, respectively. Principal surgeons operating on participants in the trial were assisted by a surgical registrar and nursing staff. Principal anaesthetists were assisted by middle-grade anaesthetists and an operating department professional. No blood transfusions were recorded, and the operation involved two sets of radiographs in the PCF arm and one set in the ACD arm for the median participant.

Perioperative antibiotic prophylaxis was administered to 93% (13/14) of participants in the PCF arm and 88% (8/9) in the ACD arm. Combination therapy with a penicillin and an aminoglycoside was the most common antibiotic regimen used: 69% (9/13) in the PCF arm and 75% (6/8) in the ACD arm. Two participants in the PCF arm and one participant in the ACD arm were given an aminoglycoside with a glycopeptide; cephalosporin was given alone (one participant in the PCF arm), in combination with penicillin (one participant in the PCF arm) or with glycopeptides (one participant in the ACD arm). Further details on the devices used and tests performed are presented in Table 14.

Among those who were recorded to have received analgesics on day 1 postoperatively, one participant in the ACD arm had missing information on the type of analgesic administered. All participants with available information received paracetamol (12 participants in the PCF arm and nine in the ACD arm). Two-thirds of PCF participants and 87.5% of ACD participants received additional medications: weak opioids (three and one participants in the PCF and ACD arms, respectively), weak opioids and neuromodulating agents (NMAs) (two participants in each arm), both weak and strong opioids with (one participant in each arm) and without NMAs (one participant in each arm), weak and strong opioids with non-steroidal anti-inflammatory drugs (NSAIDs) (one participant in the PCF arm) and strong opioids with (one ACD participant) and without (one ACD participant) NMAs.

At 6 weeks’ follow-up, eight (66.6%) and seven (77%) participants in the PCF and ACD arms, respectively, had a recorded medication, with one instance in each arm being the use of an antibiotic (Table 15). The other participants used medication for pain, which most often included drug combinations involving paracetamol and weak opioids, either without an additional analgesic (two ACD participants) or with the addition of strong opioids, NMAs and NSAIDs (one PCF participant); strong opioids (one PCF participant); NMAs and NSAIDs (two PCF participants); or NMAs (two ACD participants). The remaining participants used a single drug: paracetamol (one PCF participant), strong opioids (two participants, one in each arm), NSAIDs (one PCF participant) or NMAs (one ACD participant).

The mean length of hospital stay was 1.2 (SD 0.3) and 1.3 (SD 1) days in the PCF and ACD groups, respectively; the IQR was fixed at 1 day in both groups. The total mean cost of the initial surgical therapy was £4294 (SD £631, 95% CI £2344 to £3147) per participant in the PCF arm and £4295 (SD £1028, 95% CI £3436 to £5154) per participant in the ACD arm (Table 16). In contrast, the median total costs were £2622 (IQR £2402–2824) and £4423 (IQR £3849–4821) in the PCF and ACD arms, respectively, which were dominated by the costs of theatre. It is noted that one participant in each arm was missing information on the duration of the operation, and therefore was missing data on the cost of surgical staff input and theatre use; two participants in the PCF arm were missing data on date of discharge from hospital and on costs of hospital stay.

Health-care and social care resource use, and out-of-pocket and productivity costs

At week 6, the mean costs in the ACD arm were driven by a single participant with a recorded attendance at A&E on seven separate occasions; this participant was missing data on some of the other costs elements (ward or outpatient attendances and telephone support), and was therefore excluded in the aggregate total health-care cost variable. As a result, the total health-care costs were more than halved relative to what they would have been had the participant been included with zero values imputed for the missing cost data. However, most use of health-care services related to GP visits and contacts with the nurse at the general practice (see Appendix 6, Table 41). Apart from these two service categories, the majority of participants incurred no resource utilisation, resulting in zero costs for the median participant. The most prescribed medications were analgesics, reported by nine and four participants in the PCF and ACD arms, respectively, and antibiotics were prescribed for one and two participants in the PCF and ACD arms, respectively (see Appendix 6, Table 40). In the ACD arm, one participant recorded receipt of an analgesic and antihistamine drugs, and another recorded receipt of an antibiotic and codeine. Of the six participants who reported having paid for medications, only three provided the amount they spent on them: one participant in the PCF group spent £12; in the ACD group, one participant reported spending £5.50 and another spent £0.70. Most of these expenses were for painkillers (data not presented in Table 17). Table 17 presents further details.

When the number of days off work as a result of surgery or its complications were evaluated at the gross hourly wage in the UK according to whether the participant worked part time or full time and their age, the mean productivity costs during the first 6 weeks after surgery amounted to £1942.75 (SD £2320, 95% CI £283 to £3602) per participant in the PCF arm and £1520.51 (SD £1596, 95% CI £44 to £2997) per participant in the ACD arm (see Table 17). Since the mean figure for PCF was affected by outliers, the median productivity costs are also presented in Table 17 and are £708.29 (IQR £0–4250) in the PCF arm and £1591.20 (IQR £0–3007) in the ACD arm.

Most use of health-care services during the period was between weeks 6 and 12 post surgery and was related to GP visits (see Appendix 6, Table 42). Apart from these service categories, the majority of participants incurred no resource utilisation, resulting in zero costs for the median participant. Owing to incomplete data, the aggregate total health-care cost variable was available for only nine out of 14 participants in the PCF arm and three out of nine participants in the ACD arm, and had mean values of £115 (SD £96, 95% CI £41 to £189) and £125 (SD £71, 95% CI –£52 to £303) for the PCF and ACD arms, respectively (Table 18).

At week 12, the most prescribed medications were analgesics, reported by nine and two participants in the PCF and ACD arms, respectively, and antibiotics were prescribed to three and two participants in the PCF and ACD arms, respectively. One participant in each trial arm also reported being prescribed dressings. In the PCF arm, one participant recorded receipt of a mucolytic and in the ACD arm one participant reported use of ‘tranexamic acid for periods’. The five participants who reported having paid for medications provided the amount they spent for them: four participants in the PCF group spent a mean of £9.65 and one participant in the ACD group reported spending £1.40. All of these expenses were for analgesics, except for one participant in the PCF arm who reported use of antidepressants.

The number of days off work as a result of surgery or its complications occurring between weeks 6 and 12 was evaluated at the median gross hourly wage in the UK corresponding to the age of the individual participant and the type of work, part or full time, reported at baseline. The resulting mean productivity costs amounted to £991.60 (SD £2675, 95% CI –£992 to £2905) and £2227.27 (SD £4719, 95% CI –£2725 to £7180) per participant in the PCF and ACD groups, respectively (see Table 18). However, most participants had zero productivity costs, as only two participants in each arm out of the 10 and six in the PCF and ACD arms, respectively, with available data had incurred any costs.

Between weeks 13 and 26, the mean cost of health-care service use was £96 in the PCF arm and £284 in the ACD arm. These resulted primarily from four and five ward or outpatient attendances in the PCF and the ACD arm, respectively, two visits to A&E in the ACD arm and at least one GP visit for the majority of the participants in both arms (see Appendix 6, Table 43). The rate of missing data on health-care services used was 21% (3/14) in the PCF arm and 44% (4/9) in the ACD arm.

Taking time off work as a result of surgery or its consequences between postoperative weeks 13 and 26 was reported by two participants in the PCF arm and one participant in the ACD arm (Table 19). This information was not available for 21% (3/14) and 22% (2/9) of participants enrolled in the PCF and ACD arms, respectively. Data on the number of days of work lost for those who took time off work were missing for one participant in each arm. The one participant with available data was in the PCF arm, and reported 2 days of work lost as a consequence of the surgery or its complications over the previous 3 months.

Between weeks 27 and 39, the mean costs of health-care service use were £163 in the PCF arm and £144 in the ACD arm. These resulted mainly from four ward or outpatient attendances in each arm, two visits to A&E and eight sessions with a physiotherapist in the ACD arm, and one or two GP visits for the majority of the participants in both arms (see Appendix 6, Table 44). The rate of missing data on health-care services used was 50% (7/14) in the PCF arm and 22% (2/9) in the ACD arm.

Taking time off work because of the surgery or its consequences between postoperative weeks 27 and 39 was reported by two participants in the PCF arm and one participant in the ACD arm (Table 20). This information was not available for 42% (6/14) and 22% (2/9) of participants enrolled in the PCF and ACD arms, respectively. Data on the number of days of work lost for those who took time off work were missing for one participant in the PCF arm. In each arm, one participant with available data reported 1 day of work lost as a consequence of the surgery or its complications over the previous 3 months.

Between weeks 40 and 52, the mean cost of health-care service use was £92 in the PCF arm and £1290 in the ACD arm. Similarly to previous follow-up time points, the majority of participants had between one and two GP visits, and there were three visits to A&E in the PCF arm and three outpatient or ward attendances in the ACD arm. However, the mean costs in the ACD arm were influenced by a single participant who reported 95 contacts with a physiotherapist (see Appendix 6, Table 45). The rate of missing data on health-care services used was 29% (4/14) in the PCF arm and 44% (4/9) in the ACD arm.

No participant in either arm reported taking time off work because of surgery or its consequences between postoperative weeks 40 and 52 (Table 21). This information was not available for 21% (3/14) and 33% (3/9) of participants enrolled in the PCF and ACD arms, respectively.

In Table 22 we summarise the results of the costs analysis. There is high sensitivity of the results to attrition in the data and outliers because of the low numbers, most evident in the lack of complete data from any participant in the ACD arm to be able to calculate the total cumulative costs of health care at 52 weeks after the initial surgery. Nevertheless, the median costs up to week 12 point to productivity costs that decline to zero after week 6, and represent 26% and 37% of the costs of the initial operation in the PCF and ACD arms, respectively. The median health-care costs are of much smaller orders of magnitude and the median out-of-pocket costs are zero in both arms. Furthermore, although positive median costs were observed in both arms during weeks 1–6, the median total productivity costs for each group were zero because the three cases in each arm who were lost to subsequent follow-up were among those with positive costs in weeks 1–6.

Estimating the total costs over the 52-week follow-up duration of the trial was not feasible because of high levels of attrition after week 6. Therefore, costs were aggregated up to week 6 only. The mean costs from the NHS and Personal Social Services perspective were £2716 (95% CI £2345 to £3087; missing 6 out of 14) in the PCF arm and £4133 (95% CI £3099 to £5167; missing 3 out of 9) in the ACD arm. The median costs were £2634 (IQR £2444–2741) and £4214 (IQR £3602–4994) in the PCF and ACD arms, respectively. Combining out-of-pocket costs, productivity losses, and NHS and Personal Social Services costs produced mean per-participant total costs to society of £4608 (95% CI £2514 to £6703; missing 7 out of 14) in the PCF arm and £5015 (95% CI £2286 to £7743; missing 4 out of 9) in the ACD arm; the median costs were £4097 (IQR £2448–6591) and £4143 (IQR £4126–4284) in the PCF and ACD arms, respectively.

Health-related quality of life

On day 0 pre surgery, the baseline EQ-5D-3L scores were 0.291 (SD 0.34, 95% CI 0.07 to 0.51) in the PCF arm and 0.595 (SD 0.25, 95% CI 0.40 to 0.78) in the ACD arm. The corresponding median scores were 0.210 (IQR –0.01 to 0.60) in the PCF arm and 0.689 (IQR 0.66 to 0.69) in the ACD arm. Median changes from baseline at day 1 and at weeks 6, 12, 26, 39 and 52 were, respectively, 0.06, 0.10, 0.00, 0.00, 0.29 and 0.06 in the PCF arm and 0.00, 0.02, 0.00, 0.00, –0.03 and –0.09 in the ACD arm. The rate of missing EQ-5D-3L score data varies from a lowest rate of 14% (2/14) at baseline to a highest rate of 64% (9/14) at 39 weeks in the PCF arm and from 0% (0/9) at baseline to 33.3% (3/9) at weeks 12 and 39 in the ACD arm; no imputation was applied in the analysis. Figure 14 illustrates the variation in EQ-5D-3L scores at each follow-up point for each treatment, and a summary table of results is presented in Appendix 6, Table 39.

FIGURE 14.

The EQ-5D-3L scores at each time point, by treatment group.

Appendix 6, Table 46 presents the number of participants reporting the level of problems in each of the five EQ-5D dimensions. The distribution of frequencies of reported problems across the two arms is illustrated in Appendix 6, Figure 19, for baseline, day 1 and week 6.

Recruitment of sites

Three groups of NHS sites were approached for participation; we aimed to include some sites from each group:

1. those that took part in the FORVAD trial and recruited patients to the trial

2. those that took part in the FORVAD trial but did not recruit any patients to the trial

3. those that were approached about the FORVAD trial but did not open to recruitment.

For the qualitative study, all sites that were open and all sites that had returned a completed feasibility form and were in the process of set-up were approached to take part (11 open sites and six in set-up). In total, 17 NHS sites were approached to take part in the FORVAD qualitative study. Of these, two sites were not able to confirm continued capacity and capability for the study amendment and so declined to take part in the study. Seven sites did not respond to recruitment e-mails or follow-up recruitment e-mails. Eight sites responded to the recruitment e-mails and took part in the qualitative interview phase of the study.

Recruitment of health-care professionals

Participants were identified from authorised personnel (delegation) logs for sites that were willing to take part in the qualitative substudy. All staff identified as working on the trial were invited to participate in an interview. Initial e-mail invitations were sent out to each member of staff along with a copy of the FORVAD qualitative study participant information sheet. A follow-up e-mail was sent after 3–5 days if there had not been a response from the initial e-mail invitation.

All staff who responded to the e-mail invitation were offered a Microsoft Teams interview at a time and date that was convenient to them. A copy of the verbal consent checklist was sent by e-mail prior to the interview. At the start of the Microsoft Teams call, participants were given an opportunity to ask any questions about the study. Consent was recorded verbally at the start of the interview.

Recruitment of patients

Owing to the limited time available to conduct the qualitative study during trial close-down, recruitment for the patient interviews focused on the site with the most participants (18 out of a total of 23 randomised patients), as all other sites had recruited only a very small number of patients. All trial participants from this site were sent an invitation letter and participant information sheet for the qualitative study, along with an expression of interest (EOI) form to return if interested in taking part in the study. Reminder letters were sent out approximately 2 weeks after the initial invitations to anyone who had not returned the EOI form by this point.

Once an EOI form was received, an e-mail was sent to the participant to arrange a date and time for a Microsoft Teams interview. If a response to this e-mail was not received, a reminder was sent 1 week later; if no response to the reminder was received, an attempt was made to contact the participant by telephone. A copy of the verbal consent checklist was sent by e-mail prior to the interview. At the start of the Microsoft Teams call, participants were given an opportunity to ask any questions about the study. Consent was recorded verbally at the start of the interview.

Owing to the limited time available for the qualitative study, the patient interviews were conducted by junior doctors working at the trial site, as they did not require a research passport to conduct the interviews. These junior doctors received training in qualitative interviewing from the FORVAD qualitative research team, as well as ongoing support throughout the fieldwork period. However, it proved challenging to organise training and patient interviews around busy junior doctor schedules, which resulted in a small number of patient interviews being conducted.

Participants

Eighteen semistructured interviews were conducted with FORVAD site staff: four with a site principal investigator (PI), five with other surgeons and nine with research nurses/study co-ordinators (Table 23). Six patients responded to the initial e-mail contact; two participated in an interview. Interviews lasted between 12 and 55 minutes. In total, 12 male and 6 female health-care professionals were interviewed. Both of the patients were female, aged between 50 and 59 years and of white British ethnicity and had received PCF at randomisation. All interviews took place between March and May 2021.

In the findings that follow, all participants interviewed have been given pseudonyms and all research sites that took part in the study have been assigned a letter (e.g. site A) to maintain anonymity. The gender implied by the pseudonym does not necessarily match that of the person who took part in the study. Quotation descriptors provide job roles only to help protect the identity of the participant.

Equipoise in the FORVAD trial and in neurosurgical trials

The theme of equipoise in the FORVAD trial and in neurosurgical trials includes three subthemes: (1) theory versus applied practice, (2) preferences for surgical approach and (3) considering the patient.

Organisation and implementation

The FORVAD trial required considerable collective organisation in both its initial set-up and its subsequent implementation.

Integration of the clinical and recruitment pathways

This section will focus on the theme of the integration of the clinical and recruitment pathways, which comprises three subthemes: (1) support needs, (2) timing of randomisation and (3) identification and screening of patients.

Coherence (meaning and sense-making by participants)

Overall, there appears to have been widespread support for the aims of the FORVAD trial, and surgeons felt that the research question was important and that the results would have influenced clinical practice. The differences between the two procedures were well understood and surgeons discussed the relative risks and benefits of each, identifying certain characteristics or circumstances that would make one better than the other for specific patients. However, participating surgeons also felt that there were some patients for whom either procedure could be appropriate, and that there is a lack of evidence as to which procedure is better in these circumstances. The trial eligibility criteria were generally felt to be appropriate and identified patients for whom it was unclear which procedure would be best (i.e. when there is collective clinical equipoise).

There does not appear to have been any other significant issues with the FORVAD trial in terms of communal or individual specification. Site staff did not report any concerns around information provision, training or understanding of the trial aims or benefits, and generally both surgeons and research support staff appeared to be clear of their roles and how they fit in with the trial overall. However, one issue that did appear to present a significant challenge in terms of coherence was the timing of randomisation, which conflicted with a general move towards earlier surgical consent because of ethics and legal concerns.

Randomising on the day of surgery also challenged surgeons’ clinical relationships with patients by removing the opportunity to discuss the specific procedure the patient would have at their pre-operative outpatient clinics. Although none of the surgeons interviewed specifically identified this as a key factor in their recruitment problems, it clearly limited their ability to make sense of the trial in the context of their legal and clinical obligations. The potential impact on surgeons’ willingness to recruit and enthusiasm for the trial should therefore not be discounted, particularly for those surgeons who may have been less engaged to begin with (and who are likely to be unrepresented or under-represented in the interview data).

Cognitive participation (commitment and engagement by participants)

Commitment and engagement emerged from the thematic analysis as a crucial factor in sites’ ability to set up and recruit successfully to the FORVAD trial. This is not to say, however, that commitment and engagement was necessarily lacking at sites that struggled with recruitment, as all interviewees described being enthusiastic about the trial and keen to take part. Rather, it appears that the considerable logistical challenges of delivering a surgical trial could (in some cases) be overcome by an elevated and sustained level of commitment and engagement. Conversely, these challenges appear to have had the effect of dampening enthusiasm at some sites, and it may be that being moderately committed and engaged with the trial initially was not sufficient to maintain the level of motivation required to push through the numerous barriers to operationalising the trial in practice. The importance of having an engaged, experienced and motivated site team that was prepared to ‘go the extra mile’ to deliver the trial clearly cannot be underestimated. However, the experience of one interviewee struggling to access research nurse support despite working in a large, research-intensive hospital highlights the fact that this is not something that can always be relied on for surgical trials, and this issue has been reflected in previous literature. 116 Similarly, the role of a specialty research nurse, working specifically in clinical trials, is considered a fundamental asset of successful trial management and delivery. 117 When designing future trials, therefore, careful consideration should be given to making the trial as easy to set up and deliver as possible [taking into account the specific implementation challenges discussed under Collective action (i.e. the work participants do to make the trial function)], so that site staff can focus their limited time and energy on recruitment. It may also be beneficial to consider whether or not a model of fewer sites with more dedicated support provided to deliver the trial might be more successful in some circumstances (e.g. where there are sufficient patient numbers and the limiting factor is likely to be research capacity at sites). The benefits of this approach, however, would have to be weighed up against the risk of some of these sites still failing to recruit, leaving fewer sites overall to compensate for any shortfall in recruitment numbers.

Another important factor to consider in terms of cognitive participation is legitimisation, and in particular whether or not there is sufficient equipoise among clinical staff at sites. Some hospitals and/or surgeons may not have taken part in the FORVAD trial because of strong preferences for one of the trial interventions over the other. However, it is unlikely that this was the key factor limiting recruitment, as sites that recruited successfully did so with only a small of number of surgeons delivering the trial interventions, and the total number of sites that opened or were in set-up should have been sufficient if those sites had recruited at the same rate as the more successful sites. Furthermore, in line with their general support for the aims of the trial and agreement that there is collective clinical equipoise, the surgeons interviewed largely appeared to have sufficient individual equipoise to deliver the trial successfully. Likewise, in contrast to research on other surgical trials that found equipoise to be the key factor limiting recruitment,118 the research nurses and other research support staff interviewed for the FORVAD qualitative study did not report any significant issues with surgeon preference that affected their ability to communicate the trial to patients, yet they felt that, at times, they had to prompt the surgeon to look for potential patients during consultations as this was overlooked. When surgeons did talk to patients about the FORVAD trial, it appears that they were able to communicate equipoise successfully enough for patients to consent to taking part. The key limiting factor appears not to be the ability of the team to communicate equipoise in a trial recruitment conversation, but rather appears to be the timely identification of eligible patients so that the recruitment conversation could take place. There was some suggestion, however, that surgeons who preferred one procedure over the other might have been less open to considering specific patients for the trial, and been less able to convey equipoise to their patients during consent discussions.

Overall, this suggests that, although equipoise was certainly a factor in the FORVAD trial recruitment difficulties, there is no strong evidence that this was more of a problem than would be expected for a surgical trial of this type, and it is unlikely to have been the only (or even most significant) factor that limited recruitment. Nevertheless, adopting an expertise-based design for future trials in the area could potentially help to address some of the issues caused by a lack of individual surgeon equipoise. 119 It should be noted, however, that this design would not have addressed the issue of entire sites being unwilling to participate because they conduct only one of the interventions, as the standard expertise-based design requires both procedures to be delivered at all sites. An expertise-based design would also not necessarily improve any issues caused by surgeons struggling to convey equipoise to patients during consent discussions, and could, in fact, exacerbate these issues as surgeons are less likely to be able to talk confidently about the risks and benefits of both procedures if they are experienced in delivering only one. It would, however, allow surgeons who are in equipoise overall but do not feel personally able to deliver both of the interventions to participate, increasing the potential pool of surgeons at each site, and it could potentially make surgeons more willing to put patients forward for the trial. These benefits, however, would need to be weighed up against the additional logistical challenges introduced by an expertise-based design,120 particularly in the context of the interactional workability challenges that sites experienced with the FORVAD trial (discussed in more detail in the following section).

Collective action (i.e. the work participants do to make the trial function)

The thematic analysis and, in particular, the issues explored under the Integration of the clinical and recruitment pathways theme, suggests that interactional workability (how the trial was operationalised and its compatibility with existing work practices) is likely to have been one of the most important factors affecting recruitment to the FORVAD trial. The process for screening and approaching patients, the timing of randomisation, fitting follow-up visits into standard care pathways, the length of waiting lists, and the availability of theatre facilities and/or use of private facilities were all important aspects of operationalising the trial. Importantly, sites that recruited successfully were either not affected by these issues (e.g. waiting lists were relatively short because the site had access to theatres dedicated to minor procedures, or follow-up visits happened to fall within their usual clinical pathway) or described putting processes in place that overcame them (identifying potentially eligible patients from referral letters and triaging these patients to dedicated clinics run by participating trial surgeons). Conversely, sites that recruited less well described having issues with some or all of these activities. For example, one site experienced a significant delay in opening because of concerns about randomisation on the day of surgery, and another interviewee described arranging follow-up visits that were outside the usual care pathway taking up time that could have been spent screening. Even in cases in which interviewees felt that their recruitment problems were predominantly due to not seeing enough eligible patients, it was apparent that the site may in fact have simply failed to identify those patients because of the lack of a structured process for screening referrals. Although none of these individual issues around interactional workability would necessarily have made recruitment impossible, the cumulative effect of a number of issues combined is likely to have contributed to the low levels of recruitment seen at most sites.

In terms of skill set workability and contextual and relational integration, the thematic analysis highlighted the importance of site clinical teams and research support working together effectively, and, in particular, the crucial role played by research nurses. In addition to the importance of having surgeons with individual equipoise and an experienced and motivated team that is willing to ‘go the extra mile’ (as discussed previously), learning from each other and establishing support networks appeared to be particularly valuable for many interviewees. However, as this was reported by sites that failed to recruit as well as those that did, it is unlikely to have played a significant or direct role in recruitment, although it may have been an underlying contributory factor in maintaining the enthusiasm and motivation that has been identified as important for successful recruitment.

Reflexive monitoring (i.e. participants reflect on or appraise the trial)

The potential for reflexive monitoring of the FORVAD trial was limited to some extent by the circumstances in which the trial closed. In particular, three interlinked factors are likely to have affected participants’ ability to reflect on and appraise the trial: (1) the short length of time many sites were open limited their opportunity to reflect on the trial processes; (2) the impact of COVID-19 in the last few months of the trial meant that clinical services were severely disrupted, making it difficult to reflect on the trial itself; and (3) the early closure of recruitment means that the research question has not been answered, so participants are not able to reflect meaningfully on the outcome of the trial and the potential for changing clinical practice. Notwithstanding this, however, the thematic analysis does provide some indication of reflexive monitoring, and suggests some tentative conclusions that may be of value for future research in this area.

The surgeons interviewed were generally still supportive of the aims of the trial and felt that this is still an important clinical question to be answered, which suggests that communal appraisal of the value of learning more about ACD and PCF would not be a barrier to future research in this area. In terms of individual appraisal, and in particular the impact of the trial on individuals and their work environment, there were some concerns about the amount of time the trial required and how difficult it was for surgeons to fit this into an already full clinical schedule. This aligns with the previous discussion about the importance of having an engaged team and research support infrastructure to enable surgeons to participate effectively in trials. It does not appear, however, that the FORVAD trial was uniquely challenging in this regard and although there were some aspects of the protocol that created additional work, this does not appear to have been a significant factor that limited recruitment. It is likely, therefore, that this concern is largely applicable to trials in surgery in general, rather than being a specific issue for the FORVAD trial. It does suggest, however, that it is important to take into account the context in which site teams are working when designing surgical trials, and to ensure that site staff can focus their limited resources on screening and recruitment as much as possible. In particular, assessments and data collection should be kept to the minimum required to answer the research question and visits should be aligned as far as possible to standard care pathways to minimise the additional work required at sites.

In line with this, the thematic analysis suggests that sites had only limited ability for reconfiguration and adapting the trial procedures on the basis of their experience. Although there were some positive examples of this, such as one site setting up a follow-up visit tracker to prompt clinic visits to be arranged in the correct window after a visit had been missed, the limited flexibility allowed for in a trial protocol generally made it hard for sites to adapt trial processes to their local clinical pathways. This further increased the reliance on team members being willing to go above and beyond their usual roles, for instance research nurses coming in on the weekend to ensure that post-surgery follow-up assessments were completed on time. Again, these issues will not necessarily have directly affected recruitment rates, but will potentially have limited the time and energy available for screening and recruitment. The nature of RCTs mean that some inflexibility is inevitable, as there needs to be some consistency across sites. However, consideration should be given when designing future trials to identifying aspects of the protocol for which flexibility can be allowed without significantly affecting scientific validity, so that sites can adapt the trial to their working practices as much as possible. In addition, as one interviewee suggested, engagement with as many clinical sites as possible during the protocol design would help to ensure that the trial pathway fits in with as many variations of local clinical pathways as possible. Again, there will be a limit to how far it is practical to accommodate this recommendation. Consultation with sites takes time, and will therefore need to be balanced against the need to get the trial set up and open. Our results do suggest, however, that allocating at least some time and resources to consulting with sites earlier in the set-up process would be beneficial.

In terms of individual appraisal from the patient perspective, there are some indications that patients particularly valued being able to have their surgery as quickly as possible. The actual or perceived impact of trial participation on waiting times (whether positive or negative) is therefore likely to be an important factor in determining how appealing future trials in this area will be to patients. As discussed previously, this study was not designed to provide a comprehensive understanding of patient perspectives on the FORVAD trial, so the results cannot provide any conclusive findings on this issue. However, there are indications that this could be an important area for future research, particularly in the context of the ongoing disruption to elective services as a result of COVID-19.

Qualitative study strengths and limitations

A strength of the qualitative study was that we recruited participants from both the key professional groups involved in the FORVAD trial at sites: research nurses and surgeons. At most sites we were able to speak to at least one individual from each professional role, facilitating a comprehensive understanding of the implementation of the FORVAD trial at that site. We were successful in recruiting participants from the three groups of sites that we identified: those that had opened to recruitment and had successfully recruited patients to the FORVAD trial, those that had opened to recruitment but had not recruited and those that were still in set-up at the point that the FORVAD trial closed. A limitation was that we conducted the research at the end of the FORVAD trial, so some participants may have remembered less about challenges encountered earlier in the trial. In particular, this limited the extent to which we were able to ‘unpick’ the different factors contributing to recruitment challenges: ethnographic work at sites during recruitment might have enabled us to better understand the complex interaction between, for example, surgical preferences and trial organisation. However, sites were at varying stages in set-up and implementation and we were able to use the accumulated knowledge about the trial of other TMG members to help us to direct our recruitment and data collection effectively. We were less successful at recruiting patients, and were able to speak only to patients who had agreed to participate in the FORVAD study; any future trial in this area would benefit from including strong PPI and/or a qualitative feasibility component with patients to identify and address potential patient barriers to recruitment.

Neck Disability Index

The primary outcome measure of the FORVAD trial was NDI score at 52 weeks after surgery. In the PCF group, the median NDI score was 44.0 (IQR 36.0–62.0) pre-operatively and 25.3 (IQR 20.0–42.0) at 52 weeks. In the ACD group, the median NDI score was 35.6 (IQR 34.0–44.0) pre operatively and 45.0 (IQR 20.0–57.0) at 52 weeks, meaning that the PCF group had a worse pre-operative score and improved by 52 weeks, whereas the ACD group had a better pre-operative score and did not improve by 52 weeks. However, there is overlap in the IQRs, and the low sample size may mean that there are influential observations that influence the median. The minimum clinically important difference in NDI score is 10%, but the number of participants was too small to conclude that there is a significant difference in outcome between the two groups.

In the PCF group, the median NDI score deteriorated from 44.0 (IQR 36.0–62.0) at baseline to 61.3 (IQR 41.7–69.4) at day 1, and then sequentially improved to 45.0 (IQR 30.0–48.9) at 6 weeks, 40.0 (IQR 20.0–54.0) at 12 weeks, 42.2 (IQR 22.0–68.0) at 26 weeks, 24.0 (IQR 20.0–42.2) at 39 weeks, and 25.3 (IQR 20.0–42.0) at 52 weeks. This is in contrast to the ACD group, of which the median NDI score was 35.6 (IQR 34.0–44.0) at baseline, deteriorating to 46.0 (IQR 36.0–64.4) at day 1 then improving to the best (lowest) median score of 14.0 (IQR 8.0–53.3) at 6 weeks. Thereafter, the NDI score deteriorates again to 29.0 (IQR 13.0–49.0) at 12 weeks, 28.0 (IQR 16.0–47.0) at 26 weeks, 38.0 (IQR 32.0–44.0) at 39 weeks and 45.0 (IQR 20.0–57.0) at 52 weeks.

The data suggest that ACD may provide better early functional outcomes, whereas PCF may provide better longer-term functional outcomes. The finding that the early functional improvements following ACD are not maintained is supported by the findings of DePalma et al. 9 and Persson et al.,7 who showed that, by 12 months after the surgery, the difference between surgical and non-surgical patients diminishes. The loss of benefit was due to both ongoing improvement of the conservative group and late recurrence of symptoms in surgical patients. 7,10 Change in functional outcome over time following PCF is not well documented.

Neck and arm pain

Neither PCF nor ACD are usually considered to be effective in the treatment of neck pain. In the FORVAD trial, median neck pain scores deteriorated from a baseline of 5.5 (IQR 4.0–8.0) in the PCF arm and 5.0 (IQR 4.0–7.0) in the ACD arm to 8.5 (IQR 6.0–10.0) in the PCF arm and 7.0 (IQR 4.0–9.0) in the ACD arm on the first postoperative day. This is considered to be because of the surgical wound and provides some support for the view that PCF is a more painful procedure. 16,17 We consider that postoperative neck pain can be minimised by using small wounds, microsurgery and good operative technique. In this trial there were no cases for which a PCF minimal-access technique was used. This approach minimises muscle trauma,27 and may therefore also be effective in reducing postoperative neck pain.

Neck pain scores fell to their lowest level at week 12 [median scores of 3.0 (IQR 2.0–7.0) and 2.5 (IQR 1.0–5.0) in the PCF and ACD arms, respectively] and then increased back towards baseline levels by 52 weeks [median scores of 4.0 (IQR 2.0–5.0) and 5.0 (IQR 3.0–7.0) in the PCF and ACD arms, respectively]. The PainDETECT questionnaire was used to assess whether pain was neuropathic or nociceptive.

In the spine, nociceptive pain is usually considered to be due to muscle, joint or ligamental pathologies and responds poorly to surgical decompression, whereas neuropathic pain results from nerve compression, neurological dysfunction or injury. Neuropathic pain can be divided121 into dysaesthetic (burning) pain that does not normally respond to surgical intervention, and radiating pain that is associated with paraesthesias (pins and needles) and improves with surgical decompression. More recently, a third type of pain, termed nociplastic pain, has been described for the pain predominantly associated with central sensitisation. 122 Spinal pain may be ‘mixed pain’, for which all three components are present. 121

In the FORVAD trial data, the percentage of participants with nociceptive pain in the ACD group increased from 11.1% pre operation to 33.3% after 52 weeks, whereas in the PCF group the increase was from 14.3% pre operation to 21.4% at 52 weeks. The failure to control nociceptive neck pain in the longer term and the change in the character of the pain from neurogenic to nociceptive further suggests that the early benefits of surgery may not be maintained in the longer term. We consider that the PainDETECT questionnaire was inadequate to distinguish between dysaesthetic and radiating pain and was unable to identify any nociplastic component, which limits its utility as a tool for patient selection. Other authors have published similar conclusions relating to the lumbar spine. 123

The median arm pain score improved in both groups, from a baseline of 7.0 (IQR 4.0–8.0) in the PCF group and 6.0 (IQR 5.0–7.0) in the ACD group to 3.0 (IQR 2.0–8.0) in the PCF group and 4.0 (IQR 0.5–5.0) in the ACD group on the first postoperative day, with the lowest median scores recorded at week 12 [median scores of 3.0 (IQR 2.0–8.0) and 2.5 (IQR 0.0–5.0) in the PCF and ACD arms, respectively]. However, as with neck pain, longer follow-up demonstrated slow deterioration in median arm pain scores to 5.0 in both arms of the trial by 52 weeks (IQRs of 2.0–7.0 and 3.0–6.0 in the PCF and ACD arms, respectively).

Other markers of outcome

A total of 54% of participants reported taking time off work following PCF, whereas 62% of participants reported taking time off work following ACD, although the numbers in each arm are small.

The median changes in the HRQoL score (EQ-5D-3L) from baseline at day 1 and at weeks 6, 12, 26, 39 and 52 were, respectively, 0.06, 0.10, 0.00, 0.00, 0.29 and 0.06 in the PCF arm and 0.00, 0.02, 0.00, 0.00, –0.03 and –0.09 in the ACD arm. The baseline EQ-5D score was 0.291 (SD 0.34, 95% CI 0.07 to 0.51) in the PCF arm and 0.595 (SD 0.25, 95% CI 0.40 to 0.78) in the ACD arm; ceiling effects will have had some influence.

The mean length of hospital stay was 1.2 days (SD 0.3 days) in the PCF arm and 1.3 days (SD 1 day) in the ACD arm. The median length of stay was 1 day for both groups. Minimal-access PCF is now commonly performed as a day-case procedure,124 but no minimal-access PCF cases were included in the FORVAD trial. If day-case minimal-access PCF was to be widely adopted in the UK, a further reduction in the length of stay, and therefore in the cost of PCF, would be expected.

Arm strength and sensation was assessed using the ASIA score at day 1 and at 6 weeks. Very few participants had clinically detectable weakness or sensory loss pre operation in either group, and this did not change in the postoperative period.

Efficacy summary

The data in this study suggest that, 52 weeks after ACD, there is limited residual benefit in neck pain, arm pain, pain character or functional outcome. Similarly, 52 weeks after PCF, there is little residual benefit in neck pain or arm pain, but functional outcome improves.

Posterior cervical foraminotomy does appear to be associated with worse postoperative pain on day 1, but this does not affect length of stay, as PCF participants had a shorter mean length of stay. These findings need to be interpreted cautiously in view of the small numbers in the trial, and care should be taken to consider the spread of the data. Nevertheless, the data on NDI score and neck and arm pain are consistent with the findings of other authors27 that PCF is likely to be at least equivalent to ACD in improving outcome after 1 year.

Dysphagia (difficulty swallowing)

Dysphagia was reported as a complication in two out of nine participants following ACD. Furthermore, the EAT-10 dysphagia assessment tool shows clinically significant swallowing problems in most participants on the first postoperative day. This improved by 6 weeks and resolved by 12 weeks.

Globus is the sensation of a lump in the throat, and was assessed using the GETS. Postoperatively, there was a deterioration in the ACD group that became milder over time, but remained higher than in the PCF group throughout the whole trial period, although IQRs overlapped at multiple postoperative time points.

The finding that postoperative dysphagia and globus are common after ACD is consistent with the literature. 125 Patient-reported instruments are effective at identifying globus. The incidence of dysphagia after anterior cervical spine surgery in the literature has been reported as between 1% and 79%. Dysphagia is known to improve over time and is associated with multilevel surgery, female gender, increased operative time and older age (usually > 60 years). 125

Hoarse voice

The VHI-10 and GRBAS scores were used to assess postoperative hoarse voice. The data are similar to the dysphagia data in that the VHI-10 score appears elevated in the ACD group on the first postoperative day, but this difference resolves at later time points.

The incidence of hoarseness in this trial reflects the known incidence in the literature following ACD, which is reported as 0.06–11%, but most will resolve over time. It fails to resolve in 0–3.5% of cases. 126 Hoarseness is caused by damage to the recurrent laryngeal nerve; however, many unilateral recurrent laryngeal nerve injuries are asymptomatic, so the incidence of postoperative hoarseness is an underestimate of the incidence of recurrent laryngeal nerve palsies. 127

Reoperations

There were no reoperations in the trial. Treatment failure may sometimes necessitate that surgery be performed at the same level, but from the other aspect of the neck. However, this is uncommon within 1 year of the index surgical procedure.

In the literature, the risk of adjacent segment disease following ACD is around 3% per annum. 128 The risk of symptomatic adjacent segment disease following PCF is 1.43% per year in the literature. However, with only 1 year of follow-up, it is unlikely that adjacent segment disease will have had time to present. 129

Therefore, the absence of reoperations is likely to be due to relatively short follow-up of the participants; future studies should consider using longer follow-up periods of 2, or even 5, years to capture these data.

In addition, it is possible that the small sample size, in conjunction with low rates of reoperation in the first year, contributed to the lack of observed reoperations.

Key recommendations for future research

• Better imaging protocols, such as oblique MRI scans of the cervical nerve root foramen, should be studied to assess whether or not they could inform which surgical approach is better for different types of nerve root compression.

• Diffusion tensor imaging (i.e. a type of diffusion-weighted MRI) of the nerve root should be studied to assess whether or not it can help identify when compression is functionally important.

• In this trial, complication rates seem higher and benefits less sustained than in some of the published literature. A large prospective cohort study of all operations for cervical radiculopathy in multiple centres in the UK might better inform surgeons as to the real-life outcomes.

Principal investigators

Thank you to all of the PIs who participated in the trial: Mr Yahia Al-Tamimi (Sheffield Teaching Hospitals NHS Foundation Trust), Mr Neil Buxton (The Walton Centre NHS Foundation Trust), Mr Nicholas Haden (University Hospitals Plymouth NHS Trust), Mr Nitin Mukerji (South Tees Hospitals NHS Foundation Trust), Mr Ravindra Nannapaneni (Cardiff and Vale University Health Board), Professor Marios Papadopoulos (St George’s University Hospitals NHS Foundation Trust), Mr Anantharaju Prasad (Lancashire Teaching Hospitals NHS Foundation Trust), Mr Ivan Timofeev (Cambridge University Hospitals NHS Foundation Trust), Mr Christos Tolias (King’s College Hospital NHS Foundation Trust), Mr Senthil Selvanathan (Leeds Teaching Hospitals NHS Trust) and Mr Nitin Shetty (Whittington Health NHS Trust).

We would also like to thank all other members of the local research teams at each and every one of the FORVAD trial sites for their valued contributions to the set-up and delivery of the trial.

Recruiting surgeons

The following surgeons randomised participants to the trial: Mr Anantharaju Prasad (Lancashire Teaching Hospitals NHS Foundation Trust), Mr Simon Thomson and Mr Senthil Selvanathan (Leeds Teaching Hospitals NHS Trust), Mr Yahia Al-Tamimi (Sheffield Teaching Hospitals NHS Foundation Trust), Mr Nitin Mukerji (South Tees Hospitals NHS Foundation Trust) and Mr Nick Carleton-Bland (The Walton Centre NHS Foundation Trust).

Contributions of authors

Simon Thomson (https://orcid.org/0000-0003-4827-1961) (Consultant Neurosurgeon) was the chief investigator; was responsible for the design, conduct, co-ordination and management of the trial; and contributed to data interpretation and preparation of the final report. He also made a significant contribution to recruitment as a participating surgeon at Leeds Teaching Hospitals NHS Trust.

Gemma Ainsworth (https://orcid.org/0000-0002-9952-8440) (Senior Medical Statistician) was the trial statistician and made a substantial contribution to the analysis, interpretation and reporting of the trial, including preparation of the final report.

Senthil Selvanathan (https://orcid.org/0000-0002-3172-295X) (Consultant Spinal Neurosurgeon) contributed to the design, conduct and management of the trial, in addition to data interpretation and the review of the final report. He was also the PI at the Leeds Teaching Hospitals NHS Trust and made a significant contribution to recruitment.

Rachel Kelly (https://orcid.org/0000-0002-2912-2859) (Senior Trial Co-ordinator) played a central role in trial co-ordination and contributed to the preparation of the final report.

Howard Collier (https://orcid.org/0000-0002-0107-0604) (Senior Data Manager) was responsible for data management and contributed to the preparation of the final report.

Ruben Mujica-Mota (https://orcid.org/0000-0002-7430-2744) (Associate Professor in Health Economics) was responsible for the health economics analysis and contributed to the preparation of the final report.

Rebecca Talbot (https://orcid.org/0000-0002-3020-7585) (PhD candidate) contributed to the conduct, analysis and write-up of the qualitative substudy and was involved in the preparation of the final report.

Sarah Brown (https://orcid.org/0000-0002-1840-3786) (Principal Statistician) supervised the statistical aspects of the trial and contributed to the design, analysis, interpretation and reporting of the trial, including preparation of the final report.

Julie Croft (https://orcid.org/0000-0001-7586-3394) (Head of Trial Management) was the project delivery lead for the trial and contributed to the preparation of the final report.

Nikki Rousseau (https://orcid.org/0000-0001-8826-3515) (University Academic Fellow in Healthcare Technology Evaluation) was the qualitative substudy lead. She oversaw and contributed to the conduct, analysis and write-up of the qualitative substudy, and was involved in the preparation of the final report.

Ruchi Higham (https://orcid.org/0000-0002-0529-1041) (Qualitative Researcher) contributed to the conduct, analysis and write-up of the qualitative substudy, and was involved in the preparation of the final report.

Yahia Al-Tamimi (https://orcid.org/0000-0002-9014-2728) (Consultant Neurosurgeon) was the PI at the Sheffield Teaching Hospitals NHS Foundation Trust and contributed to participant recruitment and the interpretation and reporting of the results.

Neil Buxton (https://orcid.org/0000-0003-0589-9863) (Consultant Neurosurgeon) was the PI at the Walton Centre NHS Foundation Trust and contributed to the interpretation and reporting of the results.

Nick Carleton-Bland (https://orcid.org/0000-0002-9253-0599) (Consultant Neurosurgeon) was a participating surgeon at the Walton Centre NHS Foundation Trust and contributed to participant recruitment, and the interpretation and reporting of the results.

Martin Gledhill (https://orcid.org/0000-0002-8530-9597) (PPI representative) was involved in the trial design, advised on the PPI aspects of the trial and contributed to the preparation of the final report.

Victoria Halstead (https://orcid.org/0000-0003-2514-7351) (Speech and Language Therapist) was responsible for the hoarse voice analysis and contributed to data interpretation and the reporting of the trial.

Peter Hutchinson (https://orcid.org/0000-0002-2796-1835) (Consultant Neurosurgeon) contributed to the design, conduct and management of the trial, and the interpretation and reporting of the results.

James Meacock (https://orcid.org/0000-0002-2775-1838) contributed to the conduct and management of the trial, and the interpretation and reporting of the results.

Nitin Mukerji (https://orcid.org/0000-0002-7427-9230) (Consultant Neurosurgeon) was the PI at South Tees Hospitals NHS Foundation Trust and contributed to participant recruitment, and the interpretation and reporting of the results.

Debasish Pal (https://orcid.org/0000-0003-4120-7692) (Consultant Spinal Neurosurgeon) contributed to the design, conduct and management of the trial, and the interpretation and reporting of the results.

Armando Vargas-Palacios (https://orcid.org/0000-0002-6503-0134) (Health Economist) was involved in the health economic aspects of the trial and contributed to data interpretation and the reporting of the results.

Anantharaju Prasad (https://orcid.org/0000-0002-7245-0110) (Consultant Neurosurgeon) was the PI at Lancashire Teaching Hospitals NHS Foundation Trust and contributed to participant recruitment, and the interpretation and reporting of the results.

Martin Wilby (https://orcid.org/0000-0001-6647-9040) (Consultant Spinal Neurosurgeon) contributed to the design, conduct and management of the trial, and the interpretation and reporting of the results.

Deborah Stocken (https://orcid.org/0000-0001-8031-1738) (Professor of Clinical Trials Research) was the scientific lead and provided senior oversight in the methodology and delivery of the trial, and contributed to data interpretation and the preparation of the final report.

All authors contributed to data interpretation and the writing and/or review of the manuscript.

Data-sharing statement

All data-sharing requests should be submitted to the corresponding author for consideration. Access to anonymised data may be granted following review.

Patient data

This work uses data provided by patients and collected by the NHS as part of their care and support. Using patient data is vital to improve health and care for everyone. There is huge potential to make better use of information from people’s patient records, to understand more about disease, develop new treatments, monitor safety, and plan NHS services. Patient data should be kept safe and secure, to protect everyone’s privacy, and it’s important that there are safeguards to make sure that it is stored and used responsibly. Everyone should be able to find out about how patient data are used. #datasaveslives You can find out more about the background to this citation here: https://understandingpatientdata.org.uk/data-citation.

Disclaimers

This report presents independent research funded by the National Institute for Health and Care Research (NIHR). The views and opinions expressed by authors in this publication are those of the authors and do not necessarily reflect those of the NHS, the NIHR, the HTA programme or the Department of Health and Social Care. If there are verbatim quotations included in this publication the views and opinions expressed by the interviewees are those of the interviewees and do not necessarily reflect those of the authors, those of the NHS, the NIHR, the HTA programme or the Department of Health and Social Care.

Stop criteria (red)

If the overall mean recruitment rate is < 0.7 patients per centre per month and there are fewer than eight actively recruiting centres, then recruitment is very unlikely to reach the target sample size of 252 patients, even with a 12-month recruitment extension. Therefore, in this scenario, we would consider that it is futile to continue recruitment and will recommend stopping the trial.

Continue (green)

If the overall mean recruitment rate is at least 1.1 patients per centre per month across 12 actively recruiting centres, then we expect to reach the recruitment target of 252 patients with no remedial action required. The team will be confident that there are no concerning issues that may threaten recruitment to the trial within the planned recruitment timelines.

Amend (amber)

If the overall mean recruitment rate is between 0.7 and 1.1 patients per centre per month or fewer than 12 actively recruiting centres are opened, then a recovery plan detailing remedial actions, including increasing the number of centres and other recruitment initiatives, will be submitted; if approved, the trial will proceed with caution.

Optional speech collection

If the researcher is unable to gain 20–30 seconds of speech from the participant in the spontaneous speech section, a scripted passage may be read by the participant as an alternative. The passage used will be ‘The Rainbow Passage’. The script is as follows:

When the sunlight strikes raindrops in the air, they act like a prism and form a rainbow. The rainbow is a division of white light into many beautiful colours. These take the shape of a long round arch, with its path high above, and its two ends apparently beyond the horizon. There is according to legend, a boiling pot of gold at one end. People look, but no one ever finds it. When a man looks for something beyond his reach, his friends say he is looking for the pot of gold at the end of the rainbow.

The Rainbow Passage, a public domain text, can be found on page 127 of the 2nd edition of Grant Fairbanks’ Voice and Articulation Drillbook (New York, NY: Harper & Row). 133