Comparison of surgical or non-surgical management for non-acute anterior cruciate ligament injury: the ACL SNNAP RCT

Article history

The research reported in this issue of the journal was funded by the HTA programme as award number 14/140/63. The contractual start date was in October 2016. The draft manuscript began editorial review in May 2022 and was accepted for publication in November 2022. The authors have been wholly responsible for all data collection, analysis and interpretation, and for writing up their work. The HTA editors and publisher have tried to ensure the accuracy of the authors’ manuscript and would like to thank the reviewers for their constructive comments on the draft document. However, they do not accept liability for damages or losses arising from material published in this article.

Copyright statement

Copyright © 2024 Beard et al. This work was produced by Beard et al. under the terms of a commissioning contract issued by the Secretary of State for Health and Social Care. This is an Open Access publication distributed under the terms of the Creative Commons Attribution CC BY 4.0 licence, which permits unrestricted use, distribution, reproduction and adaptation in any medium and for any purpose provided that it is properly attributed. See: https://creativecommons.org/licenses/by/4.0/. For attribution the title, original author(s), the publication source – NIHR Journals Library, and the DOI of the publication must be cited.

2024 Beard et al.

Background

Anterior cruciate ligament (ACL) rupture is a common injury, mainly affecting young, active individuals with estimated 200,000 injuries annually in the USA. 2,3 ACL injury can have a profound effect on knee kinematics (knee movement and forces), with recurrent knee instability (giving way) as the main problem. 4 Furthermore, the injury can lead to poor quality of life (QoL), decreased activity5 and increased risk of secondary osteoarthritis (OA) of the knee. 6 Some patients, once recovered from the initial injury, are able to function well without their ACL (copers), usually after undergoing some formal rehabilitation. 7 Other patients continue with episodes of knee instability, and surgery [ACL reconstruction (ACLR) using a graft] is thought necessary to stabilise the knee.

Current treatment/management options

Current management of ACL deficiency (ACLD) includes both surgical and non-surgical treatments (rehabilitation). In the UK, however, a surgical management strategy has become the preferred treatment for individuals with ACL injuries. A survey conducted as part of the feasibility work at the outset of this study showed that the ratio of surgical intervention to non-surgical conservative intervention is 4:1 (unpublished data). Our data suggested that 80% of non-acute patients were directly listed for surgery in the NHS. In England, an estimated 15,000 primary ACLR surgeries are performed each year. 8,9 However, this is a modest estimate based on NHS Hospital Episode Statistics (HES) data, and the real figure for a UK population of 63 million may be closer to 50,000 pa (based on Swedish ACL registry data – incidence 71/100,000 pa). 10 Work related to this project showed that the age-standardised rate of ACLR in the UK increased 12-fold from 1997 to 2017. 9 The rate reached 24.2/100,000 population in 2017 with the largest increase of 48% in 2009. Based on the conservative estimate (n = 15,000), the costs of ACLR to the NHS in 2015 (at inception of this project) was approximately £63 million. Today the cost is more likely to be closer to £85 million.

Rationale for the ACL SNNAP trial

At the outset of this study, it was highlighted that despite ACLR being common, management for ACL injury was based on limited evidence. 7,11–13 At the timing of application for funding, a Cochrane systematic review examined whether surgery or non-surgical (conservative) management was superior for ACL injury14 and concluded no high-quality evidence exists on which to base practice. Surgical stabilisation of the knee joint appears a beneficial intervention, but whether the surgery is more beneficial than non-surgical intervention is unclear, particularly in the later-presenting non-acute patient. Although complications from ACLR have been shown to be infrequent,15 they can still exist and could therefore impact on the outcome for surgery and appropriately influence decision-making.

The unsupported preference for surgical management of the ACL-deficient knee was questioned by evidence obtained in a Scandinavian trial (KANON trial). 16 The benefit of surgery, for all injured patients, was shown to be uncertain, with an operation being unnecessary in many cases. Frobell et al. 16 showed that a period of prior rehabilitation before considering operation can reduce ACL surgery by up to 50%. The clinical implication was that a period of rehabilitation should always be offered prior to surgical reconstruction, and this had become accepted practice, particularly with isolated ACL tears without comorbidity [although more recent randomised controlled trial (RCT) evidence from the COMPARE trial17 has subsequently questioned this reasoning – see Evidence update since study submission]. While this original clinical decision-making evidence was valid for acutely injured individuals, it was not considered applicable to those more typically seen in the NHS, where patients are often non-acute, having sustained injury sometime earlier. By the time NHS patients are diagnosed and begin dedicated ACL injury management, up to 12 months can have passed since initial injury. 18

The mixed acknowledgement and uptake of this evidence and the uncertainty over the applicability to a less acute UK population have resulted in a highly varied approach to managing ACL injury in the NHS. 19–21

The limited evidence, particularly for the non-acute population, means that surgical management may potentially be overused for later-presenting patients. Conversely, an argument can be made in favour of bypassing further formal rehabilitation for the longer-standing ACL-deficient knee and the optimum treatment being immediate reconstructive surgery. Which strategy is the most clinically effective and cost-effective for this subpopulation remains unknown. Because surgery is expensive and may also have greater complications,8,22 provision of strong evidence for automatic default ACLR is even more important. 23

Likewise, the routine prescription of formal rehabilitation, if not beneficial, is considered wasteful and may disadvantage individuals with ACL injuries by delaying optimum treatment or fully stabilising the knee. There was a clear need to identify the most appropriate treatment strategy.

In summary, at the outset of this study there was no evidence-based management of non-acute ACLD occurring, particularly in the NHS. Moreover, little consensus existed on the management of these patients. The aim of the ACL Surgery Necessity in Non-Acute Patients (ACL SNNAP) trial was to address the gap in the evidence base regarding the clinical and cost effectiveness of these approaches and inform standards of care for ACL injury management in longer-standing non-acute patients.

Evidence update since study submission

An update of the initial literature search and review of the clinical trials registries was conducted to inform this report and set the results in context (see Table 1).

During the course of this study, the ongoing study by Reijman et al. (COMPARE trial)17,24 examining the clinical and cost effectiveness of two treatment strategies for ACL rupture has also been completed. The COMPARE trial was carried out in the Netherlands and has a sample size of 188 participants. At 2-year follow-up, the results demonstrated slightly better self-reported outcomes (knee symptoms, self-reported knee function and perception of the ability to participate in sports) in the immediate surgery group compared with the conservative group, but none of these findings were considered large enough to be clinically important. 17 As this study also evaluates the newly injured (acute) patients and replicates the Scandinavian study setting, again, it cannot be directly applied to the typical NHS pathway. However, the study does increase the contention on how best to manage the acutely injured population with the former recommendation of undergoing rehabilitation first being brought into some question. This contention, albeit in a different population, provided even greater justification for ACL SNNAP.

Research objectives

The primary objective of the ACL SNNAP study was to determine in patients with non-acute ACLD whether a strategy of non-surgical management (rehabilitation) (with option for later ACLR only if required) was more clinically effective and cost-effective than a strategy of surgical management (reconstruction) without prior rehabilitation with all patients followed up at 18 months.

Secondary objectives were to compare the two management strategies regarding the return to activity/level of sports, generic QoL, knee-specific patient-reported outcomes, intervention-related complications, health economics–cost effectiveness, ability to work (e.g. sickness absences/return to work number of days off work and subjective working ability), resource use and costs, expectations of return to activity and confidence in relation to the knee.

Clinician interviews

Semistructured, face-to-face interviews were conducted with purposely sampled 12 clinicians (6 orthopaedic knee surgeons and 6 specialist lower limb physiotherapists) from NHS hospital trusts. All clinicians were experienced in the management of ACL injuries and had expressed an interest in participating in a proposed trial of ACLD management (NIHR HTA 14/140/63).

Recordings were transcribed and analysed using thematic analysis. 33

Patient interviews

Semistructured interviews were conducted with a convenience sample of 15 patients with ACL injury referred to the Outpatient Clinic at the Nuffield Orthopaedic Centre, Oxford University Hospitals NHS Foundation Trust. The age range of patients was 21–57 years. The patients were an active population with a Tegner activity level score ranging from 3 to 10 (a score of 5 indicates participation in recreational sports to a score of 9 indicating participation in competitive sport on a non-professional level). Ten patients were interviewed face to face and five interviewed over the phone.

Recordings were transcribed and analysed using thematic analysis. 33

Summary

Comprehensive work was carried out under the heading of the ACL SNNAP to address issues of equipoise and decision-making.

Exploring clinician views on recruiting patients to the proposed ACL SNNAP trial helped the trial team highlight and understand the potential difficulties clinicians may have in adopting a position of equipoise and recruiting specific patients to the trial.

Exploring patients’ views on surgical and non-surgical treatments for ACLD helped the trial team understand factors that may influence the development of patients’ treatment preference and the potential implications for recruitment to a trial evaluating these interventions. Overall, patients viewed surgery more positively than a non-surgical, physiotherapy approach. A number of patients, however, did express a preference for an initial non-surgical approach, knowing that the option of surgery could be considered if there was no improvement.

Several strategies were employed on the basis of this work to facilitate recruitment in such an equipoise-challenged environment. These strategies and guidelines were disseminated at site visits and more widely at methodology conferences. The recruitment strategies allowed for strong preference of both patients and surgeons (patients being not eligible for the study, surgeons not being eligible to recruit), but ensured those without a strong preference could be accommodated. Both patients and surgeons were provided with contemporary information to enable an informed decision choice to enter the trial as either clinicians or participants.

Trial design

The ACL SNNAP trial was designed as a pragmatic, multicentre, superiority, RCT with two-arm parallel groups and 1:1 allocation ratio to compare non-surgical management [rehabilitation (with option for later ACLR only if required)] or surgical management (reconstruction) options for patients with a symptomatic non-acute ACL injury. A two-stage internal pilot was included with clear progression criteria regarding recruitment.

Rather than a head-to-head comparison of two interventions, the trial was designed as a ‘management’ assessment in which specific events were expected and permitted. This included the option for later surgical intervention (ACLR) in the non-surgical (rehabilitation) arm, if required.

Interventions

The trial compared two routine and well-established management strategies for patients with symptomatic non-acute ACL-injured knees: (1) non-surgical management (rehabilitation) and (2) surgical management (reconstruction).

Intervention content was based on a minimal set of pre-established criteria in order to ensure the integrity of the comparison while allowing for variation in practice in delivering the interventions between both surgeons and physiotherapists [see Non-surgical management (rehabilitation) and Surgical management (reconstruction)]. This pragmatic approach to the delivery of the intervention allowed the management approach to reflect current practice and resource use within the NHS thus aiding generalisation yet included minimal levels of standardised quality and content for both interventions.

Participants

Patients with symptomatic knee problems (instability) consistent with an ACL injury (see Inclusion criteria) were eligible for inclusion. ACLD, either partial or complete tear, was confirmed at a patient’s routine outpatient appointment through clinical assessment and magnetic resonance imaging (MRI) scan.

Anterior cruciate ligament tears can occur as isolated injuries but more commonly occur in conjunction with injuries to other structures of the knee, including menisci, articular cartilage and collateral ligaments. Apart from the pathology detailed in the exclusion criteria below, all other patients with an ACL tear combined with associated injuries were considered for participation in the trial.

Inclusion criteria

• Aged 18 years or above.

• Symptomatic ACLD of the native ligament (instability episodes of frank giving way or feeling unstable) with ACL injury (either partial or complete rupture/tear) confirmed using clinical assessment and MRI scan (patients who had undergone primary ACL reconstruction on the index knee were not eligible).

Setting and locations

Participants were recruited between 1 February 2017 and 12 April 2020 from 29 NHS secondary care hospitals from across the UK. Study sites were selected based on criteria detailed further in the published protocol. 1 This included having an established practice of ACLR and an experienced ACLR knee surgeon and physiotherapy team capable of providing contemporary care.

Whether it was feasible to run the trial at the proposed site was also explored, for example able to offer treatment (ACL surgery or rehabilitation) within the 18-week pathway (in line with current NHS waiting time targets) and provide documentary evidence of the use of a rehabilitation protocol that reflected the guidelines outlined in the protocol. Each site is listed below:

• Morriston Hospital, Swansea Bay University Health Board, Swansea.

• Glan Clwyd Hospital, Betsi Cadwaladr University Health Board, North Wales.

• Wrexham Maelor Hospital, Betsi Cadwaladr University Health Board, Wrexham.

• Countess of Chester Hospital, Countess of Chester Hospital NHS Foundation Trust, Chester.

• Frimley Park Hospital, Frimley Health NHS Foundation Trust, Frimley.

• Wexham Park Hospital, Frimley Health NHS Foundation Trust, Wexham.

• Cheltenham General Hospital, Gloucestershire Hospitals NHS Foundation Trust, Cheltenham.

• Great Western Hospital, Great Western Hospitals NHS Foundation Trust, Swindon.

• King’s College Hospital, Kings College Hospital NHS Foundation Trust, London.

• Chapel Allerton Orthopaedic Centre, Leeds Teaching Hospitals NHS Trust, Leeds.

• Manchester Royal Infirmary, Manchester University NHS Foundation Trust, Mnachester.

• Southmead Hospital, North Bristol NHS Trust, Bristol.

• Peterborough City Hospital, North West Anglia NHS Foundation Trust, Peterborough.

• Nuffield Orthopaedic Centre, Oxford University Hospitals NHS Foundation Trust, Oxford.

• Royal Berkshire Hospital, Royal Berkshire NHS Foundation Trust, Reading.

• Royal Cornwall Hospital, Royal Cornwall Hospitals NHS Trust, Truro.

• Royal Surrey County Hospital, Royal Surrey County Hospitals NHS Foundation Trust, Guildford.

• Salisbury District Hospital, Salisbury NHS Foundation Trust, Salisbury.

• Northern General Hospital, Sheffield Teaching Hospitals NHS Foundation Trust, Sheffield.

• Kings Mill Hospital, Sherwood Forest Hospitals NHS Foundation Trust, Sutton in Ashfield.

• Queen Alexandra Hospital, Solent NHS Trust, Portsmouth.

• Stepping Hill Hospital, Stockport NHS Foundation Trust, Stockport.

• Musgrove Park Hospital, Taunton and Somerset NHS Foundation Trust, Taunton.

• Pinderfields Hospital, The Mid Yorkshire Hospitals NHS Trust, Wakefield.

• University Hospital Coventry, University Hospitals Coventry and Warwickshire NHS Trust, Coventry.

• Leicester Royal Infirmary, University Hospitals of Leicester NHS Trust, Leicester.

• Warrington Hospital, Warrington and Halton Hospitals NHS Foundation Trust, Warrington.

• Wrightington Hospital, Wrightington, Wigan and Leigh NHS Foundation Trust, Wigan.

• Yeovil Hospital, Yeovil District Hospital NHS Foundation Trust, Yeovil.

Recruitment and consent

The process of patient identification and recruitment varied depending on the local treatment pathways at each participating site. Potential patients were identified in routine orthopaedic outpatient and pre-assessment clinics by the local clinical team.

The participating surgeon or member of the clinical team initially approached potential participants who meet the eligibility criteria and informed them of the study. Patients who expressed a potential interest in participating were then referred to a research nurse/physiotherapist for further details about the study and written information. Patients who wished to participate then completed an informed consent form and baseline questionnaire.

Baseline assessment

The baseline assessment included a patient self-reported questionnaire that examined knee-specific QoL [Knee Injury and Osteoarthritis Outcome Score (KOOS) and ACL-QoL], generic QoL [EuroQol-5 Dimensions, five-level version (EQ-5D-5L)], return to activity/level of sport participation (Tegner/Modified Tegner) and resource use, as detailed in Table 2. Full details of these measures are provided in the Outcomes section. Once the baseline questionnaire was completed by the patient, they were then randomised into the study as detailed below. Details of the baseline level of ACL injury and associated knee pathology from the MRI report were also collected.

Randomisation

Randomisation was performed by computer allocation (thus ensuring concealment of sequence generation) using a centrally managed web-based automated system provided by Fr3dom limited. Random allocation was to one of two management options: non-surgical management (rehabilitation) or surgical management (reconstruction) on a 1:1 basis. The allocation was generated using permuted block randomisation with varying block sizes stratified by baseline KOOS score (< 30 or ≥ 30) and recruitment site.

Randomisation by local hospital research teams took place following the baseline assessment visit. This occurred either at the time of the patient’s outpatient preoperative assessment visit or at a ‘separate research visit’ around these routine appointments, depending on the local hospital set-up. Following randomisation, the allocation details were displayed on the web-based system for each participant, and an automated e-mail also sent to the designated member of the research team at the site to inform them of the allocation.

A standard letter was used to inform the admissions, care pathway co-ordinators and general practitioner (GP) (with patient consent) of allocation.

Outcomes

The primary outcome for the study was the KOOS4 at 18 months post randomisation. This outcome measure is derived from four of five subscales: pain, symptoms, difficulty in sports and recreational activities and knee-related QoL,39 with scores ranging from 0 to 100, and a higher score indicating better health. KOOS is a validated patient-reported outcome used in ACL research (including recent RCT of acute ACL patients16,25 and large-scale databases, that is the National Ligament Registry). 40,41 The KOOS4 is sensitive and specific for detecting functional deficits due to knee instability.

Secondary outcome measures were used to further assess knee-specific QoL, return to activity/level of sport participation, patient health-related QoL, resource use, frequency of complications. The outcomes reflected consensus opinion in a patient and public involvement (PPI) group and the reference standard for evaluating ACL injury/reconstruction. 42 These were as follows:

Data collection and management

Statistical methods and study analysis

The methods outlined here are primarily for clinical effectiveness. The methods for the cost-effectiveness analysis are included in Chapter 5.

Patient and public involvement

Patients contributed to the design of the study and supported the development of the funding proposal and conduct of the study. Early in the project, the PPI group helped ensure that patient information sheets and report forms were accessible and user-friendly. A patient representative was an active member of the TSC and, as part of this role, contributed to the monitoring and supervision of the trial progress.

Ethics approval and monitoring

Favourable ethics opinion for the ACL SNNAP trial was given by the National Research Ethics Service, Oxfordshire Research Ethics Committee (REC) in October 2016 (16/SC/0502).

Trial recruitment

Twenty-nine sites across the UK recruited 316 trial participants between 1 February 2017 and 12 April 2020 (see Figure 1). One hundred and fifty-six participants were randomised to the surgical management arm and 160 to the rehabilitation arm.

FIGURE 1.

Participant recruitment throughout duration of trial.

Participant flow through trial

Figure 2 shows the Consolidated Standards of Reporting Trials (CONSORT) flow diagram for the trial, summarising participants’ movement through the trial from screening to randomisation, and on to treatment received. Also detailed are the numbers of participants who declined to be included in the trial, as well as those who were ineligible for the study.

FIGURE 2.

Consolidated Standards of Reporting Trials flow diagram. ^aReasons patients ineligible and declined to participate are detailed in Table 3.

TABLE 3 - Screening data

KL, Kellgren and Lawrence; LCL, lateral collateral ligament.

Total screened	1403
Total not eligible (reasons)	602
Other	159
Pregnancy	2
Inflammatory arthropathy	3
Grade 3 MCL/LCL injury	55
Grade 3 MCL/LCL injury + other	1
Grade 3 or 4 on KL scale	17
Grade 3 or 4 on KL scale + other	1
Meniscal pathology	148
Meniscal pathology + other	3
Meniscal pathology + grade 3 MCL/LCL injury	3
Previous knee surgery	91
Previous knee surgery + grade 3 or 4 on KL scale	2
Previous knee surgery + meniscal pathology	3
Previous knee surgery + meniscal pathology + grade 3 MCL/LCL injury	1
Previous knee surgery + meniscal pathology + grade 3 or 4 KL scale	1
Acute injury	81
Acute injury + other	9
Acute injury + grade 3 MCL/LCL injury	3
Acute injury + meniscal pathology	15
Acute injury + meniscal pathology + grade 3 MCL/LCL injury	1
Acute injury + meniscal pathology + grade 3 or 4 KL scale	1
Acute injury + previous knee surgery	1
Grade 3 MCL/LCL injury + previous knee surgery	1
Total eligible to be randomised	801
Total eligible but not participating (reasons)	485
Patient preferred surgery	276
Patient preferred physiotherapy	115
Other	67
No reason given	27
Total randomised	316

Baseline comparability

Participant randomisation was stratified by both KOOS4 score at baseline and recruiting centre. A breakdown by these two stratification factors is shown in Table 4. The randomised groups were similar for these factors as anticipated.

Table 5 shows the breakdown of baseline characteristics for each treatment group and overall. Groups were well balanced in general. The mean participant age was 33 years. A slightly higher proportion of patients were male in the surgical group compared to the rehabilitation group (71% vs. 62%). Around a third (34%) of patients had < 4 months between their injury and randomisation. The numbers of associated knee injuries reported on baseline MRI scan (see Appendix 3, Table 31) were well balanced. Baseline patient reported outcome measures (PROMs) were also well balanced between the two treatment groups.

Loss to follow-up

Data completion was generally good at the 18-month primary outcome time point, although it was poorer at the 6- and 12-month collection time points. Of the 313 forms sent out at 18 months, 248 (79%) were completed and returned. At 6 and 12 months, 64% and 54% of forms were completed and returned, respectively.

Three patients withdrew from ACL SNNAP, with one patient requesting that none of their data collected up to the point of withdrawal be used for purposes of the study.

Table 6 shows a breakdown of the reasons for injury by intervention group. Over one-third (38%) of injuries were sustained playing football. Other sports (including netball, rugby, skiing/snowboarding and non-specific sports) made up a significant proportion of other injuries.

Treatment received

Of the 156 participants allocated to surgical management, 110 (71%) received surgery and completed postoperative rehabilitation as per the original treatment pathway. Seventeen (11%) participants in the surgical management arm did not undergo surgery and instead underwent only rehabilitation. In the rehabilitation arm, 125 (78%) of 160 participants received initial rehabilitation treatment within the trial. Of these, 61 (49%) patients completed rehabilitation treatment with no subsequent surgery, 39 (31%) had subsequent reconstruction for continued symptoms and 25 (20%) started rehabilitation but did not complete the treatment (see Figure 2). Thirty-five (22%) of 160 patients in the rehabilitation group did not undergo the allocated rehabilitation treatment, of whom 26 (74%) had subsequent reconstruction. Sixty-five (41%) patients allocated to rehabilitation had subsequent reconstruction for ongoing symptoms in line with the protocol. The numbers of participants in each PP analysis population are shown in Table 7.

Details of preoperative rehabilitation for those who received surgery are shown in Table 8. Data were available for 14 participants (9%) in the surgical arm who received some form of preoperative rehabilitation, compared to 121 participants (76%) in the rehabilitation arm. As expected, due to the clinical pathway for each arm, there were far more preoperative rehabilitation sessions in the rehabilitation arm compared to the surgical arm overall, and more on average when they did receive some rehabilitation which was unrelated to an operation. Staff grades were comparable across treatment groups, as was the average session length. There was a higher proportion of one-to-one sessions in the rehabilitation group compared to the surgical arm (58% vs. 48%), and the content of sessions also varied between arms.

Data on ACLR operative data are shown in Table 9. One hundred and thirteen participants (72%) received ACLR surgery in the surgical arm, compared with 65 participants (41%) in the rehabilitation arm. A comparable proportion of surgeries were day cases (87% in surgical arm, 93% in rehabilitation arm). Operation and theatre times were similar in both arms (around 115 minutes). Most of participants (71%) in the surgical arm were classified as American Society of Anaesthesiologists (ASA) grade 1, compared to 80% in the rehabilitation arm. The majority of surgeries were arthroscopic in approach (97% in surgical arm, 100% in rehabilitation arm).

Table 10 gives details on postoperative rehabilitation sessions. Data of postoperative rehabilitation details were available for just under half of the patients who received surgery in each treatment group (surgical 48%, rehabilitation 40%). Postoperative rehabilitation was broadly similar across the treatment groups, although there were some minor differences in terms of the session content.

Primary outcome – Knee Injury and Osteoarthritis Outcome Score

The mean KOOS4 scores at baseline were 45.7 for participants allocated to surgical management and 43.3 for those allocated to rehabilitation (see Table 5).

Adjusted and unadjusted analyses were carried out on the ITT, conservative per-protocol (PPC) and pragmatic per-protocol (PPP) populations using linear regression. Unadjusted analyses included only the treatment variable in the analysis models, with adjusted analyses further adjusting for baseline KOOS4 scores and allowing for ICC between recruitment sites.

Analyses on the KOOS4 primary outcome for all analysis populations are shown in Table 11. For the ITT analysis of the KOOS4 at 18 months, 128 participants (82%) in the surgical management arm had scores available for analysis, compared to 120 (75%) participants in the rehabilitation arm. Adjusted mean KOOS4 scores at 18 months post randomisation had increased to 73.0 in the surgical management arm, and to 64.6 in the rehabilitation arm. The adjusted mean difference was 7.9 (95% CI 2.5 to 13.2; p = 0.005) in favour of surgical management. The PPP and PPC analyses supported the ITT results, with all treatment effects favouring surgical management at a level reaching statistical significance. All unadjusted analyses also produced statistically significant effects in favour of surgical management.

A secondary AUC analysis was performed on the ITT population using the KOOS4 scores at baseline, 6, 12 and 18 months (see Table 12). AUC values were calculated by extracting parameters from a mixed-effects repeated measures model. The difference in AUC values was 4.1 (95% CI 0.4 to 7.7; p = 0.028), which was statistically significant in favour of surgical management. Figure 3 shows the marginal mean KOOS4 scores over the duration of the trial period split by treatment arm. The score in both groups increases over time with mean value being very similar in both groups at 6 months but higher in the surgery arm at 12 and 18 months.

FIGURE 3.

Marginal mean and 95% CIs for KOOS4 scores at each time point (ITT population).

Figure 4 shows the marginal mean KOOS4 scores with the rehabilitation treatment arm split into those who received surgery and those who did not. Although the rehabilitation patients who did not receive surgery initially perform better than the surgical management arm in the first 6 months, this benefit levels off at 12 months and decreases between 12 and 18 months. Patients in the rehabilitation arm who received rehabilitation then required surgery start off with lower KOOS4 scores at baseline than those who did receive surgery without rehabilitation and also those who received only rehabilitation. This difference in baseline level perhaps explains to some degree why these individuals went on to have surgical treatment. Despite these participants starting off lower, the marginal means KOOS4 scores eventually surpass those of the patients in the rehabilitation arm who did not receive surgery at the 18-month time point.

FIGURE 4.

Marginal mean and 95% CIs for KOOS4 scores at each time point, broken down by receipt of surgery in the rehabilitation arm.

Additional analyses on primary outcome

Analyses were carried out on the KOOS4 score to assess sensitivity to missing data, compliance with the allocated treatment and to determine if there were any subgroup effects present.

Sensitivity to missing data was assessed using the rctmiss package in Stata. A pattern-mixture model was used to extend the adjusted linear regression model used for the primary outcome analysis, in order to show graphically the difference in treatment effect for each treatment arm if different mean values are assumed for the missing data. For up to a 5-point difference in KOOS4 scores (in favour and against surgical management group) for the participants with missing outcome data versus those for whom data were available, there remained a significant difference in favour of surgical management. The output from the rctmiss command is shown in Figure 5.

FIGURE 5.

rctmiss sensitivity analysis for KOOS4 primary outcome.

A CACE analysis was carried out on the KOOS4 outcome to assess the impact of non-compliance with allocated treatment, where compliance was defined as having had surgery at any time between randomisation and 18 months.

Of the 316 participants randomised, 248 had primary outcome data available (120 in the rehabilitation arm, 128 in the surgical management arm). Out of the 120 participants randomised to rehabilitation, 49 (40.8%) received surgery (the ‘always-takers’), while 33 out of the 128 participants (25.8%) randomised to surgical management did not receive surgery (the ‘never-takers’). The unadjusted CACE estimate can be calculated by dividing the unadjusted ITT treatment effect by the observed proportion of compliers. Accordingly, the unadjusted CACE estimate was 25.0 (95% CI 9.4 to 40.5; p = 0.002).

For the adjusted CACE analysis, the adjusted proportion of compliers was calculated using a first-stage linear regression on participants with available primary outcome data, with compliance as the outcome variable and randomised treatment as the dependent group variable, with additional adjustments as specified in the main primary outcome model. This gave an adjusted CACE estimate of 22.8 (95% CI 6.9 to 38.8; p = 0.005) (see Table 13).

TABLE 13 - Complier-average causal effect analysis of KOOS4 score at 18 months

	Surgical management – non-surgical management
	CACE mean difference (95% CI) (p-value)
KOOS4 CACE treatment effect at 18 months, adjusted for baseline KOOS4 and recruitment site	22.8 (6.9 to 38.8) (0.0051)
KOOS4 CACE treatment effect at 18 months, unadjusted	25.0 (9.4 to 40.5) (0.0017)

An additional pre-specified analysis which imputed 12-month KOOS4 scores if participants were missing their 18-month KOOS4 scores was carried out (limited form of last observation carried forward). This added an additional 12 scores to the analysis (6 in each treatment arm). The analysis results remained similar to the ITT analysis (see Table 14).

TABLE 14 - Primary outcome (ITT population) with missing 18-month KOOS4 scores replaced with 12-month scores if present

	Surgical reconstruction	Rehabilitation	Surgical management – non-surgical management
	n, mean (SD)	n, mean (SD)	Mean difference (95% CI) (p-value)
KOOS4 average treatment effect at 18 months, adjusted for baseline KOOS4 and site	128, 73.0 (18.3)	120, 64.6 (21.6)	7.5 (2.1 to 12.9) (0.008)
KOOS4 average treatment effect at 18 months, unadjusted	134, 73.1 (18.0)	126, 65.2 (21.7)	8.0 (3.1 to 12.8) (0.001)

Subgroup analyses were carried out on four subgroups: sex (male vs. female), baseline KOOS4 scores (low vs. high), age (< 40 vs. 40 years and over) and baseline Tegner Activity Scale values (moderate/light vs. high activity). No difference in treatment effect was found for any of the subgroups, although there was substantial uncertainty (see Table 15).

TABLE 15 - Subgroup analysis for KOOS4 scores at 18 months

Subgroup	Subgroup Strata	Mean difference (95% CI)	Interaction effect (95% CI)	Interaction p-value
Gender	Male (n = 209)	8.1 (−0.4 to 16.7)	0.5 (−15.4 to 16.4)	0.949
	Female (n = 107)	7.6 (−2.9 to 18.2)
KOOS4 scores	Low (< 30) (n = 76)	4.8 (−2.5 to 12.0)	−11.8 (−30.7 to 7.0)	0.210
	High (≥ 30) (n = 240)	16.6 (1.6 to 31.6)
Age (years)	< 40 (n = 236)	8.0 (−4.7 to 20.7)	0.3 (−12.2 to 12.8)	0.962
	40 and over (n = 80)	7.7 (3.1 to 12.3)
Tegner Activity Scale	Moderate/light activity (< 5) (n = 42)	6.4 (0.8 to 12.0)	−9.1 (−28.7 to 10.5)	0.350
	High activity (≥ 5) (n = 274)	15.5 (−3.2 to 34.2)

Secondary outcomes

Anterior cruciate ligament quality of life score

Anterior cruciate ligament quality of life scores at 18 months were analysed using linear regression, adjusting for baseline KOOS and ACL-QoL scores and allowing for ICC between recruitment sites (see Table 21).

Eighty-nine participants had ACL-QoL scores taken at 18 months available in the surgical management arm, compared to 81 participants in the rehabilitation arm (57% vs. 51%, respectively). The mean ACL-QoL scores were 59.7 and 48.2 in the surgical and rehabilitation arms, respectively. The adjusted mean difference was 11.6 (95% CI 4.4 to 18.8; p = 0.003) and was statistically significant in favour of surgical management at the 5% significance level.

Patient satisfaction

Patient satisfaction was assessed in two ways: asking patients the nature of their problems at the 18-month time point compared to before their treatment and also if they would still choose to have the same treatment if they were able to go back in time.

A CI for the difference in proportions of participants who said their knee was better at 18 months than before their treatment (but after their injury) was calculated using the Newcombe Method 10 (also referred to as the score method).

One hundred and twenty-three participants had available responses for the ‘better than before’ question in the surgical arm, compared to 116 in the rehabilitation arm. One hundred and two participants (83%) in the surgical management arm said their knee was better than before treatment at 18 months, compared to 79 (68%) in the rehabilitation arm. This gave a difference in proportions of 15% (95% CI 4% to 25%) in favour of surgical management.

A higher proportion of participants in the surgical management arm said they would choose the same treatment again compared to those in the rehabilitation arm (80% vs. 61%, respectively). Conversely, a higher proportion said they would not choose the same treatment again in the rehabilitation arm compared to the surgical management arm (18% vs. 5% respectively). Table 22 gives full details of the patient satisfaction outcome analysis.

Fidelity to rehabilitation

A more detailed analysis of the compliance and fidelity to the management interventions was conducted. 55 Compliance (adherence) in the rehabilitation arm was defined as participants completing sufficient physiotherapy over a period of at least a 90-day time span. Compliance in the surgical arm was defined as participants undergoing surgical reconstruction and postsurgical rehabilitation over at least a 5-month time span. This revealed that adherence (compliance) to treatment was higher in the surgical arm: 110 of 156 participants completed surgical reconstruction and were compliant with postsurgical rehabilitation for a compliance rate of 70% while compliance was 59% in the rehabilitation arm (see Table 23). Compliance to intervention allocation was higher in the rehabilitation arm (78%) than in the surgical arm (72%) (see Table 23).

Intervention fidelity is the similarity of the interventions delivered in the trial to interventions specified in the protocol. 56 In the rehabilitation arm, this was defined as participants’ physiotherapy programmes including a progression of therapeutic activities towards those targeting return to sports or functional activities. Intervention fidelity was high in the rehabilitation arm, with 72% of participants who were randomised to and began physiotherapy meeting fidelity criteria (see Table 23).

Introduction

This chapter reports the methods and results of a within-trial cost-effectiveness analysis of ACL SNNAP. Between February 2017 and April 2020, ACL SNNAP randomly allocated 316 adults with symptomatic non-acute ACL-deficient knees to receive: (1) surgical management (reconstruction n = 156), in which patients were placed on a waiting list to undergo standard ACLR procedure, or (2) non-surgical management (rehabilitation n = 160), in which patients were referred to their nearest physiotherapy department and undergo rehabilitation. Participants have been followed up to 18 months following randomisation. Information on recruitment, including inclusion and exclusion criteria, is presented in more detail in Chapter 3. Participant characteristics at recruitment and clinical results are presented in Chapter 4.

In this chapter, we compared surgery and rehabilitation in terms of QALYs gained and healthcare costs and calculated incremental cost-effectiveness ratios (ICERs), which give the additional spending required in order to generate one additional QALY.

Methods

Results

Within-trial cost-effectiveness analysis

Table 27 shows the main analysis results at 18 months (see Appendix 3, Table 36 for descriptive statistics on costs and QoL at each follow-up point following multiple imputation). Participants in the surgical arm reported higher QALYs compared to rehabilitation, but the difference was not statistically significant (0.052; p = 0.177). Total NHS costs were larger in the surgical arm (£1017; p < 0.001), as a result of higher hospitalisation costs (£1150; p < 0.001).

Adopting an NHS perspective, the ICER for surgical management programme versus rehabilitation was £19,346 per QALY gain, below the standard threshold for cost-effectiveness in the UK (£20,000–30,000 per QALY gain). Similar results were derived when using all healthcare costs (NHS and private), with the ICER for surgical management versus rehabilitation being £19,473.

Figure 6 presents the cost-effectiveness scatter plot giving differences in mean total costs and QALYs for surgical versus rehabilitation management adopting the NHS health and social care perspective. Most bootstrap replicates remained largely in the north-west quadrant of the cost-effectiveness scatter plot, indicating that surgical management resulted in higher QALYs but and also higher costs relative to rehabilitation. Figure 7 presents the cost-effectiveness acceptability curve, which gives the probability that surgical management is cost-effective compared to rehabilitation for different threshold values for a QALY (from £0 to £100,000 per QALY). Adopting an NHS or a healthcare cost perspective, the probability that surgical management is cost-effective is 51% and 72% at a threshold value of £20,000 and £30,000 per QALY, respectively.

FIGURE 6.

Cost-effective scatter plot for the base-case analysis. Scatter plot of estimated joint density of incremental costs and QALYs of surgical management relative to rehabilitation obtained by bootstrap resampling from each of the 30 imputed datasets, running the regression models on each bootstrapped dataset and extracting the estimated incremental costs and QALYs. Dashed lines represent threshold values of £20,000 and £30,000 per QALY gained. Bootstrapped results falling below the lines are deemed cost-effective. From the bootstrapped results, we calculated the probability that surgical management was more cost-effective than rehabilitation for different threshold values per QALY gained.

FIGURE 7.

Cost-effectiveness acceptability curve. Figure plots the probability (y-axis) that surgical management is cost-effective compared to rehabilitation for different willingness-to-pay thresholds per QALY gain (x-axis). Probability captures the joint uncertainty in incremental costs and QALYs of surgical management compared to rehabilitation and was obtained by estimating the proportion of bootstrapped results that were cost-effective for each threshold value. The interpretation is that, given a willingness-to-pay threshold of £30,000 per QALY gained, the probability that surgical management is cost-effective compared to rehabilitation is 0.72.

Subgroup analysis

Table 28 presents cost-effectiveness results at 18 months by subgroups of participants. There is considerable variation in the cost-effectiveness estimates between participant subgroups by age and physical activity prior to injury. Adopting a £30,000 per QALY threshold, surgical management was cost-effective for participants under 40 years of age and for those with high level of physical activity prior to injury. All subgroups by baseline KOOS4 and sex reported ICERs below the £30,000 per QALY threshold.

COVID impact

Of the total sample, 157 completed the 18-month follow-up after 23 March 2020 and 159 completed it before this date. Table 29 reports the total NHS costs and QALYs for these two subgroups of participants. Before the lockdown date, the difference in QALYs between trial arms was higher (0.09, p = 0.080) compared to after the lockdown date (0.02, p = 0.668), albeit there were no significant differences. In terms of difference in NHS costs, this was smaller before the lockdown date (£925, p = 0.001) compared to after the lockdown date (£1101, p = 0.003). These differences translated into an ICER of £10,782 using the participants who completed the 18-month follow-up before 23 March 2020 and an ICER of £54,789 using participants who completed it after this date.

Cost-effectiveness discussion

Over 18 months of follow-up in the ACL SNNAP trial, we found that surgical management led to improved health-related QoL compared to rehabilitation management but with higher healthcare costs. Using £20,000–30,000 per QALY thresholds, we report surgical management to be cost-effective in the UK setting. The probability that surgical management is the most cost-effective option was 51% and 72% at £20,000 and £30,000 per QALY gained, respectively. This is the first study estimating the effectiveness and cost effectiveness of two common interventions for non-acute ACL-injured knees.

Recent research comparing surgical management with rehabilitation in individuals with recent ACL injury (< 2 months) concluded that early ACLR was not cost-effective. 26 The study, conducted in the Netherlands, comprised an economic evaluation alongside a clinical trial where participants were followed up for 2 years. Consistent with our findings, the authors reported early reconstruction to be more effective (0.04 QALYs) but more costly than rehabilitation. The difference in costs between the two trial arms was considerably higher than what we found in the ACL SNNAP trial. This is partially explained by the larger proportion of participants undergoing surgery in the Dutch study compared to the UK study (96% vs. 78%) among those allocated to surgical management. Surgery reconstruction costs accounted largely for the difference in costs observed between the trial arms in both trials.

When trial participants were divided into those completing the trial prior to or after the UK national lockdown (23 March 2020), we found differences in the QALY gains associated with the surgical management arm. These resulted in surgical management being cost-effective in the cohort of participants that completed the trial prior to national lockdown but not cost-effective in those who completed it after the national lockdown. The number of reported rehabilitation sessions among those allocated to the surgical arm was significantly lower (four fewer sessions per individual) for those who completed the trial after the lockdown. There was no significant difference in rehabilitation sessions for those allocated to the rehabilitation arm.

Our cost-effectiveness analysis is based on the largest randomised trial comparison of surgical and non-surgical management of non-acute ACL in the world. The analysis has several limitations, including the sizeable amount of missing data on use of healthcare resources and EQ-5D-5L. We accounted for this using multiple imputation. 61 This assumes data are MAR conditional on modelled covariates and we found no strong evidence to contradict this assumption. The trial was not powered to estimate differences in incremental costs and QALYs, which may explain the probability of surgical management being cost-effective of 51% and 72% at £20,000–30,000 per QALY gained thresholds, respectively. However, with an ICER below £20,000 per QALY, surgical management is cost-effective and should be recommended for implementation relative to non-surgical management. In addition, results were calculated over only 18 months and longer follow-up could confirm whether the observed differences in costs and QoL are maintained. Finally, the trial amendment resulted in the exclusion of data concerning medication use, equipment, informal care and productivity losses. Productivity costs are likely to be high in the participants of ACL SNNAP given their age and working status. Further research would be needed to ascertain differences between the two groups.

Research question

Anterior cruciate ligament SNNAP was a large, multicentre, parallel-group, superiority RCT designed to assess the clinical and cost effectiveness of surgical management compared to initial non-surgical management (and subsequent surgery if necessary) in patients with non-acute more long-standing ACL injuries. The research question arose due to uncertainty around the best treatment for patients with a damaged ACL, particularly in patients who, in the NHS, presented late with their injury and therefore diagnosis. Previous work16,25 had already ascertained that it was beneficial to prescribe remedial exercises and undergo rehabilitation prior to surgical reconstruction, but these findings could not be applied to the less acute population as often seen in the NHS. While the findings and approach still hold for acute patients and may be appropriate for more long-standing patients, clinical anecdote suggested that patients who had been ACL deficient for longer may have already achieved their maximum level of stability with self-exercise or normal activities. An additional round of rehabilitation would therefore likely offer no further benefit. Accordingly, there were two clear treatment pathways to evaluate: firstly, immediate surgical reconstruction as soon as possible after diagnosis and, secondly, rehabilitation, followed by a surgical reconstruction option if the rehabilitation treatment failed to stabilise the knee sufficiently.

Design of the study

For these reasons, a management-style study design was selected. The two-group multi-intervention design allowed for patients in the rehabilitation arm to legitimately undergo surgical reconstruction if necessary. In a standard two-group head-to-head treatment comparison trial, this would be designated as crossover.

The trial, as all non-surgery versus surgery study designs, could not be blinded in any meaningful way as treatments were entirely disparate and the primary outcome was self-reported.

The end point of final KOOS4 score at 18 months was compared irrespective of how many patients underwent subsequent reconstruction. However, these data were still important and collected. In the event of no significant differences between treatment groups, the percentage of patients ‘failing’ rehabilitation and requiring surgical reconstruction would have become a more important variable.

One other feature of the trial that was accounted for in a comprehensive analysis plan was the fidelity and compliance of the interventions, particularly the non-surgical care. It was apparent from the outset that compliance to the rehabilitation programme had the potential to directly affect interpretation. For this reason, compliance data were collected and sub-categorisations made based on compliance/fidelity levels for insertion into different analyses. As there was a clear result from the trial, these multiple PP and compliance-based analyses were not needed for the main question but still provided a valuable insight into ACL rehabilitation compliance in a pragmatic setting.

Summary of main findings

The trial found that both groups improved over time but patients in the surgical management group had substantially better outcomes at 18 months post randomisation than those in the non-surgical management group. For the ITT population, there was a statistically significant difference between the groups for the KOOS4 primary outcome at the 18-month time point in favour of surgical management. This was supported by an AUC analysis over the duration of the study, as well as the analyses on each of the PP populations and CACE analysis which sought to compensate for non-compliance. The main result was also robust when investigating assumptions around missing data. Exploratory subgroups’ analyses suggested benefit for surgical management by age, gender, KOOS4 baseline score and Tegner Activity Score. There was a possible suggestion that those who had lower KOOS4 score at baseline and lower activity on Tegner might get more benefit.

The secondary outcome measures also generally favoured surgical management over initial non-surgical management. There were statistically significant differences shown for each of the KOOS domains, Tegner Activity Score, ACL-QoL and patient satisfaction. A higher proportion of patients said they would choose the same treatment again if they could go back in time in the surgical management group compared to the rehabilitation management group. No difference was found between the groups for the EQ-5D-5L outcome index with the surgical group having a slightly higher observed value, and there were a relatively low number of clinical events recorded in both treatment groups.

Due to the complex nature of the interventions, there were various treatment pathways participants randomised to each group could have taken. Of those randomised to surgical management, 110 out of 156 (71%) had surgery and completed their postoperative rehabilitation, which was below the level expected. Sixty-one out of 160 (38%) patients randomised to non-surgical management completed their rehabilitation without undergoing surgery. Of 160 patients, 65 (40.6%) in the non-surgical group underwent surgery, although this was allowed for in the study protocol.

A subsequent analysis that split the rehabilitation group into those who completed the rehabilitation management in full and those in that allocation who required subsequent surgery because of continued persistent symptoms revealed interesting findings. Patients in the rehabilitation only group (not requiring subsequent reconstruction) improved quickly (as per KOOS4 score) but then rapidly plateaued ending up some 12 points below the surgical group. In contrast, the patients who underwent subsequent reconstruction surgery after failed rehabilitation were slow to progress initially (presumably because of a continuation of instability) but had a final outcome (at 18 months) very similar to the allocated surgery group. This latter pattern in these patients is somewhat expected, as essentially effective treatment is ‘delayed’.

Recruitment and conduct

In the early part of the study, it was found that recruitment was slow and investigations were performed to explore this feature. The screening data collected were invaluable for this exploration and were a significant learning point for the trial. The detail and comprehensive nature of the screening data greatly assisted with re-establishing appropriate recruitment. Recruitment was inhibited for several reasons:

• lack of equipoise of patients

• lack of equipoise of clinicians (surgeons)

• eligibility and inclusion criteria features, including comorbidity and time since injury

• patient pathway issues at some sites where patients had already undertaken formal rehabilitation interventions.

Patient characteristics

Three hundred and sixteen patients with an ACL injury were recruited over 3 years. It is the largest RCT to compare a surgical management with ACLR of the ACL versus initial rehabilitation (and subsequent surgery when necessary) in a routine clinical setting. Baseline demographics were generally very similar at baseline as anticipated given randomisation and particularly for the stratification factors (site and KOOS4 at baseline). There was a small gender imbalance with more women in the non-surgical management group. The average age of the sample was 33 years.

The environment, cause or mechanism of injury showed that the highest percentage of patients in the sample had footballing injuries (n = 121, 38%). High percentages of patients also damaged their knee twisting or falling (15%), in rugby (11%) and snow sports (10%). In the UK, it is thought that this pattern of injury mechanism is generalisable. 66

The change in protocol to allow a wider acuteness boundary resulted in the recruitment of 108 patients (34%) who had sustained their injury inside 4 months previously. The remaining 66% all sustained injury longer than 4 months previously. Twenty-two per cent of patients had sustained their injury longer than a year before recruitment. While the original intention was to explore a highly ‘chronic’ and long-standing population, the trial conduct constraints precipitated some protocol changes to allow a less long-standing injury and full recruitment within the given time. While the change is listed as a limitation, the patients remained non-acute by definition, and therefore the population under evaluation remained valid and for the desired population for the research question, albeit slightly less ‘chronic’. A key interpretative point is to examine how this shift may have influenced the results and interpretation of the trial. The modified inclusion criteria created a population that would more likely benefit from rehabilitation (according to the Frobell findings and recommendations). The change would therefore have served to potentially dilute or moderate any results in favour of surgical management. However, the highly significant findings in favour of surgical management showed the converse to be true and the influence of any adjustment and amendment can be considered negligible.

The baseline characteristics of the population were unremarkable and expected. The overall KOOS4 score was 44.5/100. The patients were predominantly active with 86% of the sample having a Tegner Activity Score above level 5 prior to injury. This dropped to 8% at recruitment to the study. These results show the remarkable drop-off in activity and consequence of ACL injury.

Distribution of treatment received

The adherence (compliance) to allocation and treatment received was reviewed.

The findings for adherence and fidelity were of interest. There was substantial non-adherence in the surgical group, with only 72% of participants receiving ACLR as per allocation, including patients who ultimately declined surgery for work duties (self-employed) or an improvement in their condition. Twenty-eight per cent of patients did not have surgery for a variety of reasons. Seven per cent of patients in the surgical management group were still awaiting surgery at the final follow-up point, and 10% of patients in this group did not have surgery for unknown reasons. Despite this, the ITT analysis still provided a clear indicator of superiority for surgical management.

A substantial 41% of patients allocated to the rehabilitation arm underwent surgery following failure to stabilise the knee by rehabilitation alone. The conversion to surgery was expected and in line with the protocol, but the magnitude of conversion could not be predicted in this non-acute population. The conversion rate at around 2 years follow-up (41%) seen in SNNAP was slightly higher than expected in a non-acute population (39%)16,25 but slightly lower than the 50% reported by Reijman et al. 17 However, without examining outcome scores, it shows there is much uncertainty for individual patients around two very disparate treatment choices, surgical or non-surgical treatment. The high percentage of failure for rehabilitation management probably reflects the limited capability of remedial exercises to stabilise an unstable joint, especially for high levels of activity. This is not a new phenomenon and the characteristics of what allows some patients, and not others, to ‘cope’ without a ligament has not yet been fully delineated despite many attempts. 67

The data for partial and lower-level adherence in rehabilitation group patients were varied, with 16% of these patients starting rehabilitation, but it was unknown how much of the intervention they completed, and 6% having no record of starting rehabilitation or receiving surgery. This shows the difficulties in collecting and validating these types of data.

Of those patients who went on to have surgery after a course of rehabilitation, 16% did so very early on or had very little rehabilitation treatment and 24% completed the full course of therapy but still had an unstable knee at the end of treatment. These patients with ongoing instability (in consideration of the primary and secondary outcome results) probably reflect the overall results and again, the relative ineffectiveness of non-surgical management to stabilise the unstable knee. Nevertheless, it is important to highlight that many patients in the rehabilitation group did not undergo reconstruction and achieved knee stability without surgery. Hence, rehabilitation cannot be dismissed as a viable option and this point should be emphasised. It should be remembered that a trial is comparison and presentation of averages and does not allow or account for individual benefits. Despite evidence of superiority for one management strategy, both treatment groups did improve over time.

Overall, a similar proportion of participants complied with the randomised allocation (in some way) in the non-surgical management group (78%) and surgical group (72%). Various factors will have contributed to the completion of both groups’ management, including the impact of COVID-19 pandemic on scheduling of surgery and provision of rehabilitation, which included ‘virtual’ physiotherapy sessions in some cases. Thematic analysis of information for non-adherence in the PCRFs found that the most frequently reported reasons for non-adherence were as follows: patients not being satisfied with their progress with rehabilitation (26%), reinjury or flare-up of knee pain or instability (17%), difficulties with scheduling, child care, or getting to physiotherapy (15%), the COVID-19 pandemic (9%), participants having difficulty performing rehabilitation exercises (7%) and patient illness (6%). The cause of poor or non-adherence was unknown in 20% of those cases. In the surgical reconstruction group, the reason for not undergoing surgery was unknown in 32% of patients, the COVID-19 pandemic (14%), work commitments (7%) and participants moving out of the area (7%).

These adherence patterns demonstrated how difficult it can be to enforce rehabilitation management in pragmatic trials. Even in a high-profile trial to address a lifestyle affecting pathology, such as in ACL SNNAP, the uptake is often poor. The pragmatic nature of the trial helped give a true reflection of what could be expected in normal NHS practice. The results cannot be extrapolated to high-level sports or high-intensity rehabilitation that is sometimes offered for elite athletes. The trial assessed adequate-quality rehabilitation provision only.

Compliance and fidelity

A more detailed analysis of the compliance to the management interventions was conducted and showed that compliance was asymmetrical, being more of a problem for the rehabilitation group. Compliance (adherence) in the rehabilitation arm was defined as participants completing sufficient physiotherapy over a period of at least a 90-day time span. Compliance in the surgical arm was defined as participants undergoing surgical reconstruction and postsurgical rehabilitation over at least a 5-month time span. This revealed that adherence (compliance) to treatment was higher in the surgical arm: 110 of 156 participants completed surgical reconstruction and were compliant with postsurgical rehabilitation for a compliance rate of 70%, while compliance was 59% in the rehabilitation arm (see Table 23). Compliance to intervention allocation was higher in the rehabilitation arm (78%) than in the surgical arm (72%) (see Table 23).

Intervention fidelity is the similarity of the interventions delivered in the trial to interventions specified in the protocol. 56 In the rehabilitation arm, this was defined as participants’ physiotherapy programmes including a progression of therapeutic activities towards those targeting return to sports or functional activities.

Despite stipulating senior rehabilitation staff were required, in reality this did not always occur and the breakdown on grades (see Table 8) shows that some staff delivering the rehabilitation intervention were less senior and not especially well aligned to the protocol. It is hoped that the more junior staff had some level of supervision, again as stipulated in the protocol.

Safety

The trial did not investigate a life-threatening condition, but there were some aspects of safety to be addressed. A concern when designing the study was that a significant number of patients in the rehabilitation group, while performing exercise on an unstable knee, would have episodes of instability and sustain secondary damage to their menisci. The outcome for an ACL-injured knee with concomitant meniscal damage is potentially suboptimal. However, this did not occur. Only four patients in total had acquired meniscal damage (1%): three in the rehabilitation arm and one in the surgical arm.

Methodological points

The three pre-specified analyses to account for potential non-compliance showed surprising consistency in terms of indicating a benefit for surgical management over initial rehabilitation management in terms of KOOS4 outcome at 18 months. None of the three analyses led to a different overall conclusion with as large or larger gain in KOOS4 for surgical management estimated. It could be argued that these pre-set analyses were not needed in the light of the primary outcome results analysed as ITT.

The increase in KOOS4 over time supports the interpretation that reconstruction provides a greater benefit even if received some time later with the increase in the surgical groups. However, it is worth noting that the very large effect from the CACE will be overly optimistic, in that only non-compliance to surgery is considered and a subset of these patients will not need surgery. Eleven per cent in the surgical management group ended up undergoing rehabilitation and did not have surgery. This may be representative of real life in which a number of listed patients for ACLR do not always complete the surgery.

We also looked at the potential impact of missing data given the sizeable missing data observed at 18 months. The method we used was a pattern-mixture model-based approach where the magnitude of difference in the outcome between those observed and those missing can be specified. We assessed the impact of a difference in one group or both over a reasonable range of values (5-point difference in either direction). The findings were remarkably consistent and did not indicate sensitivity to missing data in terms of the overall conclusion.

Health economics

Over 18 months of follow-up in the ACL SNNAP trial, we found that surgical management led to improved health-related QoL compared to the rehabilitation management but with higher healthcare costs. Using £20,000–30,000 per QALY thresholds, we report surgical management to be cost-effective in the UK setting. The probability that surgical management is the most cost-effective option was 51% and 72% at £20,000 and £30,000 per QALY gained, respectively. This is the first study estimating the effectiveness and cost effectiveness of two common interventions for non-acute ACL-injured knees.

Recent research comparing surgical management with rehabilitation in individuals with recent ACL injury (< 2 months) concluded that early ACLR was not cost-effective. 26 The study, conducted in the Netherlands, comprised an economic evaluation alongside a clinical trial where participants were followed up for 2 years. Consistent with our findings, the authors reported early reconstruction to be more effective (0.04 QALYs) but more costly than rehabilitation. The difference in costs between the two trial arms was considerably higher than what we found in the ACL SNNAP trial. This is partially explained by the larger proportion of participants undergoing surgery in the Dutch study compared to the UK study (96% vs. 78%) among those allocated to surgical management. Surgery reconstruction costs accounted largely for the difference in costs observed between the trial arms in both trials.

When trial participants were divided into those completing the trial prior to or after the UK national lockdown (23 March 2020), we found differences in the QALY gains associated with the surgical management arm. These resulted in surgical management being cost-effective in the cohort of participants that completed the trial prior to national lockdown but not cost-effective in those who completed it after the national lockdown. The number of reported rehabilitation sessions among those allocated to the surgical arm was significantly lower (four fewer sessions per individual) for those who completed the trial after the lockdown. There was no significant difference in rehabilitation sessions for those allocated to the rehabilitation arm.

Our cost-effectiveness analysis is based on the largest randomised trial comparison of surgical and rehabilitation management of non-acute ACL in the world. The analysis has several limitations, including the sizeable amount of missing data on use of healthcare resources and EQ-5D-5L. We accounted for this using multiple imputation. 61 This assumes data are MAR conditional on modelled covariates and we found no strong evidence to contradict this assumption. The trial was not powered to estimate differences in incremental costs and QALYs, which may explain the probability of surgical management being cost-effective of 51% and 72% at £20,000–30,000 per QALY gained thresholds, respectively. However, with an ICER below £20,000 per QALY, surgical management is cost-effective and should be recommended for implementation relative to non-surgical management. In addition, results were calculated over only 18 months, and longer follow-up could confirm whether the observed differences in costs and QoL are maintained. Finally, the trial amendment resulted in the exclusion of data concerning medication use, equipment, informal care and productivity losses. Productivity costs are likely to be high in the participants of ACL SNNAP given their age and working status. Further research would be needed to ascertain differences between the two groups.

Equality, diversity and inclusion

As the trial was embedded in routine NHS care, it was intended to be inclusive (all patients who meet the selection criteria at participating sites were candidates). Outcomes were collected by questionnaire only to reduce participant burden; there were no additional clinical visits to routine care for participants. We maximised methods of follow-up, that is online, postal and phone calls to be inclusive of patients’ preferences for completion.

In addition to study participants, we included a representative sample of clinicians and trial personnel at recruiting sites. The trial was promoted at national conferences to increase awareness of the project and encourage interest and geographic diversity of trial sites. To help remove any potential barriers perceived by inexperienced teams/sites, we emphasised the support available throughout the study from the central study team to enable them to participate. As this was a management study including both rehabilitation exercises (physiotherapy) and surgery, we encouraged co-principal investigators from each discipline to lead the study locally where possible.

Patient and public involvement

Patients contributed to the design of the study, supported the development of the funding proposal and conduct of the study. The proposed research was presented and discussed at a PPI study development meeting supported by the Research Design Service South Central. Early in the project, the PPI group of four patients helped ensure that patient information sheets and report forms were accessible and user-friendly. In addition, the patients from Swansea and Oxford supported work around the testing of the EDC and method of follow-up. A patient representative was an active member of the TSC and, as part of this role, contributed to the monitoring and supervision of the trial progress.

Limitations

Several potential limitations and influences on the study have already been discussed including modifications to the inclusion criteria, issues of equipoise, adherence (compliance) to allocated treatment and fidelity of treatment content. Most of these have been well described and caveats provided to explain why any threat from these factors might be low. There are, however, other aspects that warrant mention.

A surgical versus non-surgical trial, such as this, suffers from an imbalance in sequencing options. Rehabilitation is nearly always considered as the first option and surgery the second. The surgery option is perceived as an escalation in intervention (and therefore can be perceived as having greater a priori efficacy). This has many manifestations. It can produce a perceived suboptimal treatment option in the non-surgical group and create resentful demoralisation, especially if not progressed as soon as expected. We have no evidence or detailed information to describe the effect or influence of this in SNNAP, but it is likely present.

Another potential limitation is the ability to control the professional profiles of clinicians treating patients in the different arms of the study. Unfortunately, we could not describe or enforce any grade or profile of the staff (mainly physiotherapists) treating non-surgical patients a priori, as was possible for surgeons. This information could only be obtained post study and is presented in Results section (see Table 8). Physiotherapy treatment was administered in line with a pragmatic trial design, which resulted in varied experience and expertise. However, a senior physiotherapist was required to overview the treatment regime and provide guidance. This issue is common to most surgery versus non-surgery trials with a surgical operation being highly specialised with risk for serious, even life-threatening complications. Therefore, ACLR should not be carried out by surgeons without the necessary experience, for which it was thought necessary to stipulate for the trial (> 50 operations). The nature of the two main interventions, one a very clearly defined surgical procedure and the other a less stark and defined intervention (rehabilitation) with over 80 different therapeutic interventions, can create difficulties with interpretation. The quality of compliance data then becomes an important factor. Despite detailed recording and reporting for compliance for the rehabilitation arm, it remains a limitation of the RCT. Fortunately, the signal from the ITT analysis was so great in favour of one group that it counteracted this concern and was sufficient to allow confidence in reporting that patients in the surgical arm had greater benefit.

The follow-up, while adequate for power, was not as great as hoped, particularly with early data points. Many strategies were included to improve this, but, as with all surgical trials, solutions to obtaining high rates of follow-up from self-reported instruments remain elusive.

As previously discussed, the change in the injury acuteness of the population should be highlighted. The study set out to examine just long-standing ACL-injured patients but ended up including some patients much closer in time to their original injury.

Strong opinion on the compliance issues from rehabilitation is negated by the pragmatic design. However, it is not possible to say whether the results may have been different if compliance had been higher for the rehabilitation group. As a question we were interested in the real-world delivery and not proof of principle.

Crossover could be considered an issue, but it must be remembered that the change from rehabilitation to surgery was expected and accounted for in the protocol. Several patients in both groups failed to complete treatment, but these were not considered crossovers. The only true crossover patients are those who were allocated to surgery and chose not to undergo reconstruction for various reasons (n = 17). The ITT analysis accounted for all the non-compliance.

It should also be noted that the trial had a substantial number of patients who had a strong preference and were ineligible. The trial results therefore can only apply to patients without preference (in equipoise) over treatment.

Recommendations for future research

The trial was pragmatic and studies to explore the influence of treatment fidelity and compliance, especially in the rehabilitation arm, will be useful. Other recommendations for future research include evaluation of innovative surgical reconstruction or even ligament repair. The best form and content of postoperative rehabilitation following ACLR should also be explored. Future research could also focus on identifying characteristics of those patients who are more likely to benefit from either surgery or rehabilitation. This can be used to support shared decision-making on management and treatment. Further methodology work exploring the challenges in conducting trials comparing surgical versus non-surgical interventions, including patient preference and clinician equipoise, and compliance would also be beneficial. How to obtain improved compliance for rehabilitation (and reporting), perhaps using wearables or assessing other biomarkers, may be of value.

Conclusion and clinical impact

The ACL SNNAP trial showed that, while benefit can be obtained from initial non-surgical treatment (with reconstruction performed only if required), treatment with more immediate surgical reconstruction of the ACL without further rehabilitation provides superior outcome. The null hypothesis was rejected, and the effect size observed for this surgical trial can be considered unusual for a surgical trial exploring true uncertainty.

The superiority of surgical management was shown in self-reported outcome of function and pain, complication rate and all secondary outcomes, including activity level. Interestingly neither group had good evidence of returning to their pre-injury level (28% surgery and 24% rehabilitation). In terms of patient satisfaction, 83% of patients in the surgery group reported they were better than before treatment compared to only 68% in the rehabilitation group. Eighty per cent of patients allocated to the surgical management would have the treatment and management strategy again compared to 61% in the rehabilitation group. Eighteen per cent of patients undergoing initial non-surgical management would not opt for that treatment again. Immediate ACLR management, despite generating higher costs, was shown to be cost-effective in the NHS.

The clinical implications are clear and important. In a shared decision-making setting with patients who have injured their ACL previously and are no longer acute, it should be communicated that ACLR (as soon as possible to stabilise the joint) is likely to provide superior outcome compared to attempting a rehabilitation strategy with subsequent reconstruction, if and when, needed. Patients who do not want surgery (for any reason) should be reassured that they can still improve with non-operative care and the option for later surgical reconstruction remains open.

Pathways being provided for ACL injury should incorporate this evidence and with practice change where appropriate. Sufficient availability and resourcing of ACLR for the non-acute population are warranted. Non-operative management and access to physiotherapy should still remain available, but it may be that efforts for progressing rehabilitation of ACL injury should be directed to postoperative care and return to sport/activity after surgery.

Contributions of authors

David J Beard (https://orcid.org/0000-0001-7884-6389) (Professor of Musculoskeletal and Surgical Science) was chief investigator; conceived and planned the study; acted as study lead; led the writing of the report and all aspects of study design, conduct and analysis.

Loretta Davies (https://orcid.org/0000-0002-4721-356X) (Senior Trial Manager, Clinical Trials Unit) contributed to conception, design and conduct; was project manager; and contributed to the writing, editing and review of the report.

Jonathan A Cook (https://orcid.org/0000-0002-4156-6989) (Senior Statistician, Clinical Trials Unit) contributed to the conception, design and conduct; oversaw the statistical analyses, and contributed to the writing, editing and review of the report.

Jamie Stokes (https://orcid.org/0000-0002-5279-2332) (Trial Statistician, Clinical Trials Unit) was the trial statistician, led the design of the statistical analysis plan, undertook the statistical analyses and contributed to the writing of the report.

Jose Leal (https://orcid.org/0000-0001-7870-6730) (Senior Health Economist) provided expertise in health economics, co-designed the health economics analysis, carried out the analysis of the health economics data and contributed to the writing and review of the report.

Heidi Fletcher (https://orcid.org/0000-0002-6518-6024) (Data Manager, Clinical Trials Unit) contributed to the design and management of the trial, data analysis and review of the report.

Simon Abram (https://orcid.org/0000-0002-4452-6499) (Clinical Lecturer in Orthopaedic Surgery) contributed to the design and management of the trial and review of the report.

Katie Chegwin (https://orcid.org/0000-0002-8671-5976) (Trial Administrator) contributed to the management of the trial and review of the report.

Akiko Greshon (https://orcid.org/0000-0002-8006-2943) (Data Management Assistant) contributed to the management of the trial and review of the report.

William Jackson (https://orcid.org/0000-0002-1851-5921) (Consultant in Trauma and Orthopaedic Surgery, Oxford University Hospitals) was a trial co-investigator and contributed to the study design and review of the report.

Nicholas Bottomley (https://orcid.org/0000-0001-8441-630X) (Clinical Lead Consultant in Trauma and Orthopaedic Surgery, Oxford University Hospitals) was the principal investigator for one of the recruiting sites; and contributed to the review of the report.

Matthew Dodd (https://orcid.org/0000-0002-8229-8175) (Consultant in Orthopaedic Surgery, Swansea University Hospitals) was the principal investigator for one the recruiting sites; and contributed to the review of the report.

Henry Bourke (https://orcid.org/0000-0003-3341-8545) (Consultant in Orthopaedic Surgery) was principal investigator for one of the recruiting sites; and contributed to the review of the report.

Beverly A Shirkey (https://orcid.org/0000-0002-4347-6784) (Trial Statistician) was the trial statistician, contributed to the design and conduct and review of the report.

Arsenio Paez (https://orcid.org/0000-0002-7984-514X) (Doctoral Student) contributed to the analysis and interpretation of the compliance data and the write-up and review of the report.

Sarah E Lamb (https://orcid.org/0000-0003-4349-7195) (Professor of Rehabilitation) was a trial co-investigator and contributed to the study design and review of the report.

Karen L Barker (https://orcid.org/0000-0001-9363-0383) (Professor of Physiotherapy) was a trial co-investigator and contributed to study design and review of the report.

Michael Phillips (https://orcid.org/0000-0002-6381-0727) (Founder, Director of PRO-MAPP Ltd) contributed to the study design, data collection processes and review of the report.

Mark Brown (https://orcid.org/0000-0002-9618-2480) (Programming PRO-MAPP Ltd) contributed to the study design, data collection processes and review of the report.

Vanessa Lythe (https://orcid.org/0000-0002-6372-2743) (Health Economics) provided expertise in health economics, carried out the analysis of the health economics data and contributed to the writing of the report.

Burhan Mirza (https://orcid.org/0000-0003-0369-7707) (Health Economics) provided expertise in health economics, carried out the analysis of the health economics data and contributed to the writing of the report.

Andrew Carr (https://orcid.org/0000-0001-5940-1464) (Nuffield Professor of Orthopaedics) was a trial co-investigator and contributed to the study design and review of the report.

Paul Monk (https://orcid.org/0000-0002-1673-2148) (Clinical Lecturer in Orthopaedics) was a trial co-investigator and contributed to the study design and review of the report.

Carlos Morgado Areia (https://orcid.org/0000-0002-4668-7069) (Trial Manager, Clinical Trials Unit) contributed to the design and management of the trial and review of the report.

Sean O’Leary (https://orcid.org/0000-0002-5650-9768) (Consultant in Orthopaedics) was a trial co-investigator, site lead for Reading and contributed to the study design and review of the report.

Fares Haddad (https://orcid.org/0000-0002-6272-0156) (Consultant in Orthopaedics) was a trial co-investigator, site lead for King’s College, London, and contributed to the study design and review of the report.

Chris Wilson (https://orcid.org/0000-0001-5730-0651) (Consultant in Orthopaedics) was a trial co-investigator and contributed to the study design and review of the report.

Andrew Price (https://orcid.org/0000-0002-4258-5866) (Consultant in Orthopaedics) was a trial co-investigator contributed to conception, design and conduct, advised on clinical aspects of trial. Contributed to the writing, editing and critical review of the report.

All authors contributed to the design and development of the trial protocol. All authors read, gave input and approved the final manuscript.

Acknowledgements

ACL SNNAP study group

Trial Management Group

Chief investigator: David Beard.

Trial co-investigators: Karen Barker (Oxford), Helen Campbell (Oxford) (submission of funding application), Andrew Carr (Oxford), Jonathan Cook (Oxford), Loretta Davies (Oxford), Fares Haddad (London), William Jackson (Oxford), Sallie Lamb (Oxford then Exeter), Jose Leal (Oxford), Paul Monk (Oxford), Sean O’Leary (Reading), Andrew Price (Oxford), Chris Wilson (Cardiff).

Research fellows: Simon Abram, Arsenio Páez.

Trial management: Loretta Davies, Carlos Morgado Areia (until October 2018), Jackie Davies (until June 2016).

Trial administration: Katie Chegwin, Jiyang Li (until February 2018).

Data management: Heidi Fletcher, Akiko Greshon.

Database/programming management: Christina Bagg, Mark Brown, Michael Phillips (Fr3dom Limited).

Trial statisticians: Jamie Stokes, Beverly Shirkey (until April 2020).

Health economists: Jose Leal, Vanessa Lythe, Burhan Mirza.

Qualitative researcher: Francine Toye (Oxford).

The study was managed by the Surgical Intervention Trials Unit (SITU) in collaboration with the Oxford Clinical Trials Research Unit (OCTRU).

Data-sharing statement

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

Ethics statement

Favourable ethics opinion for the ACL SNNAP trial was given by the National Research Ethics Service, Oxfordshire Research Ethics Committee (REC) in October 2016 (16/SC/0502).

Information governance statement

The University of Oxford is committed to handling all personal information in line with the UK Data Protection Act (2018) and the General Data Protection Regulation (EU GDPR) 2016/679. Under the Data Protection legislation, The University of Oxford is the Data Controller, and you can find out more about how we handle personal data, including how to exercise your individual rights and the contact details for our Data Protection Officer here: https://compliance.web.ox.ac.uk/individual-rights

Disclosure of interests

Full disclosure of interests: Completed ICMJE forms for all authors, including all related interests, are available in the toolkit on the NIHR Journals Library report publication page at https://doi.org/10.3310/VDKB6009.

Primary conflicts of interest: David Beard holds a NIHR Senior Investigator award. Andrew Price was President of the British Association for Surgery of the Knee (BASK) during this period. William Jackson reports FH Ortho consulting fees and stock: Smith and Nephew. Fares Haddad reports grants, royalties/licenses and consulting fees from: Corin, MatOrtho Limited, Smith and Nephew, Inc., Stryker. Leadership roles: director ISEH; divisional director UCLH and EIC at Bone and Joint Journal.

Publications

Monk AP, Davies LJ, Hopewell S, Harris K, Beard DJ, Price AJ. Surgical versus conservative interventions for treating anterior cruciate ligament injuries. Cochrane Database Syst Rev 2016;4:CD011166.

Abram SGF, Judge A, Beard DJ, Price AJ. Rates of adverse outcomes and revision surgery after anterior cruciate ligament reconstruction: a study of 104,255 procedures using the National Hospital Episode Statistics Database for England, UK. Am J Sports Med 2019;47(11):2533–42

Davies LJ. Design Issues and Challenges in Clinical Trials with a Surgical and Non-surgical Comparison. Doctoral dissertation. Oxford: University of Oxford; 2019.

Davies L, Cook J, Leal J, Areia CM, Shirkey B, Jackson W, et al. Comparison of the clinical and cost effectiveness of two management strategies (rehabilitation versus surgical reconstruction) for non-acute anterior cruciate ligament (ACL) injury: study protocol for the ACL SNNAP randomised controlled trial. Trials 2020;21(1):405.

Stokes JR, Beard DJ, Davies L, Shirkey BA, Price A, Cook JA. ACL Surgery Necessity in Non-Acute Patients (ACL SNNAP): a statistical analysis plan for a randomised controlled trial. Trials 2022;23(1):389.

Beard DJ, Davies L, Cook JA, Stokes J, Leal J, Fletcher H, et al. ; ACL SNNAP Study Group. Rehabilitation versus surgical reconstruction for non-acute anterior cruciate ligament injury (ACL SNNAP): a pragmatic randomised controlled trial. Lancet 2022 Aug 20;400(10352):605–15. https://doi.org/10.1016/S0140-6736(22)01424-6. PMID: 35988569.

Paez A. Faithful but Flexible: Intervention Fidelity in Clinical Trials of Complex Interventions in Healthcare. Doctoral dissertation. Oxford: University of Oxford]; 2023.

Leal J, Mirza B, Davies L, Fletcher H, Stokes J, Cook JA, et al. Cost-effectiveness analysis of a pragmatic randomized trial evaluating surgical reconstruction versus rehabilitation in patients with long-standing anterior cruciate ligament injury. Bone Joint J 2024;106-B (1), 38–45. https://doi.org/10.1302/0301-620X.106B1.BJJ-2023-0175.R1

Disclaimers

This article 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.

Independent members of the Data Monitoring Committee

Richard Emsley (Chair) (King’s College London).

George Peat (Keele University).

Martyn Snow (Royal Orthopaedic Hospital, Birmingham).

Independent members of the Trial Steering Committee

Marion Campbell (Chair) (University of Aberdeen).

Tessa Howell (Patient Representative).

Hilary Johnson (Gloucestershire Hospitals NHS Foundation Trust).

Stephen McDonnell (Cambridge University Hospitals NHS Foundation Trust).

Thomas Pinkney (University of Birmingham).

Mark Williams (Oxford Brookes University).

Trial co-investigators

Helen Campbell, Andrew Carr, Jonathan Cook, Loretta Davies, Jose Leal, Paul Monk and Andrew Price (University of Oxford).

Sarah E Lamb (College of Medicine and Health, University of Exeter).

Fares Haddad (University College London Hospitals NHS Foundation Trust).

Karen Barker and William Jackson (Oxford University Hospitals NHS Foundation Trust).

Sean O’Leary (Royal Berkshire NHS Foundation Trust).

Chris Wilson (Cardiff and Vale University Health Board).

Research fellows: Simon Abram and Arsenio Páez (University of Oxford).

Trial management: Loretta Davies, Carlos Morgado Areia and Jackie Davies (University of Oxford).

Trial administration: Katie Chegwin and Jiyang Li (University of Oxford).

Data management: Heidi Fletcher and Akiko Greshon (University of Oxford).

Database/programming management: Christina Bagg, Mark Brown and Michael Phillips (Fr3dom Limited).

Trial statisticians: Jamie Stokes and Beverly Shirkey (until Apr 2020); Health economists: Jose Leal, Vanessa Lythe and Burhan Mirza (University of Oxford).

Qualitative researcher: Francine Toye (Oxford).

Principal investigators (PIs) and research teams:

Laura Haywood, Anne Nicholson, Joanne Riches, Sean Symons (PI) and Mark Vertue (Basildon and Thurrock University Hospitals NHS Foundation Trust).

Louay Al Mouazzen, Rachel Bray, Damian Clark, James Coulthard, Tim Holland, Nick Howells, Andrew Jones, Richard Kapur, Alastair Kiszely, Harry Krishnan, Karen MacDonald-Taylor, Jon Manara, James Murray (PI), Corina Negrut, Vishai Pai, Andrew Porteous, Sven Putnis, James Robinson, Shav Rupasinghe, Veenesh Selvaratnam, James Smith, Nick Smith, Jarrad Stevens, Clare Taylor, Anthony Theodorides, Niraj Vetharajan, Helen Vint and Lucy Young (North Bristol NHS Trust).

Susan Bullock, Rebecca Cook, Alexander Dodds (PI), Amanda Freeman-Hicks and Paula Hillout (Gloucestershire Hospitals NHS Foundation Trust).

Thomas Cornell, Abbie Coutts, Suzy Dean, Nicki Devooght-Johnson, Emma Ferrell, Eve Fletcher, Chrissie Hall, Benjamin Kent, Sandra Kessly, Robin Kincaid (PI), Mohamed Lazizi, Ahmed Mostafa, Toby Nisbett, Tim Powell, Peter Riddlestone, Andrew Roberton, Jessica Summers, Lucy Whitbread and Belinda Wroath (Royal Cornwall Hospitals NHS Trust).

Emma Fenlon, Andrew Hall, Helen Jeffrey and Raghuram Thonse (PI) (Countess of Chester Hospital NHS Foundation Trust).

Debra Dunne, Andy Metcalfe (PI), Kerri McGowan, Simon Middleton, Feisal Shah, Tim Spalding, Charlie-Marie Suddens, Tamar Sweed, Joanna Teuke, Peter Thompson and David Wright (University Hospitals Coventry and Warwickshire NHS Trust).

Justine Amero, Emma Brown, Hugh Chissell, Andrea Croucher, Gareth Dickinson, Catherine Hawkes-Blackburn, Alice Peacocke, Graham Smith (PI) and Carol Snipe (Frimley Health NHS Foundation Trust).

Kim Dearnley and Reza Mayahi (PI) (Hull and East Yorkshire Hospitals NHS Trust).

Barry Andrews (PI), Massimo Barcelona (PI), Hazel Giles, Abdulkerim Gokturk, Paul Harnett, Katie Jeeves, Joyce Kadunyi, Sheena Mendoza, Ines Reichert, Marta Santamaria and Harshinder Virdee (Kings College Hospital NHS Foundation Trust).

Sanjeev Anand (PI), Nayef Aslam-Pervez, Stephen Draycott, Faye Howarth, Irfan Jina, Niall Maher, Denise Ross (PI) and Lindsey Worstenholme (Leeds Teaching Hospitals NHS Trust).

Abdul Baig, Arun Bhaskaran, Daniel Banks, Tracy Brear, Carla Christie, Laura Cowen, Jack Davis, Ross Dixey, Colin Esler (PI), Amirah Essop-Adam, Christina Haines, Linzy Houchen-Wolloff, Husein Varachia and Richard Wood (University Hospitals of Leicester).

Glaxy Gray, Jessica Nichols, Alice Panes, Susan Partridge, Lawrie Rogerson, Pankaj Sharma (PI), David Triggs, Ian Venables and Danielle Wilcock (Manchester University NHS Foundation Trust).

Sarah Buckley, Thelma Darian, Elizabeth Denis, Jo Duncan, Charlotte Hirst, James Newman, Fern Richardson and Jon Smith (PI) (The Mid Yorkshire Hospitals NHS Trust).

Megan Adcode, Megan Cottingham, Eliza Foster, Andrew Kelly (PI), Niamh McKay, Jane Rewbury, Alison Whitcher, James Williams and Esther Zebracki (Taunton and Somerset NHS Foundation Trust, Musgrove).

Llinos Davies, Jayadeep Jayachandran (PI), Alison Tardivel and Victoria Whitehead (Betsi Cadwaladr University Health Board, North Wales).

Simon Abram, Carlos Morgado Areia, Martha Batting, Amy Bond, Nick Bottomley (PI), Loretta Davies, Marc Deakin, Christopher Dodd, Alison Hudak, Samantha Hynes, Will Jackson, Luke Jones, Gail Lang, David McKenna, Susan Morris, Andrew Price, Clare Scott-Dempster, Adam Sykes, Iason Vichos and Simon Wood (Oxford University Hospitals NHS Foundation Trust).

Rupert Clifton (PI), Stephanie Diaz, Craig Hendy, Nishil Modi, Brendan O’Mahony, Susan O’Sullivan, Nicola Parker, Mira Pecheva and Rowan Rumonovic (North West Anglia NHS Foundation Trust, Peterborough).

Emma McLoughlin, Jeremy Rushbrook (PI) and Anna Thornhill (Solent NHS Trust, Portsmouth).

Emily Bannister, Chloe Brown, Debbie Burden, Terence Campbell, Emma Craig, Rashmi Easow, Julie Foxton, Alexandra Hazlerigg, Chethan Jayabev, Rosie Murdoch, Sean O’Leary (PI) and Georgie Parsons (Royal Berkshire NHS Foundation Trust).

Harry Brown, Paula Carvelli, Rugaia Montaser, Ali Pepper, Sinduja Sivarajan, Oliver Templeton-Ward (PI) and Eva Wilson (Royal Surrey County Hospitals NHS Foundation Trust).

Julie Cronin, Sarah Diment, Victoria King, Katharine Shean, Leonidas Vachtsevanos (PI), Katharine Wilcocks and Ben Wilson (Salisbury NHS Foundation Trust).

Paul McNestry, Joanna Ollerenshaw (PI), James Stoddard and Paul Sutton (Sheffield Teaching Hospitals NHS Foundation Trust).

Sanjay Anand, Judith Bell, Albert Chikate, Diane Daniel, Timothy Davies, Tom Finnigan, Antonio Frasquet-Garcia, Susan Hopkins, Sharon Kerrison, Angela McGowan, David Sands Johnson (PI), Lara Smith, Philip Turner and Helen Wilkinson (Stockport NHS Foundation Trust).

Lynne Allsop, Deborah Anthony, Rebecca Boulton, Sarah Brown, Vikram Desai (PI), Mandy Gill, Cheryl Heeley, Sushrut Kulkarni, Wayne Lovegrove, Dominic Nash, Terri-Ann Sewell, Sarah Shelton and Katie Slack (Sherwood Forest Hospitals NHS Foundation Trust, Sutton in Ashfield).

David Beard, James Cartwright, Lynda Connor, Andrew Davies, Caroline Davies, Matthew Dodd (PI), Glyn Gainard, Dave Graham-Woollard, Carl Murphy, Leanne Quinn, Caradog Thomas, Jenny Travers and Marie Williams (Swansea Bay University Health Board, Swansea).

Amanda Bell, Sunny Deo, Katharine Francis, Tracy Jackson, Laura McCafferty, Basalingappa Navadgi, Karan Plank, Venkat Satish (PI) and Claire Thelwall (Great Western Hospitals NHS Foundation Trust, Swindon).

Fares Haddad (PI), Rachel Knight, Rahul Patel and Bruce Paton (University College London Hospitals NHS Foundation Trust).

Ashutosh Acharya, Utuman Aland, Miltiades Areirobulos, Pascal de Feyter, Lisa Ditchfield, Hafiz Iqbaz, Daniel Massey and Gareth Stables (PI) (Warrington and Halton Hospitals NHS Foundation Trust).

Sarah Appleby, Henry Bourke (PI), Michael Brown, Sarah Cable, Alexander Damen, Joana Da Rocha, Louise Foster, Elizabeth Hamilton, Catriona Hatton, Cassie Honeywell, Kunal Kulkarni, Lucy Markham, Haadiya Mohammed and John O’Grady (Frimley Health NHS Foundation Trust, Wexham).

Yogesh Joshi (PI), Heather Mclintock, Tania Morgan, Jane Stockport and Victoria Whitehead (Betsi Cadwaladr University Health Board, Wrexham).

Pranshu Agrawal, Jo Armstrong, Shannon Briggs, Ben Coupe (PI), Anne Evans, Rob Gilbert, Sandra Latham, Aslam Mohammed, Valerie Parkinson, Valerie Parkinson, Rafael Sales, Katja Van De Snepscheut-Jones, David Wilcock and Daniel Wright (Wrightington, Wigan and Leigh NHS Foundation Trust).

Joanna Allison, Simon Baker, Kate Beesley, Gill Ferrari, Benedict Lankester (PI), Alison Lewis, Joanne Lyons, Jamie O’Callaghan, Sarah Sutcliffe and Dianne Wood (Yeovil District Hospital NHS Foundation Trust).