Multi-centre randomised controlled trial with economic evaluation and nested qualitative study comparing early structured physiotherapy versus manipulation under anaesthesia versus arthroscopic capsular release for patients referred to secondary care with a frozen shoulder (Adhesive Capsulitis)

Design: Randomised controlled trial with economic evaluation and nested qualitative study, with an internal pilot to check assumptions on trial recruitment and feasibility. Setting: We need to recruit from 25 NHS Hospitals in the United Kingdom. Target population: Patients, including diabetics, to be included if they are aged 18 years or older, with clinical diagnosis of frozen shoulder characterised by restriction of passive external rotation in the affected shoulder to less than 50% of the contralateral shoulder, and have radiographs that exclude glenohumeral arthritis. This is a recommended consensus definition for diagnosis of frozen shoulder that can be applied across centres [1]. We will exclude patients with bilateral concurrent frozen shoulders, those secondary to trauma (i.e. trauma to the shoulder that required hospital care) or secondary to other causes (e.g. recent breast surgery), those for which any of the trial treatments are contraindicated, and those unfit for general anaesthesia. Health technologies being assessed: York Trials Unit will perform unequal random allocation (1:2:2), to allow for the potential difference in effect between treatment comparisons, stratified by presence of diabetes [2] of the following interventions: 1) Early structured physiotherapy (ESP) for 12 weeks ('control') [3]. 2) Manipulation under anaesthesia (MUA) without Arthroscopic capsular release (ACR)within 18 weeks from randomisation. 3) MUA with ACR within 18 weeks from randomisation. We have chosen to evaluate these interventions for the following reasons: a)MUA and ACR are the most commonly used surgical interventions in secondary care; they are both costly, with ACR costing twice as much as MUA. b)There is no randomised evidence on MUA or ACR to indicate superiority of one over the other. c)It is important to have a control arm, as the benefits & risks of MUA & ACR have not yet been established. ESP based on national clinical physiotherapy guidelines with active monitoring is our proposed control, as physiotherapy is a commonly provided conservative treatment and will also provide the opportunity to evaluate the national physiotherapy guidelines in secondary care. For those patients who do not improve with ESP they will be referred for further treatment in consultation with the orthopaedic surgeon at a 12 week assessment. When further treatment after ESP involves surgical intervention (e.g. MUA or ACR), patients will be placed on the normal waiting list for such treatment. Any further treatment provided will be recorded. Based on a survey of potential PIs for this study, we will standardise the surgical interventions as follows: MUA without ACR will be supplemented with steroid injection (which is 80% standard practice according to the survey); for ACR there will be no steroid injection as the rare risk of infection from arthroscopy can be increased from the presence of steroid. MUA with ACR, where arthroscopic rotator interval and anterior capsular release will be performed followed by MUA to complete the release. Additional procedures like posterior capsular release, subacromial decompression and supplementary steroid injections for pain control occur in a minority of patients and will be left pragmatic as a ‘needs based’ intervention. Any additional interventions will be recorded. Post procedural physiotherapy will be standardised by consensus agreement between centres. Measurement of costs and outcomes: Analysis will be conducted in accordance with CONSORT. Our primary outcome is the Oxford Shoulder Score (OSS) at baseline, 3, 6, and 12 months after randomisation. Twelve months after randomisation is the primary end-point. We believe these time points are essential to assess rate of recovery and to not miss important differences. We will estimate the proportion of patients achieving the Minimal Clinically Important Difference (MCID) for OSS at the fixed study time points. In addition, the Disabilities of Arm, Shoulder and Hand (QuickDASH), EQ5D-5L, and Numerical Rating System pain questions will be collected at those time points. Time to resolution of symptoms will be recorded using a simple global question administered at the fixed study time points and has been defined in consultation with patients undergoing therapy for frozen shoulder. Complications will be recorded at 12 months. Adverse events will be collected routinely for 12 months. An economic evaluation will be conducted to assess the most cost-effective management option. EQ5D will be used to generate within trial estimates of changes in health status and in accordance with NICE guidance health benefits will be expressed as quality adjusted life years (QALYs). Health care resource use will be collected prospectively from hospital forms and patient questionnaires. The analysis will take an NHS and Personal Social Services perspective consistent with methods used by bodies such as NICE. Results will be summarised in an incremental cost effectiveness ratio. To explore uncertainty around adoption decision a probabilistic sensitivity analysis will be used to quantify joint uncertainty around costs and QALYs. Sample size: The primary outcome is the OSS and will be assessed for three treatment comparisons: ESP compared with MUA; ESP compared with ACR where for both of these comparisons we are testing for a 5 point difference on the OSS; and MUA versus ACR where we are testing for a 4 point difference. There is data to suggest a 5 point improvement can be found on the OSS (effect size of 0.42) between surgically and conservatively treated patients, with a stable standard deviation of 12 points across different populations. This larger effect size will be required to justify the greater costs and potential risks associated with surgery. A smaller difference of 4 points on the OSS (effect size of 0.33) is expected to distinguish between MUA and ACR. To observe the above effect sizes with 90% power and 5% two-sided significance, adjusting for a conservative estimate (r=0.4) of the correlation between OSS over 12 months and allowing for 20% attrition, a total sample size of 500 patients is required (ESP: 100; MUA: 200; ACR: 200). Owing to the a priori specified sequence of treatment comparisons, multiplicity should not be a concern. Thus no adjustments are made to the calculation. Project timetables & recruitment rate: Please see research plan section for further detail.