Subacromial decompression versus diagnostic arthroscopy for shoulder impingement: randomised, placebo surgery controlled clinical trial

Abstract Objective To assess the efficacy of arthroscopic subacromial decompression (ASD) by comparing it with diagnostic arthroscopy, a placebo surgical intervention, and with a non-operative alternative, exercise therapy, in a more pragmatic setting. Design Multicentre, three group, randomised, double blind, sham controlled trial. Setting Orthopaedic departments at three public hospitals in Finland. Participants 210 patients with symptoms consistent with shoulder impingement syndrome, enrolled from 1 February 2005 with two year follow-up completed by 25 June 2015. Interventions ASD, diagnostic arthroscopy (placebo control), and exercise therapy. Main outcome measures Shoulder pain at rest and on arm activity (visual analogue scale (VAS) from 0 to 100, with 0 denoting no pain), at 24 months. The threshold for minimal clinically important difference was set at 15. Results In the primary intention to treat analysis (ASD versus diagnostic arthroscopy), no clinically relevant between group differences were seen in the two primary outcomes at 24 months (mean change for ASD 36.0 at rest and 55.4 on activity; for diagnostic arthroscopy 31.4 at rest and 47.5 on activity). The observed mean difference between groups (ASD minus diagnostic arthroscopy) in pain VAS were −4.6 (95% confidence interval −11.3 to 2.1) points (P=0.18) at rest and −9.0 (−18.1 to 0.2) points (P=0.054) on arm activity. No between group differences were seen between the ASD and diagnostic arthroscopy groups in the secondary outcomes or adverse events. In the secondary comparison (ASD versus exercise therapy), statistically significant differences were found in favour of ASD in the two primary outcomes at 24 months in both VAS at rest (−7.5, −14.0 to −1.0, points; P=0.023) and VAS on arm activity (−12.0, −20.9 to −3.2, points; P=0.008), but the mean differences between groups did not exceed the pre-specified minimal clinically important difference. Of note, this ASD versus exercise therapy comparison is not only confounded by lack of blinding but also likely to be biased in favour of ASD owing to the selective removal of patients with likely poor outcome from the ASD group, without comparable exclusions from the exercise therapy group. Conclusions In this controlled trial involving patients with a shoulder impingement syndrome, arthroscopic subacromial decompression provided no benefit over diagnostic arthroscopy at 24 months. Trial registration Clinicaltrials.gov NCT00428870.

This supplementary material has been provided by the authors to give readers additional information about their work.
Stand sideways against a wall with your upper arm close to your side and elbow at a right angle. Push the forearm to the side against the wall. Hold approx. 5 secs.
Stand in a doorway with you elbow close to your body and bent at a right angle. Place your hand against the wall. Push your hand inwards against the wall. Hold 5 secs. Relax.

Repeat 3 x 3 times.
Stand with your upper arm close to your side, elbow at a right angle and the back of your hand against a wall. Push the back of your hand against the wall. Hold approx. 5secs.

Phase II (4-5 weeks)
© PhysioTools Ltd Sit. Place your hand on the front of the opposite shoulder. Feel your shoulder with your fingers making sure that the shoulder does not come forward. Move your shoulder gently 1 cm back and 2 cm up. Later on, exercise is done without palpation or visual confirmation of movement.
Sit on a chair with your arm supported on a table.
With your other hand push the top of your upper arm downwards. Hold approx. 2 secs.
Sit. Place your hand on the front of the opposite shoulder. Feel your shoulder with your fingers. Move your shoulder gently 1 cm forward and Stand with your back against the wall. Keep your upper arm close to the side and elbow at a right angle. Push the elbow back against the wall and hold for 5 secs.
Stand with your upper arm close to your side, elbow at a right angle and the back of your hand against a wall. Push the back of your hand against the wall. Hold approx. 5 secs.
Stand in a doorway with you elbow close to your body and bent at a right angle. Place your hand against the wall. Push your hand inwards against the wall. Hold 5 secs.

Phase III (6-8 weeks)
Sit or stand holding on to a rubber exercise band with both hands. Pull the band with both arms pushing the shoulder blades together.
Stand or sit. Place a rolled towel between your upper arm and side. Your upper arm should be slightly forward and your elbow at a right angle. Hold an exercise band. Pull the band taking your forearm out 45 degrees. Hold for 5 secs.
Stand straight holding an exercise band. Move your arm forward, pulling the band. Slowly return.
Stand with one foot forward. Keep your back straight. Place the palm of your hand against a wall using a ball or a piece of cloth. Lean your body weight onto your hand. Control your shoulder position and move hand in small horisontal motions.
Hold an exercise band with both hands. Pull the band outwards to about 45-60 degrees with back of your hand leading the movement. Hold for 5 secs.

Repeat 3 x 15-25 times
Stand keeping your upper arm close to the side and elbow at a right angle. Hold a rubber exercise band. Pull the band toward your stomach and hold for 5 secs.

Phase IV (9-12 weeks)
Stand keeping your upper arm close to the side and elbow at a right angle. Hold a rubber exercise band. Pull the band toward your stomach.
Stand or sit with the hand of the arm to be exercised on your opposite hip. Hold on to a rubber exercise band. Pull the band up towards the opposite side.
Stand with your arm up and out to the side. Hold a rubber exercise band. Pull the band down and across your body letting your thumb lead the movement.
Alternatively, Sit on a chair with your arm lifted to the side and your elbow at a right angle supported on a table. Hold on to a rubber exercise band which is fastened behind you. Pull the band keeping elbow bent and resting on the table.
Alternatively, Sit on a chair with your arm lifted out to the side and elbow at a right angle supported on a table. Hold on to a rubber exercise band which is in front on you. Pull the band keeping your elbow bent and resting on the table.
Stand keeping your upper arm close to the side and elbow at a right angle. Hold a rubber exercise band. Pull the band by turning your forearm outwards.

Patient Satisfaction and the Responder Analysis
Global assessment of satisfaction with the treatment How satisfied are you with the treatment given? Place a vertical mark to the line below to indicate your current satisfaction to the treatment given.
Completely dissatisfied ___________________________________________________Very satisfied

Satisfaction with the treatment outcome
How satisfied are you with the treatment outcome? Mark the answer closest to your situation.
1. Very satisfied, my shoulder has healed completely.
2. Satisfied, I have only minor, activity related symptoms. My shoulder is much better than before treatment.
3. Somewhat satisfied, I have only minor symptoms. My shoulder is better than before treatment.
4. Dissatisfied, my shoulder is the same as before treatment.

Responder analysis
Instead of focusing only on the statistical significance of the mean differences between treatment groups in the VAS (i.e., the mean improvement from baseline to 24 months), "a responder analysis" was also carried out. We anchored our responder analysis to the patient's assessment of satisfaction with the shoulder treatment outcome: Patients reporting very satisfied or satisfied were categorised as "Responders".

STUDY SYNOPSIS
Introduction: Arthroscopic subacromial decompression (ASD) is the most commonly performed surgical intervention for shoulder pain, yet evidence on its efficacy is limited. The rationale for the surgery rests on the tenet that symptom relief is achieved through removal of a bony acromial spur and the resulting decompression of the tendon passage. Acknowledging the potential placebo effect of surgery, the primary objective of this superiority trial is to compare the efficacy of ASD versus diagnostic arthroscopy (DA) in patients with shoulder impingement syndrome (SIS), where DA differs only by the lack of subacromial decompression. As a non-surgical treatment option, a third group of supervised progressive exercise therapy (ET) will allow for pragmatic assessment of the relative benefits of surgical vs. non-operative treatment strategies. Methods/Design: FIMPACT trial is an ongoing multicentre, three-group randomised controlled study with a primary objective of assessing the efficacy of the ASD vs. DA and a secondary objective of comparing ASD to exercise therapy (ET) in a pragmatic setting. We performed two-fold concealed allocation, first by randomising patients to surgical (ASD or DA) or conservative (ET) treatment in 2:1 ratio and then those allocated to surgery further to ASD or DA in 1:1 ratio. Our two primary outcomes are pain at rest and arm activity assessed with visual analog scale (VAS), while the secondary outcomes are functional assessment (Constant score and Simple shoulder test), quality of life (15D and SF-36), patient satisfaction, proportions of responders and non-responders, reoperations/treatment conversions, all at 2 years post-randomisation, as well as adverse effects and complications. We recruited a total of 210 patients from 3 tertiary referral centres. We will conduct the primary analysis on the intention-to-treat basis.

STUDY OBJECTIVES AND OUTCOMES
This statistical analysis plan (SAP) is accompanying the actual study protocol of the FIMPACT trial, a document that elaborates the methods used in detail. All outcomes were inquired from participants at baseline and follow-ups (6 and 24 months) and selected additional measures at 3 and 12 months (for details, see Table 1). The last patient reached the primary endpoint, the 24-month follow-up, in September 2015.

DESCRIPTIVE OUTCOMES
At screening, the participants filled out a questionnaire to record gender, age, hand dominance, weight, height, level of education (socioeconomic status), workload (type of work), physical activity level, sports discipline, subjective health, symptoms (onset, frequency, and severity), use of pain medications, prior treatments, expectations to treatment, generic health state, and disease-specific scores. To exclude patients with concomitant shoulder pathology (particularly rotator cuff rupture), magnetic resonance imaging with contrast (MRA) was acquired for each participant.

OBJECTIVES AND PRIMARY OUTCOME
The primary objective of this trial is to compare the efficacy of arthroscopic subacromial decompression (ASD) versus diagnostic arthroscopy (DA) in patients with SIS. The trial is designed as a superiority trial, i.e. we expected in the power calculation that the ASD will result in greater pain relief at 24-month follow-up than DA (or ET). The 24-month follow-up was chosen as the primary endpoint, since this time point is a commonly held "minimal requirement" for any procedure in the field (orthopaedics) and most commonly used in the trials assessing the treatment of SIS.
The primary hypothesis: The primary hypothesis of our FIMPACT trial is that ASD is superior to DA in patients with SIS.
To enable pragmatic assessment of the relative benefits of surgical vs. non-operative treatment strategies on SIS, a non-surgical (third) treatment option of supervised progressive exercise therapy (ET) is also included (ASD vs. ET).
Additional hypothesis: The relative benefits of ASD and ET will be assessed without a priori hypothesis on the superiority of one or the other.
As the primary outcome measure, a visual analogue scale (0-100) was used to measure the patient's perceived pain intensity at rest and at arm activity during the 24 hours preceding the assessment. We considered 15 as the minimal clinically important difference (MCID) for SIS. 1

SECONDARY OUTCOMES
Our secondary outcome measures are listed below. These outcomes will only be supportive, explanatory and/or hypothesis generating, which is why multiplicity is not considered to be a problem 2 .

Constant-Murley score
Constant-Murley score (CS) is the most commonly used scoring system for evaluation of various disorders of the shoulder 3 . It consists of both objective (range of motion and strength) and subjective measurements (pain assessment, work load, and leisure time activities), which are summarised in a score between 0 and 100. A higher score indicates better shoulder function. The minimal clinically important improvement (MCII) of the Constant score is 17 for patients with SIS 4 .
In addition, as night pain is considered one of the hallmark symptoms in patients with SIS and our two primary outcome measures (patient's perceived pain intensity at rest and at arm activity in the last 24 hours) do not specifically address this issue, a specific question from the Constant-Murley score (unaffected sleep: "Yes" or "No") will be analysed separately.

SST
The simple shoulder test (SST) was developed to assess any impairment of the patient's activities of daily living 5 . The SST consists of 12 questions with yes (1) or no (0) response options. The maximum SST score is 12 indicating normal shoulder function, minimum score of 0 points refers severely diminished shoulder function. The SST has good reliability and responsiveness in patients with rotator cuff symptoms 6 . The MCID for the SST in rotator cuff disease is 2 points 7 .

15D
The 15D instrument is a generic health-related quality of life (HRQoL) instrument comprising 15 dimensions 8 . For each dimension, the respondent must choose one of the five levels that best describes his/her state of health at that the moment (the best level being 1 and the worst level being 5). A set of utility or preference weights is used in an addition aggregate formula to generate a single index number, the utility or 15D score. The maximum 15D score is 1 (no problems on any dimension) and the minimum score is 0 (being dead). The responsiveness, reliability and validity of 15D have been thoroughly established, and this instrument has been used extensively in clinical and healthcare research 9 10 .

SF-36
Short form or SF-36 is a generic HRQoL instrument to quantify the physical, functional, and psychological aspects of healthrelated quality of life. It consists of 36 questions in eight subscales that assess physical, functional, social, and psychological well-being 11 . Score ranges from 0 to 100, where a higher score is associated with better health. The physical and mental component summary scales (PCS and MCS, respectively) are then calculated as composites of the related subscales. SF-36 is one of most widely used measure of health-related quality of life 12 .

Patient satisfaction and Responder analysis
We elicited patients' global assessment of satisfaction to the treatment with this question: "Are you satisfied with the treatment you have received?" We used a VAS scale ranging from 0 (completely disappointed) to 100 (completely satisfied).
Additionally, we elicited patient satisfaction to the treatment outcome with the following question at each follow-up time point (Table 1): "How satisfied are you with the outcome of your treatment?" on a 5-item scale. Participants who reported very satisfied or satisfied will be categorised as "Responders" and patients who responded very dissatisfied or dissatisfied as "Nonresponders".

Return to previous leisure activities
Similarly, at each follow-up (Table 1), participants were asked to respond to the following question: "Have you been able to return to your previous leisure activities?" ("yes" or "no").

Patients' perception of operative treatment-group assignment
At the 3-month follow-up point, the patients in the two operative groups were asked to guess whether they had undergone ASD or DA.

Health resource utilisation and costs
For the cost-effectiveness analysis, at each follow-up visit the participants were asked to fill in a questionnaire inquiring about the use of healthcare resources. The questionnaire contains a list of items of healthcare resources available and the participants were asked to fill in the number of visits per item during the recall period of each follow-up time point. The resource use will be calculated based on the number of visits times unit cost per item and expressed as mean costs by items of resource use, and the mean direct total health care resource costs. All costs will be discounted to the 2016 price level.

Time to return to work
Information about return to work was recorded at each follow-up time point (Table 1).

Complications and adverse effects
Complications directly related to the interventions were registered. The participants were also encouraged to contact the participating hospitals if any adverse effects occurred and contacts to the health care system were monitored at every followup visit. Potential adverse effects (AE) were categorised to serious ad-verse effects (SAE) and minor adverse effects (MAE) if the participants sought treatment. Death, cardio-vascular or gastrointestinal effects, deep venous thrombosis, pulmonary embolism, systemic or local infection were categorised as SAEs and shoulder symptoms like pain, swelling and decreased range of motion were categorised as MAEs. The number and severity of complications and adverse effects will be assessed.

EXPLORATORY OUTCOMES
We have identified three potentially important effect modifying factors. We will perform subgroup analyses with the primary endpoint as the outcome and the direction of hypothesised effect described as below 14 .

Duration of symptoms
We will compare the treatment effects stratified based on the duration of symptoms (those with < 6/12 months vs. those > 6/12 months). We hypothesise that subacromial decompression will work better in patients with duration of symptoms > 6 months than for patients with symptoms < 6 months.

Severity of symptoms
We will compare the treatment effects in patients with severe (VAS 70 or more), moderate (VAS 55 to 69), and mild (VAS less than 55) symptoms at baseline. We hypothesise that subacromial decompression will work better in patients with more severe (VAS 70 or more) than moderate (VAS 55 to 69) or mild (VAS less than 55) symptoms at baseline.

Acromial anatomy
We will compare the treatment effects in patients with flat (type I), curved (type II), or hooked (type III) acromion according to classification by Bigliani et al. 15 We hypothesise that subacromial decompression will work better in patients with hooked (type III) than curved (type II) or flat (type I) acromion at baseline.

STUDY DESIGN Sample size
The sample size calculation was based on the two primary outcome measures, VAS at rest and at arm activity, at 24 months post randomisation. FIMPACT trial was powered to detect a minimal clinically important improvement (MCII) in a VAS pain score (improvement of at least 15; assumed standard deviation 25) between ASD and DA (or ET). To achieve a somewhat unconventional (stringent) 90% study power and using a two-sided Type I error rate (5%), our trial requires 68 patients per study group to show clinically meaningful advantage of ASD over DA (or ET). Acknowledging the stringent power threshold, only 3% surplus was reserved for potential loss to follow up/crossovers (3%), and accordingly, the recruitment target was set at 70 patients per treatment group.

Randomisation and blinding
To obtain three balanced study groups (of similar group size), we performed a two-fold, sequential randomisation. In Phase I, the participants were randomised into non-surgical or surgical treatment with allocation ratio 1:2. In the Phase II, those allocated to surgical treatment were further randomised to ASD or DA with 1:1 ratio. An independent statistician with no clinical involvement in the execution of the trial prepared separate randomisation lists for each study centre using a computer-generated algorithm. Randomisation was carried out using sequentially numbered sealed opaque envelopes. The envelopes were kept in a secure, agreed location at each centre. To ensure concealment, block randomisation was applied using blocks varying in size randomly, the block size known only by the statistician. To initially enter a participant into the study (Phase I), an envelope containing the treatment assignment [non-surgical (ET) or surgery (ASD or DA), ratio 1:2] was opened during the baseline appointment. Participants randomised to ET started standardised physiotherapy within 2 weeks of the baseline appointment. Participants allocated to surgical treatment were scheduled for surgery aimed to be completed within 12 weeks of randomization. At the day of surgery, an arthroscopic examination was first carried out to confirm the eligibility of the participant (to rule out full-thickness RC tear and other obvious intra-articular pathology). Research/staff nurse then completed the randomisation procedure (Phase II) by opening an envelope containing the surgical treatment allocation (ASD or DA, ratio 1:1). The allocation was revealed to the surgeon by showing the paper, but not expressed verbally.
The full follow-up process is shown in figure 1. In brief, the participants filled in the above noted (mailed) outcome questionnaires at 3, 6, 12 and 24 months post randomisation, in addition to which they were also assessed clinically at 6 and 24 months (and 5 and 10 years) post randomisation by a study physiotherapist unaware of treatment allocation, treatment given or possible unblinding. Outcome assessors were instructed not to inquire anything about prior treatment. Further, participants wore a t-shirt on all follow-up examinations. Data analysis will be done in a blinded manner by the study statistician (JR) not directly involved in the study.

STUDY POPULATION Subject disposition
Study procedures, including recruitment strategies and inclusion and exclusion criteria, are presented in detail in the accompanying actual study protocol.

STATISTICAL ANALYSIS
Data will be analysed in a blinded manner. All p-values will be reported to 3 decimal places with those less than 0.001 reported as p < .001. The criterion for statistical significance will be set at alpha = 0.05.

Primary analysis
The primary analysis will be carried out according to the intention-to-treat (ITT) principle: participants are retained in the groups to which they were initially randomised. The primary comparison on the efficacy of ASD (ASD vs. DA) will be performed as a between-group comparison using a repeated measures mixed-effects model (RMMM). Study group and time of assessment (baseline, 3, 6, 12 and 24 months) will be included as fixed factors and patient as a random factor. The model will include interactions between study group and time of assessment. The baseline value will be included as a covariate. An unstructured covariance structure will be assumed. If the model cannot be fitted, compound symmetry will be assumed instead. The number of degrees of freedom will be assessed using Satterthwaite's method. The RMMM model will be used to quantify the treatment effect as the difference between the groups in pain scores (VAS) with the associated 95% confidence interval (CI) and p-value at 24 months post-primary randomisation. To safeguard against potential multiplicity bias 2 , we will require a statistically significant treatment effect on both of our primary outcome variables, i.e., pain at rest and pain at activity ( Table  2). The same statistical model will also apply to the pragmatic comparison of the relative benefits of surgical vs. non-operative treatment strategies on SIS (ASD vs. ET) ( Table 3).

Secondary analyses
We will also use the RMMM model to analyse secondary outcomes (Table 2 and 3) where applicable. The results will be reported as the differences between the groups with the associated 95% confidence interval (CI) and p-value at 24 months postprimary randomisation. Furthermore, instead of focusing only on the statistical significance of the mean differences between treatment groups in the VAS (i.e., the mean improvement from baseline to 24 months), we will also carry out "a responder analysis". In principle, this analysis allows physicians to inform a patient of his or her chance of experiencing a clinically meaningful improvement from the treatment, both in absolute terms and in comparison, to a control group. The difference between responders and nonresponders can be considered the net-benefit of the treatment. One proposed means to carry out a responder analysis relies on the assessment of the proportion of patients reaching the patient-acceptable symptom state (PASS) and the patientdisappointing symptoms state (PDSS). As no universal consensus exists on either the PASS or the PDSS in the context of SIS, we chose to anchor our responder analysis to the patient's assessment of satisfaction with the shoulder treatment outcome: Patients reporting very satisfied or satisfied will be categorised as "Responders" and those reporting very dissatisfied or dissatisfied as "Non-responders". Given the obvious coarseness of this approach, we plan to evaluate the appropriate criteria for PASS and PDSS in more detail in the future, exploring the potential contribution of, e.g., arm pain at rest and at activity, shoulder function, and night pain.
Categorical variables, the rates of unblinding, reoperation, treatment conversion, complications and adverse effects will be analysed using logistic regression analysis or Poisson regression dependent on whether subjects with complications or (multiple) complications (per subject) are analysed. These secondary analyses will be supportive, explanatory and/or hypothesis generating, which is why multiplicity is not a problem 2 .

Sensitivity analyses
The following two sensitivity analyses will be carried out: 1) per-protocol analysis, in which the above noted primary analyses will be carried out again with patients who received the interventions as allocated will be redone; 2) and potential effects due to the treatment providing centres.
As all the participants in the ASD group have received the critical therapeutic element (subacromial decompression), no treatment group conversion is possible in this group.
In the per-protocol comparison of the efficacy of ASD (ASD vs. DA), we define the DA per-protocol population as those participants who have not received ASD during the 24-month follow-up (who have not crossed over to ASD).
In the per-protocol comparison of the effectiveness of ASD (ASD vs. ET), we define the ET per-protocol population as those participants who have not received ASD during the 24-month follow-up (who have not crossed over to ASD).

INTERPRETATION OF RESULTS
To safeguard against potential risk of bias during interpretation, a method of "blinded data interpretation" will be used 17 . In brief, an independent statistician will provide the Steering/Writing committee of the FIMPACT trial with blinded results from the analyses with study groups labelled as group A, group B, and group C. This data will be presented to the Steering/Writing Committee, who will then contemplate on the interpretation of the results until a consensus is reached and agree in writing on all alternative interpretations of the findings. Once reaching a consensus, we will record the minutes of this meeting as a statement of interpretation document signed by all members of the Writing Committee. Only after reaching this common agreement will the data manager and independent statistician break the randomisation code.
There was also variation in the actual execution of the follow-up assessments, particularly in the earlier time-points (3-and 6month follow-up visits).

IMPLEMENTATION OF ANALYSIS PLAN
This SAP will be used as a work description for the statistician performing the analyses. All analyses will be performed by the same statistician and none of the investigators involved in this trial will perform any of the statistical analyses.
The implementation of the SAP will be as follows: 1. A 'data collection form' will be outlined in a collaboration between the database manager (Leena Caravitis), statistician and principal investigators (Mika Paavola and Teppo Järvinen).
2. The database manager will code each treatment arm into 'treatment A', 'treatment B' and 'treatment C', thus leaving all others blinded to group assignment during the analyses.
3. Blinded data will be delivered to the statistician according to the 'data collection form'.
4. Primary, secondary and exploratory endpoint analyses will be made blinded to group assignment.
5. Results will be presented to the trial Writing and Steering committee, any uncertainties will be clarified and blinded interpretations of the primary endpoint results will be conducted prior to unblinding of data.