Rotator cuff is essential to function, but at risk of injury
The shoulder joint is essential for placing the hand in functional positions and as such requires mobility, strength, and stability. To provide mobility, the joint has a small shallow glenoid upon which the humeral head is permitted a large range of movement. Unlike other socket joints like the hip where bony structure provides stability, in the shoulder, stability is provided primarily by muscles. This defines the critical importance of the rotator cuff muscles and the cascade of events that follow when cuff integrity is compromised. The rotator cuff muscles are responsible for dynamic stabilization of the glenohumeral joint during shoulder motion. During abduction, the supraspinatus and the deltoid muscle work synergistically to elevate the arm. The infraspinatus, subscapularis, and teres minor depress the humeral head and maintain joint stability by opposing the upward forces created by the supraspinatus and deltoid. A cuff tear can occur after a single extreme overload, but typically it is the result of a degenerative process that compromises tendon integrity, starting with supraspinatus  and progressing to the other cuff muscles. A rotator cuff tear destroys the balance of forces at the shoulder and as a result, the humeral head is driven superiorly, resulting in pain, loss of strength, and reduced motion – all of which contribute to a profound functional loss. The progression of a complete cuff tear leads to rotator cuff arthropathy (osteoarthritis).
There is a high incidence of tears
It has been established that degeneration, i.e., frictional wear in the presence of hypovascularity, is the primary mechanism of rotator cuff tears [1–4] with anatomic factors, i.e., acromial shape, having a smaller or uncertain role [5, 6]. Thus, aging and overuse both contribute to high rates of cuff pathology. There is consensus across epidemiological studies that full thickness tears increase with age [1–3, 7–9], with some studies reporting an incidence of 5% in patients in their fourth decade and 80% of those in their eighth decade [1–3, 5, 7, 10]. While rates vary across studies, there is consensus that incidence is high and that changes are often bilateral and age dependent.
A cuff tear has a large effect on disability and quality of life
The impact of rotator cuff pathology on overall health, work productivity, and quality of life is profound. A study of over 500 shoulder problems, including 111 rotator cuff tears, indicated that health status was equally compromised in these patients as compared to 5 major medical diagnoses (hypertension, congestive heart failure, acute myocardial infarction, diabetes mellitus, and clinical depression) . Others have confirmed the impact of cuff disease on health at the primary care level . A substantial loss in work productivity is evident in the high claims rates and costs attributed to cuff disease in the worker's compensation system [9, 13, 14], where cuff pathology ranks second to back pain as a reason for lost time from work in manual workers . Cuff pathology results in high levels of pain, disability, and reduced quality of life compared to age-matched controls . It has been shown that the presence of cuff pathology is a primary determinant of health status (SF-36) (R2 = 72%) . Furthermore, rotator cuff surgery has been shown to have a significant impact; not just on shoulder symptoms, but also on overall health status .
Studies on arthroscopic and mini-open repairs are low quality and inconclusive
There has been a move across surgical disciplines to minimize the extent of surgical dissection required to perform procedures. Procedures that have traditionally been performed though open techniques are now performed arthroscopically or through mini-open procedures. For example, there have been sufficient numbers of randomized controlled trials (RCTs) comparing open and endoscopic carpal tunnel release that systematic reviews have now been completed on this topic . Recently, both arthroscopic and mini-open procedures have been introduced into shoulder surgery. While both interventions are less invasive than open procedures, there are variable advantages and concerns with these two procedures . Repair by an all-arthroscopic procedure is less invasive, but requires more extensive training . While a more rapid recovery might be expected, the adequacy of the repair and risk of complications have been questioned . Some surgeons are reluctant to invest time in training for this procedure, unless it can be shown that an all-arthroscopic repair provides superior results [18, 20].
A few low-quality (level 4) case series have evaluated the outcomes of these two surgical interventions. A completely arthroscopic technique has been reported as effective across a spectrum of tears [21–28], specifically in small to medium , moderate to large , and large/massive tears . There is consensus across these studies that clinical improvements can be obtained in the majority of cases, although a recurrent defect was reported in 17/18 large or massive tears . Similarly, other cases series report a high percentage of good/excellent results in patients treated with arthroscopically assisted mini-open repairs across a spectrum of tear sizes, specifically for both small/moderate to large  tears.
Three low-quality (level 3) studies have reported head-to-head comparison of mini-open versus all-arthroscopic rotator cuff repair [31–33]. Two studies (n = 19  and n = 76 ) found similar disability and strength scores at follow-up; another study  reported similar long-term disability, but faster return of motion in the arthroscopic group as compared to the mini-open procedure. In these studies [31–33], patients were not randomized, follow-up was retrospective, and in one study , patients received a mini-open procedure following "technical failure" of arthroscopic repair. Thus, these low-quality studies from different subpopulations may be biased since groups potentially differ in their prognostic balance. Further, comparison across previous studies is difficult due to variation in patient selection techniques, symptom duration, disability levels, and extent of pathology. To date, no randomized clinical trial has attempted to compare the results of an all-arthroscopic repair to an arthroscopically assisted mini-open repair. High-quality evidence is required to assist surgeons to determine whether the move towards less invasive procedures in cuff repair is appropriate.
There are challenges in conducting surgical trials in orthopaedic surgery
There are a number of barriers to the conduct of randomized clinical trials in surgical practices. One such barrier is the influence of surgeon skill as a component of the treatment process . Surgeons are rarely equally competent at two different surgical procedures for the same clinical problem, and even more rarely do they believe that two surgical options are equally beneficial . Thus, the classic randomized trial developed in medicine to evaluate drug therapies can fall short when applied to the evaluation of surgical interventions, where skill and preference for a particular procedure have the potential to bias the results or challenge the feasibility of the trial. In fact, it has been shown that surgeons are less likely to participate in Randomized Controlled Trials (RCTs) as compared to other medical specialties . Surgeons may be reluctant to participate in trials because they do not consider themselves to be in a position of equipoise or may be reluctant to enroll all patients because of treatment preferences. High-quality trials require a broad level of participation to achieve the sample sizes required for adequate power.
To address these challenges, shoulder surgeons across Canada formed a national collaborative group focused on multi-centered clinical trials for shoulder conditions (Joint Orthopaedic Initiative for National Trials of the Shoulder) (JOINTS Canada). Most recently, JOINTS Canada obtained CIHR funding to complete strategic planning on the priority research question facing shoulder surgeons and to identify trial methods suited to meet the challenges inherent in surgical trials.
Expertise-based random allocation may enhance feasibility and validity
In 1980, randomizing subjects to clinicians was first suggested by Van der Linden [36, 37]. Since that time, however, this design has been little used. The expertise-based randomized controlled trial has been recently revisited as a potential alternative to minimizing these barriers . In a conventional RCT, both interventions are offered by each health care provider. This works well in drug trials where there is no interaction between the drug and its provider and where the provider can be blinded to the intervention they are dispensing. However, in surgical trials there are problems with this approach that have been recently described by the Expertise-Based Group . The primary concern is that the conventional RCT may actually introduce a bias, due to "differential expertise" within a treatment group that may act as a confounder leading to spurious or biased results. It has been well established that the patients of surgeons who have more experience can expect better outcomes following a surgical intervention. Thus, the patients of surgeons participating in a conventional clinical trial will tend to have better outcomes if they are assigned to the treatment in which their surgeon is more experienced. Greater imbalance in expertise between surgeons compounded by an imbalance in recruitment rates between surgeons will result in greater differential expertise bias and lower trial validity. In fact, extensive experience may be required to achieve superior outcomes. Thus, conducting a specific number of "run-in cases" prior to enrolling patients (used in conventional RCTs) does little to impact on this form of bias.
Not only do surgeons have differential expertise, but also this is often associated with strong beliefs about which intervention is more effective. In fact, initial beliefs of trial surgeons about which intervention is superior has been shown to persist at the end of the trial . Therefore, it is possible that surgeons will introduce differential treatment bias by managing cases with their preferred intervention in a more rigorous manner. Although self-report measures can obviate the impact of surgeon bias on measurement of outcomes, these cannot control for differential treatment bias. Another form of bias tends to manifest itself in the conventional design because a greater number of procedural crossovers (surgeon does not comply with the random assignment) occur when the patient is assigned to the less preferred treatment arm. Increased rates of crossing over also compromise the validity of the RCT results. Finally, from an ethical viewpoint, it is preferable that patients receive care from a surgeon using a procedure that they were experienced in performing.
Despite these potential advantages, the expertise-based design has rarely been used except in cases where the interventions arise from different disciplines . After consulting with experts , we devised a strategy to maximize valid participation of trial surgeons. Given the potential for differential expertise bias, we concluded that only surgeons with long-term (>2 years) and recent experience in both procedures were appropriate for the conventional RCT in our study. Those who demonstrated differential expertise or preference are required to participate in the expertise-based RCT. While this requires two designs, it ensures that the potential bias due to differential expertise or preference for one procedure over another is reduced.
Systematic reviews confirm that treatment is based on inadequate evidence
Two separate research teams have attempted to conduct systematic reviews to summarize the available evidence for management of shoulder problems, including rotator cuff disease. A 1998 systematic review included all randomized trials that addressed shoulder pain and any related treatment intervention and found insufficient evidence to make any specific recommendations with regards to management of rotator cuff disease . A recent systematic review of surgical or conservative interventions included a broader range of evidence (including level 2 and 3 studies) and specifically focused on patients with rotator cuff disease. This review found weak evidence to suggest that open and primary surgery were more effective than arthroscopic debridement and revision surgery . However, these conclusions do little to provide guidance on current practice trends as these procedures have different indications and recent techniques were not addressed. No randomized clinical trials have compared mini-open to an all-arthroscopic repair technique.
A trial is needed now
The wait-list for this procedure ranges from 6 months to 2 years in Canada, indicating a high demand for effective rotator cuff repair procedures. Canadian shoulder surgeons who are active members of JOINTS Canada obtained funding from CIHR to mount a team-planning meeting where we identified and reached a consensus that this research question was the top priority for shoulder surgeons and their patients. The timing of this clinical trial is critical. An RCT mounted too early will have difficulty recruiting a sufficient number of surgeons with the experience required to perform the study interventions. If the trial is delayed until less invasive surgical techniques become the standard of practice, a large group of patients will be exposed to suboptimal care should this practice prove less effective. Further, it will be more difficult to change clinical practice after this shift has occurred. Currently, surgeons are making decisions about abandoning traditional repairs in the search of the ideal minimally invasive procedure. Without the benefit of high-quality evidence to support these decisions, premature selection may expose patients to potentially useless or risky procedures. Certainly, haphazard incorporation of new surgical approaches exposes the health care system to inefficient allocation of its limited resources.