The expert panel comprised physical therapists purposely selected based on their experience and knowledge in RCRSP.41,42 Only physical therapists were included in the study to make this study more relevant to physical therapy teaching, research, and practice. Additionally, trying to establish an acceptable diagnosis/label such as RCRSP may help patients to understand their problem and facilitate adherence to a treatment (i.e., exercise)18 that has been shown to be as effective as a surgical procedure.43 To assemble a representative group of experts, a list of inclusion criteria was established and the Web Of Science database was queried using a search of relevant terms related to the study topic (Supplementary File - Table 1). Additionally, a manual search was performed to verify that a series of additional experts proposed by the working group also met the inclusion criteria. The selection process is depicted in Fig. 1.

Fig. 1.

Flow diagram of experts.

(0.38MB).

RCRSP, Rotator Cuff Related Shoulder Pain; ROM, range of motion

The working group consisted of five investigators, physical therapists (NRS, RFM, ELG, RLT, JTL), who were responsible for designing the survey and collecting and analysing the data from each round of questions.

For all three Delphi rounds, the experts received an invitation by email with a link to an online questionnaire. The experts had 3 weeks to complete each round, with reminders emailed weekly.

First, the expert panel was sent three documents via email: a letter inviting them to participate in the study, an informed consent document created by the working group, and a questionnaire. The first part of this questionnaire included a list of sociodemographic questions. The second part was composed of items structured into six domains (subjective examination, patient-reported outcome measures [PROMs]), diagnostic examination, physical examination, special tests, and functional tests) as described in a previous DELPHI study.36 These initial items were chosen after two members of the working group (ELG and JTL) carried out a three-round non-published pilot Delphi study (with the same methodology as stated below for the final Delphi study) with 10 shoulder experts from February to March 2019 (Table 1). The use of a previous pilot study has been reported in some DELPHI studies44,45 and it has been suggested that it could ensure greater rigor, especially regarding the design of the first round questions.46 For the first round of the current Delphi study, participants were asked to rate the items from the pilot study using a five-point Likert scale (Strongly Disagree, Disagree, Neutral, Agree, Strongly Agree). In addition, in this first round, experts were invited to add items to each of the six examination domains.

The level of agreement among the experts was analysed by means of the Aiken's V coefficient of validity (V). This coefficient is used to quantify the content validity or relevance of an item with respect to a content domain evaluated by several experts’ judgement. Aiken's V is calculated as the ratio of the sum of agreement score obtained from all authors for a given item, with respect to the maximum possible score (i.e. maximum value of the Likert scale * number of experts rating that item). The value of Aiken's V ranges from 0 to 1, the latter representing perfect agreement. An Aiken's V ≥ 0.7 was considered reflective of group consensus, as recommended for Delphi studies.47

The second round of questions included items that reached at least a 0.7 validity index in the first round and the suggestions made by the expert panel members. No open-ended questions were included, and the experts were asked to indicate their degree of agreement with each proposed item using the previously mentioned five-point Likert scale.

For round 3, participants were informed of the results from round 2 using descriptive statistics and asked to reconsider their degree of agreement with each item before expressing their final opinion. Each participant was, therefore, asked to re-grade the criteria that reached consensus in round 2 using the same Likert-type scale.

All analyses were performed with statistical software R version 4.1.0 (R Core Team (2021). R: A language and environment for statistical computing. R Foundation for Statistical Computing, Vienna, Austria. URL https://www.R-project.org/). Mean and standard deviation (SD) and absolute and relative frequencies were used for the descriptive analyses of the continuous and nominal variables, respectively. The agreement between experts was analysed using the Aiken's Validity Index (Aiken's V), an agreement index for ordinal data with scores ranging from 0 (no agreement) to 1 (perfect agreement).48

For the first round, 56 physical therapists who were experts in the study topic were selected using the inclusion criteria, and 29 participants responded to the initial email. Of the 29 respondents, 14 declined to participate in the study for different reasons (most common: not having enough time) and 15 (26.78%) completed the three rounds of questions. The final expert panel consisted of 15 physical therapists from diverse work fields (research, clinical practice, education, or mixed). The group had an average ± standard deviation of 23.3 ± 8.1 years of experience with RCRSP and an average of 24.9 ± 17.3 scientific publications related to shoulder pain. Demographic information for the expert panel members is provided in Table 2.

At the end of the three rounds, a total of 24 RCRSP-related items were obtained (Supplementary File -Table 2). They were distributed amongst the six diagnostic domains as follows: subjective examination (n = 11), PROMs (n = 1), diagnostic examination (n = 4), physical examination (n = 6), and functional tests (n = 2). No items reached consensus within the “special tests” diagnostic domain. Two items from the "subjective examination" were merged (“Pain with movement of the arm – typically abduction or overhead activities is indicative of RCRSP” and “Pain and weakness with elevation or overhead activities are indicative of RCRSP“) and one from "diagnostic examination" (“Diagnosis imaging is required if the patient has a history of trauma”) was eliminated after reaching consensus among the working group because it was highly similar to another item. A total of 22 clinical descriptors for diagnosing RCRSP were obtained. However, some were unified and grammatically edited by the working team to simplify understanding, leaving a total of 18 items (Fig. 2). The original items of Round 3 without unification or editing are presented in Supplementary File - Table 2.

Fig. 2.

Final descriptors of the Delphi study.

(0.72MB).

Overall, 44.4% (8/18) of the items from the initial Delphi pilot study remained at the end of the third round and 92.31% (24/26) of the items that reached consensus in round 2 remained in round 3. The complete item selection process is represented in Fig. 3.

Fig. 3.

Flow diagram of the Delphi study.

(0.42MB).

The respondent group included 15 experts. There is currently no consensus on the ideal sample for a panel of experts with some authors recommending a minimum of 15,51 10,52 or even seven members.53 The quality of the panel of experts seems to be more important than the number when judging the level of representativeness of the results obtained in a Delphi study.49,54 Additionally, the criteria for defining an expert within a Delphi study are not clearly established.55 Levels of knowledge and professional experience have been proposed as main criteria.55 After analysing the characteristics of the panel members, their years of clinical experience treating people with RCRSP (23.3 ± 8.1), the number of scientific publications in a topic related to shoulder pain (24.9 ± 17.3), and the percentage of experts with PhD as the highest professional degree (80%), the quality of the panel of experts of this Delphi study seems high. This fact is important when considering the results.

The subjective examination is such a fundamental part of a physical therapist assessment that 75 to 83% of diagnostic decisions are reached based on its results.56,57

Pain related to mechanical load emerged as one of the main descriptors indicative of RCRSP in the subjective examination domain. Although the origin of RCRSP is multifactorial,13 poor load management (e.g., excessive loading, alterations to regular loading on subacromial structures) seems to be the most determining causal factor58 and plays a critical role in its progression.59 Furthermore, although the nociceptive sources of RCRSP are not only tendinous,13,23 it seems likely that the pain behavior in RCRSP is similar to that described for tendinopathies (e.g., on/off pain behavior dependent on the amount of load applied).60 A clear evidence of the influence of mechanical load on the genesis of RCRSP is that this condition is related to the performance of activities involving load in shoulder elevation,61–63 and is more prevalent on the dominant side, in active, working populations61 and overhead sports.64,65 In addition, RCRSP is more prevalent in individuals who have performed overhead activities for a long time.66–69

Regarding the location of the pain, the experts agreed that the most frequent area reported by those with RCRSP is the deltoid region, which is in line with a previous consensus report34 and experimental pain models.70 Pain reaching the forearm has also been reported in people with RCRSP.71 To the authors’ knowledge, the utility of pain distribution (e.g., using pain drawings) to assist with the differential diagnosis of shoulder pain has not fully been investigated. As experimental pain models have considerable limitations as models for clinical shoulder pain,72 future research may focus on exploring differences between patients in terms of clinical pain distribution.

Another item obtained by consensus was that RCRSP affects sleep. The relationship between the presence of shoulder pathology and sleep disturbances has been already established in other studies.73,74 Overall, aspects related to a patient's lifestyle such as sleep quality can be relevant in people with musculoskeletal pain.75,76 For example, according to the majority of the studies, sleep deprivation produces hyperalgesic changes.77 Furthermore, poor sleep can compromise the physiological processes underlying tendon recovery.78 Sleep disturbances are frequently associated with anxiety and depression, which highlights the value of performing a psychological evaluation in people with RCRSP.79,80 Also, there seems to be a dose and time-dependent relationship between tobacco consumption and the appearance of pathological changes in the rotator cuff,81 shoulder pain, and rotator cuff tear size.82 Other lifestyle-related factors (e.g., body mass index, alcohol consumption, and comorbidities like diabetes) may also be related to the development of RCRSP.66,83–85

The most frequent age of RCRSP onset related to tendon changes is between 45 and 55 years old.4 As such, being older than 50 years is a significant factor for developing rotator cuff tendinopathy among active populations.66 Moreover, being over 60 years old is a risk factor for rotator cuff tears.86 However, the item “age typically over 40″ did not reach consensus in the last round of questions (Supplementary file - Table 2). Rotator cuff disorders can appear in a wide age spectrum. Some authors even consider “two different rotator cuff diseases,” differentiating between patients under 40 years old with rotator cuff tendinopathy and older adults suffering from full-thickness tears.87 The main differences between these two populations are the potential for tissue healing, etiology, activity levels, physical demands and long-term expectations of recovery.87

Regarding age, the item “Could be secondary to atraumatic or microtraumatic instability in younger patients (< 40 years)” also reached consensus. It seems that in young people the development of rotator cuff tears may have a traumatic or atraumatic origin, this latter typical of overhead sports in the context of shoulder “microinstability.”87,88 The concept of shoulder microinstability is used to describe excessive laxity of the anterior capsule secondary to repeated movements of shoulder abduction and external rotation (i.e. overhead sports) which leads to abnormal glenohumeral biomechanics and internal shoulder impingement.89 The role of shoulder microinstability in the development of rotator cuff disorders is however controversial and some authors even consider that excessive laxity of the shoulder may act as a protective mechanism for avoiding impingement.90

The absence of neurological symptoms was reported by experts to be indicative of RCRSP, which agrees with a previous consensus report.34 Collecting sufficient data during the subjective examination can help to rule out a neurological component to the generation of symptoms.91

There are several PROMs for assessing shoulder pain and function, such as the Disabilities of the Arm, Shoulder and Hand (DASH) questionnaire, the quick DASH, the Shoulder Pain and Disability Index (SPADI), the Constant (Murley) Score (CS), and the American Shoulder and Elbow Surgeons (ASES).92,93 Some questionnaires including the Rotator Cuff-Quality Of Life and the Western Ontario Rotator Cuff (WORC) index have been developed specifically for people with RCRSP.94 In the current Delphi study, PROMs were considered useful for guiding treatment and prognosis but not to make a diagnosis of RCRSP.95 Shoulder pain questionnaires have demonstrated to be valid, reliable, and responsive to changes93,96,97 but some are not specific to the shoulder joint (i.e., DASH) and few are specific to RCRSP.94

Diagnostic imaging was considered necessary for people with RCRSP in three situations: red flags, history of trauma, and lack of response to a minimum of 3 months of conservative treatment. Imaging is widely used in medical practice for diagnosing patients with musculoskeletal pain.98–100 In people with RCRSP, magnetic resonance and ultrasound imaging have proven to be valid for diagnosing massive rotator cuff tears, with comparable levels of sensitivity and specificity.101,102 However, their sensitivity decreases with less extensive rotator cuff lesions.101,103,104 Furthermore, there is still controversy about the utility of imaging for guiding clinical decisions in people with shoulder pain including RCRSP, because it correlates poorly with symptoms.9,28,29,105 In patients with shoulder pain, it is difficult to identify the exact structure responsible for the symptomology due to the coexistence of multiple tissue alterations.32,106 However, radiological examination can help to rule out red flags.34 Finally, if we accept that exercise is key in the management of RCRSP107 and a minimum of 12 weeks is recommended to determine whether the results thereof are satisfactory,34,108 it follows that radiological examination be requested only after a lack of response to a minimum 3 months of exercise.

There was consensus amongst the experts that pain with resisted movements, in particular in shoulder abduction and external rotation, is indicative of RCRSP. This finding is described by others,13,34,109 which together suggests that pain with resisted movements is key for diagnosing RCRSP. Depending on the stage of pathology in which the patient with RCRSP is58 and the level of tissue sensitization23,110 pain could probably be reproduced with other resisted movements (i.e., resisted flexion).

Assessment of muscle strength was reported as necessary for suspected RCRSP. In fact, RCRSP has also been called "weak and painful shoulder".21,34 Subacromial pain induces alterations in the electromyographic activity of shoulder musculature and its ability to produce strength.111–114 Although deficits in shoulder abduction and external and internal rotation strength have been described in those with RCRSP,115–117 there is conflicting evidence.118

No special tests achieved the level for consensus at round 3 required to become a relevant clinical descriptor for RCRSP diagnosis. This finding is consistent with the current literature questioning the usefulness of orthopedic tests in the diagnosis of shoulder pain.24–26 Most orthopedic tests used for diagnosing shoulder pain,24–26 including RCRSP,119–122 have low diagnostic accuracy. This is likely due to the inability to isolate and specifically stress one single anatomic structure with any special test,13 the coexistence of several altered anatomical structures,23,123 and potential changes in the mechanosensitivity of local tissues.110 Therefore, currently, it is recommended to stop teaching and using special tests to establish the origin of shoulder pain.27

Interestingly, the item "depending on the condition some tests may be needed" was very close to reaching consensus in the final round (Aiken's V = 0.68). Some special tests and test clusters have been shown, for example, to improve the clinical diagnostic process for arriving at a pathology-based diagnosis in people with shoulder pain.124 In our opinion, such tests should be a component of a bigger picture, not a standalone diagnostic tool and are likely to be more useful for detecting full-thickness rotator cuff tears.

Performance-based functional tests are performed by clinicians in an attempt to quantify and discriminate impairments in patient's body functions.125 Several upper-extremity functional tests have been described and evaluated in the shoulder region both at rehabilitation and injury-prevention settings126 and their psychometric properties have been recently summarized.125

Experts considered functional tests to be relevant for patients with RCRSP reporting subtle symptoms or at end-stage rehabilitation. However, no functional test in isolation reached a sufficient level of consensus to be useful for diagnosing RCRSP. Based on these results, functional tests may then be useful for assessing people with RCRSP with low levels of irritability.127 Functional testing was also considered useful by experts as a basis for symptom modification. Symptom modification procedures based in part on symptom reproduction with patient-specific functional movements have been recommended for assessing people with RCRSP.128

As a final consideration, the diagnosis of RCRSP is, to a certain extent, a diagnosis by exclusion. This implies that other conditions must be ruled out before accepting a diagnosis of RCRSP as valid. Interestingly, experts did not reach consensus, for example, on screening the cervical spine129 or assessing passive shoulder movements130 to determine the involvement of the cervical spine as a source of pain contribution or exclude a frozen shoulder, respectively. The absence of distal neurovascular symptoms is another parameter that should be taken into account when excluding other pathologies.131

This study involved a highly experienced panel of experts who were chosen using a systematic search strategy to avoid excessive selection bias. Furthermore, the preliminary pilot study likely strengthens the validity of the final results. On the other hand, the low response rate in the initial round (26.8%) might represent a limitation when considering the external validity of the proposed descriptors. Secondly, the decision of including only physical therapists in this study could consequently affect the external validity of the diagnostic criteria across other health professionals who manage shoulder pain (e.g. general practitioners, rheumatologists, sports doctors, orthopedic surgeons). Further studies are needed to validate the proposed clinical descriptors across other health professions.