Background: Adhesive capsulitis often is difficult to diagnose in its early stage and to differentiate from other commonly seen shoulder disorders with the potential to cause pain and limited range of movement.

Objectives: The purpose of this study was to establish consensus among a group of experts regarding the clinical identifiers for the first or early stage of primary (idiopathic) adhesive capsulitis.

Design: A correspondence-based Delphi technique was used in this study.

Methods: Three sequential questionnaires, each building on the results of the previous round, were used to establish consensus.

Results: A total of 70 experts from Australia and New Zealand involved in the diagnosis and treatment of adhesive capsulitis completed the 3 rounds of questionnaires. Following round 3, descriptive statistics were used to screen the data into a meaningful subset. Cronbach alpha and factor analysis then were used to determine agreement among the experts. Consensus was achieved on 8 clinical identifiers. These identifiers clustered into 2 discrete domains of pain and movement. For pain, the clinical identifiers were a strong component of night pain, pain with rapid or unguarded movement, discomfort lying on the affected shoulder, and pain easily aggravated by movement. For movement, the clinical identifiers included a global loss of active and passive range of movement, with pain at the end-range in all directions. Onset of the disorder was at greater than 35 years of age.

Conclusions: This is the first study to use the Delphi technique to establish clinical identifiers indicative of the early stage of primary (idiopathic) adhesive capsulitis. Although limited in differential diagnostic ability, these identifiers may assist the clinician in recognizing early-stage adhesive capsulitis and may inform management, as well as facilitate future research.

Adhesive capsulitis of the shoulder is a disorder frequently encountered by primary health care professionals. It often is difficult to identify and correctly diagnose in its early stage. Labeled “frozen shoulder” by Codman in 19341 but subsequently termed “adhesive capsulitis” by Neviaser2 to better describe the pathology, this condition generally is characterized by pain and a gradual progressive loss of shoulder active and passive range of motion.3 It has been reported that its prevalence is 2% to 3% in the general population.35 This figure is higher in the diabetic population,6 with a prevalence of 30% reported in patients with type 2 diabetes mellitus.7 Adhesive capulitis also is reported to be more common in women, especially between the ages of 40 to 60 years.3,5,8,9 The condition usually very slowly progresses toward spontaneous resolution; however, the findings of several long-term studies1014 suggest that ongoing impairment may persist in some patients.

Adhesive capsulitis is described as being either primary or secondary.10,15,16 Primary adhesive capsulitis is due to an unknown cause (ie, it is idiopathic), whereas secondary adhesive capsulitis results from a known cause or surgical event.4 Published descriptions of the condition commonly further subdivide it into 3 or 4 stages. Following arthroscopic evaluation, Neviaser and Neviaser8 identified 4 stages of involvement. These 4 stages have been correlated with clinical examination findings and histological appearance of the tissues.4 The more-recent literature, however, generally describes adhesive capsulitis as consisting of 3 stages.3,5,15 These stages have been identified as the painful stage (first), the adhesive stage (second), and the resolution stage (third). The painful stage in this nomenclature includes both stage 1 (the pre-adhesive stage) and stage 2 as described by Neviaser and Neviaser.8 The current study was concerned with identifying primary adhesive capsulitis in the painful or first stage of the condition.

Although “textbook” descriptions of diagnostic criteria for adhesive capsulitis, including variable pain and movement characteristics, are present in the literature, validation of these descriptions is lacking. Currently, the diagnosis of primary adhesive capsulitis is based on the findings of the patient history and physical examination. No specific clinical test or definitive investigation has been reported in the literature, and there remains no gold standard to diagnose this disorder. A varying range of “typical” signs and symptoms, such as pain aggravated by shoulder movement,4,5 pain at night,8 and multidirectional limitation of active and passive joint movement accompanied by pain at the extremes of range,3 have been proposed instead. To date, however, there are no agreed-upon or validated diagnostic criteria for the disorder.

The lack of validity and reliability for the diagnostic classification of shoulder pain has been a topic of controversy for some time.1722 In a study of interobserver agreement between general practitioners and physical therapists, this deficit has been particularly highlighted.23 However, the need for diagnostic labels for shoulder disorders has been questioned, as there is some evidence that the outcomes of treatment may be similar for heterogeneous groups of patients with shoulder pain lacking a specific diagnosis.2427 Conversely, other authors28,29 have suggested that a uniform method of defining shoulder disorders is necessary. In a systematic review of randomized controlled trials of interventions for the painful shoulder,28 the authors commented that, in the studies sampled, no standard diagnostic definitions were used, and indeed conflicting criteria were used to define the same condition in various trials. These limitations make drawing conclusions across studies difficult. Although a set of diagnostic criteria may not exclusively represent a single pathological entity, it may represent a subgroup of patients to whom randomized controlled trials may be directed.

Similarly, early and accurate identification of diagnostic criteria is recommended for determining prognosis as well as optimizing treatment outcomes in the clinic.30 Early presentation of shoulder disorders has been associated with a favorable outcome.31 Some authors4,8 have recommended that treatment and prognosis for adhesive capsulitis should be tailored to the stage of the disorder. Consequently, it is arguably appropriate to establish diagnostic criteria for each stage rather than for the disease process as a whole.

The difficulty faced by clinicians in the diagnosis of shoulder disorders recently was addressed by Mitchell and colleagues.32 They proposed a simple model to assist in the diagnosis of rotator cuff, glenohumeral, and acromioclavicular joint disorders, as well as referred cervical spine pain. Although potentially facilitating aspects of the clinical reasoning process, this model fails to recognize the various stages of adhesive capsulitis. Agreed-upon diagnostic criteria for early-stage adhesive capsulitis, therefore, remain to be established.

The aim of this study was to reveal such consensus that may currently exist among a group of experts regarding the clinical signs and symptoms indicative of the first stage of primary adhesive capsulitis. The establishment of such consensus is the first step in the process of identification and validation of agreed-upon diagnostic criteria for this disorder.


The Delphi technique was chosen to explore this issue because it is an established and recognized method of deriving the opinion of experts to determine the degree of consensus where there is a lack of empirical evidence.33,34 This technique has the advantages of maintaining anonymity among respondents, allowing time for participants to consider their response while not being influenced by dominant individuals and enabling recruitment from diverse geographical locations and clinical backgrounds.35,36 Using a panel of experts, the Delphi technique is a multistage process using a series of sequential questionnaires or rounds linked by feedback. Each round of the process builds on the results of the previous one and results in consensus by the final round. This technique has been widely used in establishing consensus on various diagnostic descriptors and clinical identifiers.3742


The participants were a group of experts involved in the diagnosis and treatment of adhesive capsulitis and recruited from several disciplines. These disciplines included rehabilitation medicine, physical medicine, orthopedic surgery, physical therapy, chiropractic, and osteopathy. Medical practitioners invited to participate in the study were required to hold postgraduate qualifications in a relevant specialty or be members of a special interest group in a discipline relevant to the study. Rehabilitation medicine physicians were recruited from the Musculoskeletal Medicine and Pain Special Interest Group, a subgroup of the Australasian Faculty of Rehabilitation Medicine. Members of the Australasian Faculty of Musculoskeletal Medicine also were included, as were members of the College of Physical Medicine. As a special interest group of the Australian Orthopaedic Association, members of the Shoulder and Elbow Society of Australia were approached. Physical therapist participants were members of Shoulder and Elbow Physiotherapists Australia (a physical therapy subgroup of the Shoulder and Elbow Society of Australia), as well as coordinators of postgraduate musculoskeletal physical therapy programs at Australian and New Zealand universities. In addition, specialist musculoskeletal physical therapists recognized by the Australian Physiotherapy Association and the Australian College of Physiotherapists were included. Australian and New Zealand authors who had published on the topic of adhesive capsulitis in peer-reviewed journals or texts in the past 10 years also were invited to participate. These potential participants were identified by searching MEDLINE and CINAHL databases using the search terms “adhesive capsulitis” and “frozen shoulder.” Only articles published in the English language between February 1996 and February 2006 were identified. The reference lists of identified articles also were scrutinized in an attempt to identify any texts or other references that may have been published during this period. Where contact details indicated the authors were located in Australia or New Zealand, these individuals were included in the expert group. Finally, chiropractors and osteopaths who were coordinators of postgraduate musculoskeletal programs offered at Australian and New Zealand universities were approached. A total of 185 potential participants were contacted in the first round.

Pilot Study

A pilot study, using a sample of convenience comprising 6 participants representative of the overall sample, was performed prior to the commencement of the main study to determine whether the instructions to participants were clear and to identify any improvements to the method. Following the pilot study, it was determined that 2 reminders should be issued to nonresponding participants to maximize the response rate. It also was determined that documents should be highlighted to more clearly indicate that stage 1 of adhesive capsulitis was being investigated, not the later, more easily recognizable stages.


The study was correspondence based, and the questionnaires were distributed by the researchers to the participants’ work addresses. Addresses were obtained from the appropriate organizations, and all contact details were available in the public domain, with the exception of the rehabilitation medicine physicians, whose members were approached through the chairperson of the Musculoskeletal Medicine and Pain Special Interest Group. In this case, the letter of invitation was sent to the chairperson of the group, requesting it be forwarded to members. Those members who were potentially interested in participating were asked to contact the researchers directly. Members of the Faculty of Musculoskeletal Medicine also were approached through the chairperson of the faculty, who provided names and contact details of members. All of the participants who were clinicians were approached at their private clinics.

Experts were asked to participate in 3 rounds of questionnaires. For the first round, potential participants were sent a letter of invitation along with the first questionnaire and were given 2 weeks to reply. Participants were given the opportunity to receive the subsequent questionnaires electronically and to supply a contact telephone number. A reminder was sent if a response was not received in the specified time, and, if necessary, a second reminder was issued after a further 2 weeks. The same approach and time frame for reminders were used for the 2 subsequent rounds. Telephone contact was used in the second and third rounds for the second reminder if a telephone number was made available by the participant.

Round 1.

The first questionnaire requested participants to list as many or as few diagnostic criteria as they considered necessary and sufficient to diagnose stage 1 primary adhesive capsulitis. Respondents were given the opportunity to provide a rationale for their criteria if they felt this appropriate. The responses were independently reviewed and collated by each of the 3 researchers, using a series of operational rules. These rules involved listing all of the criteria (individual responses) proposed, grouping the criteria into relevant clinical categories, eliminating single responses, merging repeated responses, and discarding unclear responses. Responses clearly inconsistent with the literature or obviously relating to secondary adhesive capsulitis or the later stages of the target disorder also were discarded. Following initial independent review, the researchers met and reached a consensus on the criteria to constitute the second round.

Round 2.

The second round used the criteria identified in round 1 by all participants. In this round, participants were asked to score the importance of each criterion in the diagnosis of stage 1 adhesive capsulitis using the following 5-point Likert scale adapted from Cook et al39:

  1. Strongly agree: the selected criterion is extremely important in the diagnosis of stage 1 of primary adhesive capsulitis.

  2. Agree: the selected criterion is important in the diagnosis of stage 1 of primary adhesive capsulitis.

  3. Undecided: uncertainty about the importance of the selected criterion in the diagnosis of stage 1 of primary adhesive capsulitis.

  4. Disagree: the selected criterion is not important in the diagnosis of stage 1 of primary adhesive capsulitis.

  5. Strongly disagree: there is absolutely no importance whatsoever of the selected criterion in the diagnosis of stage 1 of primary adhesive capsulitis.

Round 3.

The third round provided feedback to the participants in the form of the percentages for each of the 5 response options as to how all participants rated each criterion in round 2. In the light of this information, participants were requested to rescore each criterion on the same Likert scale used in round 2.

Data Analysis

The data were analyzed initially using simple descriptive statistics. The Cronbach coefficient alpha then was used as a measure of the level of consistency of opinion among the respondents regarding the agreed-upon criteria. Finally, to determine the underlying structure of the criteria, a factor analysis was performed.


From the 185 potential participants approached in the first round, 89 responses (48.1%) were received. From the 89 respondents from round 1, 75 responses (84.3%) were received following round 2. Seventy (93.3%) of these respondents completed the final round. Overall, 37.8% of the original sample completed all 3 rounds. The response rate of participants in each discipline is indicated in Table 1, and the flow of participants through the study is depicted in Figure 1.

Figure 1.

Flow of participants through the study.

Table 1.

Composition and Response Rate of Participants in Delphi Study

Following the first round, 367 criteria were generated. Collation of the data resulted in the establishment of 60 diagnostic criteria structured into 6 sections to form round 2. These criteria are outlined in Table 2. Following round 3, the data were analyzed initially using descriptive statistics. As the purpose of the study was to seek strongly held views of experts and the initial request had been for necessary and sufficient criteria, it was determined that only the “strongly agree” response would be analyzed. Therefore, the number of respondents scoring “strongly agree” was calculated and is graphically represented in Figure 2. In order to determine the criteria to be used in further analysis, several principles were applied. First, the Pareto principle,43 which suggests that 20% of the items would determine 80% of the value or benefit in deciding what is important in diagnosis, was used to commence analysis. By applying this principle, 12 criteria were identified. Second, the pattern of drop-off of frequency for these items resulted in a delineation at 10 criteria. As this delineation was in reasonable agreement with the Pareto principle, it was considered that this was an appropriate cutoff to select. As a result, 10 criteria (in descending order, criteria 13, 14, 25, 42, 12, 15, 34, 22, 60, and 26) were selected for further analysis.

Figure 2.

Percentage of respondents scoring a criterion as “strongly agree” (n=70).

View this table:
Table 2.

Items Generated Following Round 1

In order to measure the internal consistency of the criteria, Cronbach alpha was used. Using SPSS version 15,* an analysis of the 10 selected criteria resulted in a Cronbach alpha value of .63. Stepwise removal of items whose inclusion reduced the alpha value was performed (criteria 42 and 60). Removal of these 2 criteria maximized Cronbach alpha to .71. Eight criteria were established as a result of this analysis and are presented in Table 3. As the underlying structure of these criteria was of interest and factor analysis was proposed, a Kaiser-Meyer-Olkin measure of sampling adequacy was performed to determine whether factor analysis would be of benefit. The value of this test was .661. A value above .60 indicates that it is worthwhile proceeding with factor analyis.44 A factor analysis using varimax rotation, therefore, was performed on the remaining 8 criteria to examine their underlying structure. Figure 3 demonstrates the scree plot for this calculation. The result of this factor analysis determined 2 discrete dimensions of pain and movement into which the criteria clustered. This is represented in Figure 4. These factors together accounted for 56.3% of the total variance of the expert responses, with the pain factor accounting for 36% and the movement factor accounting for 20.3%. The relative weights of the 8 criteria are shown in Table 4, which provides factor loadings for each criterion in the 2-factor solution.

Figure 3.

Scree plot of final components selected.

Figure 4.

Component plot of diagnostic criteria following factor analysis.

Table 3.

Diagnostic Criteria Achieving Consensus

Table 4.

Factor Loadings Following Principal Components Factor Analysis of Clinical Criteria


The Delphi technique was used successfully in this study to establish consensus among a group of musculoskeletal professionals on 8 clinical identifiers for the first stage of primary (idiopathic) adhesive capsulitis. Although the initial aim of the study had been to establish diagnostic criteria and instructions to participants had been to respond as such, following data analysis it was considered more appropriate to alter the nomenclature of the set of resultant criteria to clinical identifiers. In a recent Delphi study of lumbar zygapophyseal joint pain,42 a similar dilemma was encountered, with experts in medical disciplines applying different definitions to the term “diagnostic criteria.” Following the first round of that study, it was decided to replace the phrase “diagnostic criteria” with “criteria indicative” of lumbar zygapophyseal joint pain to more appropriately reflect the responses received. At the conclusion of the current study, the term “clinical identifiers” was similarly determined to be more appropriate for the set of criteria established, as they could not be regarded as a gold standard for diagnosis or provide a differential diagnosis, but rather are a set of clinical identifiers that may assist the clinician in diagnosis, as well as help form the basis for further research.

Unlike many earlier published studies using the Delphi technique, the application of rigorous statistical analysis, rather than only simple descriptive statistics, was used to determine consensus in this study. Notably, factor analysis in this study has resulted in identifiers clustering into 2 discrete domains of pain and movement.

Clinically, diagnosis of adhesive capsulitis is made through the history and physical examination. Textbook descriptions of the clinical characteristics of adhesive capsulitis identify a number of features present in each of the various stages of the disorder.45 These features encompass onset and description of pain, as well as effect on movement. Similarly, in published studies such as a recent systematic review of physical therapy for adhesive capsulitis,46 many of the clinical identifiers proposed by respondents in the present study are described, despite a lack of validation. Although these identifiers (including descriptions of pain and movement) are commonly proposed, they have not previously been subjected to formal evaluation. Using the Delphi technique, the present study is the first to subject these descriptors to scrutiny and begin the process of validation.

To date, there has been no agreement on the necessary criteria or clinical identifiers required for diagnosing adhesive capsulitis in its early stage.20,45,47,48 However, it has been suggested that although the exact identifiers are poorly defined, pain is a significant feature in this stage.4 Our study supports this premise, with several dimensions of pain being qualified and achieving consensus. A strong component of night pain, a marked increase of pain with rapid or unguarded movements, discomfort lying on the affected shoulder, and pain easily aggravated by movement were the 4 descriptors of pain on which consensus was achieved. Although not validated, night pain or sleep disturbance has been described previously as a feature of this disorder in the early stage.8,10,46,47 There also are descriptions in the literature of pain easily aggravated by movement.4,5 Although probably not exclusive to adhesive capsulitis, these descriptors of pain may reflect the pathology of inflammatory synovitis that has been demonstrated at this stage.8,49 The panel of experts in this study concur that these identifiers are necessary to diagnose early-stage primary adhesive capsulitis. Although the identifiers describing location and intensity of pain did not reach consensus, the pain identifiers described and for which consensus was reached may assist the clinician in the diagnosis of early-stage adhesive capsulitis.

The exact characteristics of movement dysfunction in the early stage of adhesive capsulitis are not clearly described in the literature. Although the effect on movement in the later stages of the disorder usually is described and even quantified, description of any movement deficit in the early stage generally is minimal. Nonetheless, general restriction of movement in all directions at this early stage has been described previously.3,5,10 This study achieved consensus on the clinical identifiers of global loss of both active and passive ranges of movement, accompanied by pain at the end-range in all directions. Although no specific quantification of the loss at this stage has been determined, the fact that loss is global, rather than related to a specific direction, is the key feature in this clinical descriptor. Unlike many other shoulder pathologies, adhesive capsulitis is a disorder mainly affecting the glenohumeral joint capsule.8 Global loss of active and passive range of motion is consistent with pathology of this structure. In addition, pain at the end-range in all directions is a feature that also may raise the level of clinical suspicion of adhesive capsulitis and also is consistent with capsular pathology.3

Demographic factors of adhesive capsulitis, including the age of onset, are considered relevant clinical features important in diagnosis. Generally, it is suggested in the literature that patients affected by this disorder are over 40 years of age.3,4,8,9 Following round 1, a variety of responses quantifying age were received from the expert panel, such as “not seen less than 30 years of age,” “middle aged 45–60,” and “age 50s.” The most-frequent response was captured in criterion 22 (“the onset is generally in people greater than 35 years of age”). Interestingly, criterion 23 (“The onset is generally in people less than 60 years of age”), which was descriptive of the upper age limit for this disorder, did not achieve consensus. Therefore, in this study, there was consensus that the age of onset of the disorder generally is greater than 35 years. This finding is consistent with previous published literature, although no explanation was offered.3,4,8,9 The higher incidence of women in the 40- to 60-year age group, which failed to reach consensus, has been hypothesized to coincide with menopause and perimenopause,50 but as yet this hypothesis remains unproven. The factor analysis determined that those respondents who regarded clinical identifiers in the pain dimension as diagnostically important consistently reported age (criterion 22) alongside the pain identifiers. As pain behavior and age generally are considered patient-reported data and not physical examination findings, it is appropriate that the clinical identifier describing age clustered with identifiers describing pain rather than with movement findings.

Interestingly, the 8 clinical identifiers established in this study did not include any negative findings. Instructions to participants were not to limit responses to positive findings, and indeed negative findings were solicited; however, they failed to reach consensus. This finding is relevant, as the presence of pathology in structures other than the glenohumeral joint capsule may elicit differing clinical characteristics that would raise doubts about a diagnosis of adhesive capsulitis. Acute cervical radiculopathy or rotator cuff tendinitis, for example, may be recognized by other clinical features that would contribute to a differential diagnosis. As such features did not reach consensus in the current study, the limitation of the results in assisting differential diagnosis is acknowledged. A further consideration is whether the resultant group of identifiers should be regarded as a set or as individual items. Instructions to participants had been to give a “set of necessary and sufficient diagnostic criteria”; however, it remains to be determined whether all or only some are necessary in diagnosis. This is particularly relevant, as some of the identifiers also may be present in other acutely presenting shoulder disorders of differing pathology.

The recent suggestion that attempting to place diagnostic labels on groups of patients in clinical research trials is of little value22 may overstate the case. Arguably, one of the aims of establishing diagnostic criteria is to identify a homogenous subgroup of patients with which to evaluate treatment outcomes and make comparisons across trials more meaningful. Although there is some evidence that the outcomes of treatment may be similar in heterogeneous groups,2427 it remains to be seen whether subgroups of patients with common clinical features experience greater benefits with particular interventions.

The Delphi technique, and its application in this study, has a number of limitations. However, it was chosen because it enabled the engagement of a large number of musculoskeletal experts from a range of relevant professions and across a wide geographical area. One limitation often described is that there may be a poor response rate to the questionnaires.35,36 In this study, the initial round had a moderate response rate of 48.1%, whereas the second and third rounds had high response rates of 84.3% and 93.3%, respectively. It has been suggested that a poor response rate may characterize the final rounds35; however, this did not occur in the current study. The overall response rate for this study was 37.8%, which compares favorably with recent studies that also had a large sample but achieved a response rate of only 8.4%.38,39 Researcher bias also has been proposed as a weakness of the Delphi technique. The use of an open initial response in round 1 achieved a richness of collected data; however, this required care in reducing data to a more manageable volume for the subsequent rounds. Strict operational definitions were used by the 3 researchers to minimize bias. Furthermore, following round 3, rather than just using simple descriptive statistics as in many earlier studies, a more rigorous analysis was used to provide a more independent insight into the data.

Composition and size of the expert panel in Delphi studies vary across the literature. In an article discussing the methodology of the Delphi technique, Williams and Webb51 noted that there is no agreement regarding the optimal size of an expert panel. They commented that the panel size of studies reported in the earlier literature varied from 10 to 1,685 participants. In the current study, the inclusion criteria for potential participants determined the size of the expert panel. These inclusion criteria were established to recruit musculoskeletal practitioners and leaders in several fields with expertise in clinical, research, and educational facets of shoulder pain. Although medical practitioners were represented, omission of rheumatologists, who may assess and treat musculoskeletal disorders, could be regarded as a limitation of this study. This omission occurred because it was not possible to identify a defined special interest group in musculoskeletal medicine or orthopedics within the Australian Rheumatology Association. Regional differences in prevalence or characteristics of adhesive capsulitis are not described in the literature. However, as the participants in this study were recruited from Australian and New Zealand experts, the results may reflect only views held in this region.

The present study not only addressed the difficulty faced by clinicians in the diagnosis of shoulder disorders as described by Mitchell and colleagues,32 but also is the first of its kind to establish a set of clinical identifiers for the early stage of primary adhesive capsulitis. Although a specific diagnostic test or negative findings that may contribute to differential diagnosis did not achieve consensus in this study, several parameters of patient presentation have been established. These agreed-upon clinical identifiers should assist in the clinical decision-making process and aid in the early recognition of this disorder. They also represent the first step in the longer process of identification and validation of the agreed diagnostic criteria for this disorder.


The results of this study provide a framework for the validation of clinical identifiers for early primary adhesive capsulitis in further studies, as well as potentially facilitating comparisons across future clinical trials. Although the identifiers established in this study do not constitute an exclusive or discriminatory set of diagnostic criteria, they may be of assistance to the clinician confronted with the diagnostic dilemma of recognizing the early stage of primary adhesive capsulitis.


  • All authors provided concept/idea/research design and data analysis. Ms Walmsley and Dr Rivett provided writing. Ms Walmsley provided data collection. Dr Rivett provided institutional liaisons. Dr Rivett and Mr Osmotherly provided consultation (including review of manuscript before submission).

  • This study was approved by The University of Newcastle Human Research Ethics Committee.

  • An interactive poster presentation of selected findings of this study was given at the International Federation of Orthopaedic Manipulative Therapists (IFOMT) Conference; June 8–13, 2008; Rotterdam, the Netherlands.

  • * SPSS Inc, 233 S Wacker Dr, Chicago, Il 60606.

  • Received October 26, 2008.
  • Accepted May 19, 2009.


View Abstract