Background A variety of therapeutic interventions is available for restoring motion and diminishing pain in patients with frozen shoulder. An overview article concerning the evidence for the effectiveness of these interventions is lacking.
Objective To provide an evidence-based overview regarding the effectiveness of conservative and surgical interventions to treat the frozen shoulder.
Methods The Cochrane Library, PubMed, Embase, Cinahl and Pedro were searched for relevant systematic reviews and randomised clinical trials (RCTs). Two reviewers independently selected relevant studies, assessed the methodological quality and extracted data. A best-evidence synthesis was used to summarise the results.
Results Five Cochrane reviews and 18 RCTs were included studying the effectiveness of oral medication, injection therapy, physiotherapy, acupuncture, arthrographic distension and suprascapular nerve block (SSNB).
Conclusions We found strong evidence for the effectiveness of steroid injections and laser therapy in short-term and moderate evidence for steroid injections in mid-term follow-up. Moderate evidence was found in favour of mobilisation techniques in the short and long term, for the effectiveness of arthrographic distension alone and as an addition to active physiotherapy in the short term, for the effectiveness of oral steroids compared with no treatment or placebo in the short term, and for the effectiveness of SSNB compared with acupuncture, placebo or steroid injections. For other commonly used interventions no or only limited evidence of effectiveness was found. Most of the included studies reported short-term results, whereas symptoms of frozen shoulder may last up to 4 years. High quality RCTs studying long-term results are clearly needed in this field.
Statistics from Altmetric.com
Frozen shoulder is a common problem in general practice and its incidence is estimated to be 2–5%.1 2 Primary frozen shoulder is mentioned as one of the 23 specific disorders in the CANS model,3 which classifies complaints of the arm, neck and/or shoulder (CANS).3 In The Netherlands injuries of the upper extremity are increasingly classified according this model.4 The aetiology and pathogenesis of primary frozen shoulder are not known, but patients with diabetes and hemiplegia are predisposed.5,–,7 Shoulder injuries are common in athletes. Frozen shoulder occurs less frequent, but has a high impact due to pain and functional disability. Frozen shoulder is characterised by three phases: the freezing phase (pain and loss of motion of the glenohumeral joint in all directions), frozen phase (stiffness reaches its maximum) and the thawing phase (range of motion (ROM) returns to normal).8 Although frozen shoulder is often considered to be self-limiting, full resolution of symptoms does not always occur. Only 59% of the patients had a near normal shoulder after 4 years. However, persistent symptoms are commonly mild.1
No therapeutic intervention is currently universally accepted as most effective for restoring motion and diminishing pain in patients with frozen shoulder.9 An overview regarding the effectiveness of possible conservative and surgical interventions is lacking. Therefore, this study aimed to provide an evidence-based overview on the effectiveness of interventions for primary frozen shoulder.
A search of systematic reviews on primary frozen shoulder was performed in the Cochrane Library. In addition, reviews and randomised clinical trials (RCTs) in PubMed, Embase, Cinahl and Pedro were searched for interventions included in the systematic reviews from the date of the search strategy of the review up to April 2008 (ie, recent RCTs) and from the beginning of the database up to April 2008 for interventions not (yet) included in systematic reviews (ie, additional RCTs). See supplementary appendix I (available online only) for the complete search strategy.
Systematic reviews and RCTs were included if they fulfilled all of the following criteria: (1) the study included patients with frozen shoulder; (2) the disorder was not caused by an acute trauma or systemic disease; (3) an intervention for treating frozen shoulder was evaluated; (4) results on pain, function or recovery were reported; (5) the article was written in English, French, German or Dutch.
If a subset of the total patients included in a study met our inclusion criteria, the study was included only if the outcomes of the subset were assessed and reported separately.
Two reviewers (BMAH and SG) independently applied the inclusion criteria to select potentially relevant studies from the title and abstracts of the references retrieved by the literature search. A consensus method was used to solve disagreements concerning the inclusion of studies. A third reviewer (BWK) was consulted if disagreement persisted.
RCTs published after the search date mentioned in the (Cochrane) review and RCTs investigating interventions not summarised in a (Cochrane) review were included in this study.
Categorisation of the relevant literature
Relevant studies are presented as systematic reviews describing all Cochrane and Cochrane-based systematic reviews; recent RCTs, containing all RCTs published after the search date of the systematic review on the same intervention; and as additional RCTs describing all RCTs concerning an intervention that has not yet been described in a systematic review.
Methodological quality assessment
We adapted the score for the methodological quality10 given to the included articles in the Cochrane reviews, ranked as A (adequate), B (unclear), C (inadequate), D (not used). The number of quality criteria differed per Cochrane review, but we defined high quality as a score of 50% or more. To identify potential risks of bias of the included recent and additional RCT two reviewers (MR and MMF) independently assessed the methodological quality of each RCT. The 12 quality criteria and operationalisation of these criteria were adapted from Furlan et al.11 Each item was scored as ‘+’, ‘−’, or ‘?’. High quality was defined as a score of 50% or more (ie, a ‘+’ score on 50% or more of the criteria) on the methodological quality assessment. A consensus procedure was used to solve disagreement between the reviewers.
One author (MMF) extracted the data, which were checked by a second author (BMAH). Information was collected on the study population, interventions used, outcome measures and outcome.
For the data synthesis we used the results of the relevant individual papers identified in each Cochrane review plus the results of the recent and additional RCTs together. A quantitative analysis of the studies was not possible due to the heterogeneity of the outcome measures. Therefore, we did not perform a meta-analysis but summarised the results using a best-evidence synthesis.12 The article was included in the best-evidence synthesis only if a comparison was made between the groups (treatment vs placebo/control/treatment) and the level of significance was reported. The study results were labelled significant if it reported significant results on one of the three outcome measures. The level of evidence was ranked and divided as:
Strong evidence for effectiveness: consistent (when ≥75% of the trials report the same findings) positive (significant) findings on pain, function or recovery within multiple (two or more) higher quality RCTs.
Moderate evidence for effectiveness: consistent positive (significant) findings on pain, function or recovery within multiple lower quality RCTs and/or one high quality RCT.
Limited evidence for effectiveness: positive (significant) findings on pain, function or recovery within one low quality RCT.
Conflicting evidence for effectiveness: provided by conflicting (significant) findings on pain, function or recovery in the RCTs (<75% of the studies reported consistent findings).
No difference found in favour of the effectiveness of the invention: RCTs available, but no (significant) differences between the intervention and control groups were reported.
No systematic review or RCT found.
The follow-up period was categorised into short term (0–3 months), mid term (4–6 months) and long term (>6 months).
The literature search resulted in the identification of five Cochrane reviews, 63 RCTs and no Cochrane-based reviews. Five Cochrane reviews and 18 RCTs (17 in PubMed, one in Embase) met our inclusion criteria. Of these Cochrane reviews, three examined a mixed population with shoulder pain; two concentrated on frozen shoulder alone.
Supplementary appendix II (available online only) presents the data extraction of five Cochrane reviews (appendix IIa), 12 recent RCTs (appendix IIb) and six additional RCTs (appendix IIc). Table 1 lists the evidence for the effectiveness of each intervention.
The methodological quality of the studies from the Cochrane reviews is presented in table 2. The Cochrane review that reported on oral steroids consisted of 80% high quality studies. The Cochrane review that examined arthrographic distension has the lowest proportion (20%) of high quality studies.
The methodological criteria and quality of the 18 recent and additional RCTs are presented in table 3. Of all included studies Buchbinder et al13 met all the quality assessment criteria. Three out of 18 studies had less then six positive scores, indicating a high risk of bias. The most prevalent methodological shortcoming appeared to be the item ‘no blinding of the care provider’. Item 11 (concerning compliance) was not applicable in five studies because measurements were carried out directly after treatment.
Oral steroids are given to relieve pain and stiffness in patients with frozen shoulder. One Cochrane review was found concerning the use of oral steroids in the treatment of frozen shoulder, and included five trials (n=179).14 In three high quality trials15,–,17 oral steroids were compared with placebo or no treatment. No significant differences were found in the short term on pain (no data available)16 and in the long term on pain (RR 0.5; 95% CI 0.1 to 5.1) and ROM (abduction: RR −4.0; 95% CI −14.6 to 6.6).15 Buchbinder et al17 found at 3 weeks significant differences in favour of oral steroids for pain (RR 2.7; 95% CI 1.4 to 4.0) and ROM (abduction: RR 23.3; 95% CI 11.3 to 35.5). At 7, 13, 26 and 52 weeks no significant differences were found. In two trials, one of high18 and one of low19 quality, oral steroids were compared with other interventions. Oral steroids given before manipulation under anaesthesia were superior to manipulation alone with regard to ROM (external rotation: RR 4.7; 95% CI 1.2 to 18.6) at 6 weeks.18 The second study compared oral steroids with intra-articular steroids.19 At 1 week a significant difference was found both for pain (p=0.022) and for cure rate, defined as 90% of normal abduction and external rotation (RR 0.2; 95% CI 0.1 to 0.8). No specific data were provided and it was not reported which treatment was superior.
Intra-articular corticosteroid injections are a commonly used intervention in treating shoulder pain in general. Corticosteroids may be injected into the glenohumeral joint by an anterior, lateral and/or posterior approach, into the subacromial space, tendon sheaths or locally into trigger or tender points.20
One Cochrane review20 was found that reported on the effectiveness of corticosteroid injections in treating shoulder pain, including 13 RCTs (n=656) on frozen shoulder. One of these RCTs, of low quality21 (n=101), compared intra-articular corticosteroid injections with placebo. No significant differences were found. However, there was a trend towards steroid injections being more effective.
Three high quality recent RCTs22,–,24 (n=62) compared steroid injections with placebo. All found significant differences in favour of intra-articular steroids for pain p=0.02 (no values were given)22 (SEM 36.7.5 (5.1) vs 18.9 (5.1))23 in the short and mid term (95% CI 11.6 (0.4 to 22.8)).24 On ROM no significant differences were found (95% CI 5.5 (−1.7 to 12.8)).24
Different dosages, anatomical sites or approaches of steroid injections
In the Cochrane review20 that reported on the effectiveness of corticosteroid injections in the treatment of shoulder pain, six RCTs21 25,–,29 compared different dosages, anatomical sites or approaches of steroid injections with each other. Three of these RCTs were of high quality.25,–,27 No significant differences on pain or ROM were found.
Steroid injection versus physiotherapy
Four low quality studies (n=288)28 30,–,32 included in the Cochrane review compared steroid injections with several modalities of physiotherapy (not specified). No significant differences for pain or ROM were found. One high quality RCT33 in the Cochrane review found significant benefits of intra-articular injections over physiotherapy concerning pain (weighted mean difference (WMD) 14.0; 95% CI 3.1 to 24.9) and ROM (WMD 15.00; 95% CI 9.34 to 20.66) at 3 and 7 weeks follow-up, respectively. After 26 weeks only external rotation remained significant (WMD 9.00; 95% CI 1.67 to 16.33).
Three high quality recent RCTs compared steroid injections with several physiotherapy modalities ranging from interferential modalities to exercises.22,–,24 From 6 weeks to 4 months follow-up all trials showed significant benefit of intra-articular injections over physiotherapy alone or placebo on pain. On ROM, the combination of steroid injection with physiotherapy was more effective than physiotherapy or steroid injection alone (p<0.05).23 In contrast to van der Windt et al,33 two studies22 24 reported significant differences in favour of physiotherapy on external rotation at 3–4 months (p=0.02 and p<0.05, respectively).
Steroid injection versus other treatments
Three RCTs (n=93)5 34 35 of low quality included in the Cochrane review compared steroid injections with surgical interventions. No significant differences between injections and manipulation under anaesthesia on ROM were reported. Two studies34 35 compared steroid injections with arthrographic distension. Significant differences were reported in favour of distension on ROM (external p=0.0007, flexion p=0.03, extension p=0.01).34 35
One recent high quality RCT36 found no significant differences between steroid injection and steroid injection in combination with hydrodilatation on ROM.
In a Cochrane review37 including 26 trials, seven trials reported (n=381) on the effectiveness of physiotherapy for frozen shoulder. Two38 39 of them (n=64), of high quality, reported on laser therapy. Laser therapy was of more benefit than placebo on improvement defined as an excellent or good result (RR 3.71; 95% CI 1.89 to 7.28),39 (RR 8.00; 95% CI 2.11 to 30.34)38 after 15 treatments.
In one recent high quality RCT40 (n=63) laser therapy was in favour of placebo regarding pain and disability at 16 weeks (no data were provided).
In the Cochrane study one high quality RCT (n=20)41 showed short-term significant benefits of exercises in combination with mobilisation compared with exercises alone on ROM (internal rotation: WMD 6.1 (95% CI 0.1 to 12.1) and abduction: WMD 17.4 (95% CI 5.5 to 29.3)).
In one recent high quality RCT (n=40)42 two different physiotherapy modalities were compared with each other. The combination of deep friction massage with exercises (Cyriax method) was in favour of physiotherapy including hot pack and short-wave diathermy application after 2 weeks regarding ROM (p<0.05).
Three recent high quality RCTs compared different mobilisation techniques with each other.9 43 44 High-grade mobilisation techniques were compared with low-grade mobilisation techniques in 100 patients.43 In both groups subjects improved significantly between baseline and 1-year follow-up. High-grade mobilisation techniques were significantly more effective than low-grade mobilisation techniques in improving glenohumeral joint mobility (p values 0.01–0.05) and reducing disability (p=0.03). The effect of anterior versus posterior glide joint mobilisation on external rotation was compared in 18 patients.9 A significant difference between the groups was found in favour of the posterior joint mobilisation technique after three treatment sessions (p=0.01) for external rotation.
In a trial of 28 patients44 the use of three mobilisation techniques, end-range mobilisation, mobilisation with movement and mid-range mobilisation was compared in the management of subjects with frozen shoulder. Significant benefits of end-range mobilisation and mobilisation with movement over mid-range mobilisation techniques (p<0.05) were found after 12 weeks. Furthermore, significantly more improvement of the scapulohumeral rhythm was found after 6 weeks in favour of mobilisation with movement compared with end-range mobilisation (WMD 14.3; 6 to 22).
One additional, high quality RCT (n=125)5 reported on home exercises in combination with manipulation under anaesthesia. This treatment shows significantly better results than exercises alone regarding ROM (flexion: WMD 8; 95% CI 0 to 16) at 3 months.
Another additional RCT, of low quality, reported on oral treatment with ibuprofen (n=122).45 Ibuprofen in combination with physiotherapy was superior to ibuprofen alone concerning ROM at 3 weeks (p= 0.002).
Acupuncture is postulated to work by releasing endogenous opioids in the body that relieve pain, by overriding pain signals in the nerves, or by allowing energy or blood to flow freely through the body.46
One recent, low quality, RCT46 (n=70) compared electro-acupuncture and interferential electrotherapy with each other and with placebo. Significant differences were found between both treatment groups and the control group, on pain and function (all p<0.001) at 4 weeks.
In a Cochrane review47 the effectiveness of acupuncture for shoulder pain was studied. It included nine studies: three RCTs (n=233)48,–,50 reported on frozen shoulder. One high quality study48 (n=35) reported a significant difference in favour of acupuncture and exercises compared with exercises alone on shoulder function (9.40 WMD; 95% CI 0.52 to 18.28) at 20 weeks. A second (low quality) study49 (n=100) showed significant differences in favour of suprascapular nerve blocks (SSNB) compared with acupuncture on pain (1.33; WMD 95% CI 1.22 to 1.44) and ROM (flexion: −7.00; −11.17 to −2.83) 30 min after treatment. The third (low quality) study50 showed a significant difference between two types of acupuncture on recovery (RR 1.50; 95% CI 1.08 to 2.09) in favour of acupuncture according to Jing Luo over traditional acupuncture. The follow-up time was not stated.
Arthrographic distension of the glenohumeral joint is thought to disrupt adhesions that might be restricting the shoulder ROM. It can be achieved with combinations of saline, local anaesthetic, steroid, contrast medium and air.13
One Cochrane review (n=196)51 reported on distension for frozen shoulder. Five trials were included.34 35 52,–,54 One compared distension with placebo.52 In this high quality trial, distension showed significant differences on ROM (abduction: WMD 20.40; 95% CI 5.14 to 35.66); (internal rotation: WMD 3.40; 95% CI 1.22 to 5.58) at 3 weeks. Three included RCTs34 35 53 compared distension with corticosteroid injections. Only one35 reported significant differences in favour of distension on ROM (see Steroid section). Distension followed by physiotherapy was found to be more effective than physiotherapy alone with respect to pain (p<0.05) and ROM (p<0.01) at 8 weeks.54
This finding is in line with the results reported in a recent, high quality, RCT on the effectiveness of active physiotherapy after distension (n=144).13 Active physiotherapy (manual therapy and directed exercise) was compared with passive physiotherapy (sham ultrasound), both following distension. Significant results were found in favour of active physiotherapy at 12 weeks.
In an additional, low quality, RCT55 (n=60) pain levels were lower after distension in combination with transcutaneous electric nerve stimulation compared with distension alone. Significant differences were found between the high-intensity group and a 38% difference between the low-intensity group compared with controls (p<0.01 and p<0.05, respectively) postoperatively.
Suprascapular nerve block
One high quality trial56 was found that studied the effectiveness of (bupivacaine) SSNB compared with placebo (n=34). SSNB are used to manage acute or chronic pain.57 A reduction of pain in the treatment group compared with the placebo group was observed at 1-month follow-up (62% vs 13%, p=0.03). Two high quality RCTs56,–,58 were found that studied the effectiveness of SSNB. One RCT (n=41)57 reported on two different techniques in SSNB; needle tip guided by superficial bony landmarks was compared with the near-nerve electromyographically guided technique. The latter was more successful in providing and maintaining pain relief for up to 60 min; this difference was not significant. In the second RCT (n=30)58 the effectiveness of SSNB compared with a series of intra-articular injections was studied. Significantly better results were found for pain (p<0.001) and ROM (p<0.05) in favour of SSNB at 12 weeks.
One additional, high quality, RCT59 reported on the effect of polarity exchangeable permanent magnet (PEPM) on frozen shoulder pain. Significantly greater pain relief was found with north–south PEPM compared with non-PEPM on pain.
The goal of this study was to provide an overview concerning evidence for the effectiveness of interventions to treat primary frozen shoulder. In the short term, moderate evidence was found for the effectiveness of oral steroids before manipulation, SSNB compared with placebo or intra-articular injections, and for acupuncture or distension in addition to exercises or active physiotherapy, respectively. For the effectiveness of physiotherapy we found moderate evidence for mobilisation techniques in the short and long term and strong evidence for the effectiveness of laser therapy in the short term, although the latter is not based on exact data. Conflicting evidence was found for steroid injections compared with arthrographic distension and compared with physiotherapy. However, when subdividing outcomes on pain and ROM we found strong evidence in favour of steroids, and for ROM we found moderate evidence in favour of physiotherapy. Also strong evidence was found for the effectiveness of steriod injections compared with placebo in the short term and moderate evidence in the mid term. For other interventions limited or no evidence was found.
This review has several limitations. First, due to the lack of an unambiguous definition for frozen shoulder there were differences in selection criteria (eg, differences in loss of ROM and duration of injuries). Therefore, we used a wide-ranging search strategy, although the broadly accepted clinical definition is ‘restriction of ROM in all directions’. As frozen shoulder is a temporary condition, it is essential to consider the time factor when interpreting treatment results.60 It is striking that almost every trial presented results of less than 3 months, whereas symptoms of frozen shoulder may last up to 4 years.1
Second, there were differences in study designs. Pooling of studies on the same interventions was not possible due to the heterogeneity of the interventions. Besides, differences in the outcome measures used made it difficult to compare results. Most studies presented results on pain and restricted ROM.1 However, they were measured using different outcome scales. Furthermore, both active and passive ROM ranged from one to all movement directions. Moreover, in most studies it was not clear if the total ROM (ie, scapula movement plus glenohumeral movement) was presented, or only the ROM of the glenohumeral joint. This is important because the loss of ROM for frozen shoulder is localised in the glenohumeral joint. It was reported that lateral rotation of the scapula occurs earlier in patients with frozen shoulder to compensate the loss of ROM in the glenohumeral joint.61 The greater the limitation of glenohumeral ROM, the higher the scapular rotation.62 As a result, scapular movement alone can reach a forward flexion of 90°. This compensatory pattern of the scapula was analysed in only one study.44
One of the most common medical interventions in treating frozen shoulder are non-steroidal anti-inflammatory drugs (NSAIDs).63 Nevertheless, RCTs reporting on the effectiveness of NSAIDs in treating frozen shoulder are lacking, as are RCTs studying the arthroscopic release of the glenohumeral capsular. The latter intervention is, however, the main operative treatment for frozen shoulder at present.63
Relieving pain and restoring shoulder function are common aims in treating frozen shoulder. It is important to note the phase being treated2 because of differences in symptoms at each phase. In the freezing phase8 pain is most prominent. Intra-articular corticosteroids provide rapid pain relief, mainly in the short-term period. We found strong evidence for steroid injection on pain in the short term. At 6 weeks to 9 months after onset, restricted ROM is predominant. In this frozen phase8 therapy should concentrate on increasing ROM, such as mobilisation techniques or distension for which limited evidence was found. A possible explanation for this may be the wide variety of mobilisation techniques used. For both treatments the follow-up period of the RCTs was probably too short to expect fully restored ROM, except for two of the studies.13 43 Moreover, it was reported that the distension procedure should be reserved for patients in this phase who do not progress despite participating in a physiotherapy programme.60 In studies on the effectiveness of distension to treat frozen shoulder, this was not always the case. In the thawing phase there is a minimum of pain and progressive improvement in ROM. As pain and muscular inhibition result in compensatory movements of the scapula, the role of adaptation of scapular motion could be important in managing rehabilitation in frozen shoulder. Therefore, it is suggested that both the treatment modality and the evaluation of treatment effects should be individualised based on the stage of the frozen shoulder.
Disuse of the arm, due to the insidious onset of pain, results in loss of shoulder mobility. Pain and muscular inhibition result in compensatory movements of the scapula to minimise the pain.1 Continued use of these compensatory movements can result in pain and dysfunction elsewhere; for example development of a kissing coracoid.64 Therefore, when the pain has ceased and the ROM returns to normal (thawing phase)8 restoring the normal scapular movement should be an important goal in treating frozen shoulder. Consequently, more research is needed to assess the effectiveness of interventions used to normalise the scapulothorecal movement.
In this review we found strong evidence for the effectiveness of laser therapy and steroid injections on pain in the treatment of frozen shoulder in the short term. Moderate evidence was found in favour of mobilisation techniques in the short and long term and for steroid injections in the mid term. Moderate evidence was found for the effectiveness of distension alone and as an addition to active physiotherapy in the short term. Furthermore, moderate evidence in the short term was found for the effectiveness of oral steroids compared with no treatment or placebo and SSNB compared with acupuncture, placebo or steroid injections. For other commonly applied interventions no differences or only limited evidence was found regarding their effectiveness. Because most studies concentrated on short-term results, high quality RCT studying the effectiveness of interventions in the long term are needed.
What is already known on this topic
▶ The incidence and disease course of frozen shoulder as well as which patients are predisposed are known.
▶ A variety of therapeutic interventions is available for restoring motion and diminishing pain in patients with frozen shoulder.
▶ The treatment of frozen shoulder is still controversial despite the large number of clinical trials.
What this study adds
▶ This study provided an overview regarding the effectiveness of interventions both conservative and surgical, in the treatment of primary frozen shoulder.
▶ A systematic, evidence-based overview of the literature was lacking.
▶ This review incorporates a clinical perspective by categorising the effectiveness of interventions into the different stages of the disease process.
The authors would like to thank Dr S Glerum (SG) for participating in the study selection and Dr M Randsdorp (MR) for participating in assessing the methodological quality of the included RCT.
Funding This study was funded by Fonds Nuts Ohra.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.