Intended for healthcare professionals

Clinical Review

Frozen shoulder

BMJ 2005; 331 doi: (Published 15 December 2005) Cite this as: BMJ 2005;331:1453
  1. Richard Dias, specialist registrar in orthopaedics1,
  2. Steven Cutts, specialist registrar in orthopaedics (stevenfrcs{at},
  3. Samir Massoud, consultant orthopaedic surgeon1
  1. 1 Royal Orthopaedic Hospital, Birmingham B31 2AP
  1. Correspondence to: S Cutts
  • Accepted 3 November 2005

Frozen shoulder is a painful, often prolonged, condition that requires careful clinical diagnosis and management. Patients usually recover, but they may never regain their full range of movement.


Frozen shoulder is a disabling and sometimes severely painful condition that is commonly managed in the primary care setting. True frozen shoulder has a protracted natural history that usually ends in resolution. In this article we consider how to diagnose frozen shoulder and how to distinguish it from other painful shoulder conditions. We also look at the current aetiological theories and the effectiveness of conservative and operative management. We reviewed the current literature on this topic and discussed papers of historical interest with consultants in our department. We have also made reference to key papers cited in Clinical Evidence (

What is frozen shoulder?

The term “frozen shoulder” was first introduced by Codman in 1934.w1 He described a painful shoulder condition of insidious onset that was associated with stiffness and difficulty sleeping on the affected side. Codman also identified the marked reduction in forward elevation and external rotation that are the hallmarks of the disease.

Long before Codman, in 1872, the same condition had already been labelled “peri-arthritis” by Duplay.w2 In 1945, Naviesar coined the term “adhesive capsulitis.”w3 Although still in use, this more recent term is unfortunate since, although a frozen shoulder is associated with synovitis and capsule contracture, it is not associated with capsular adhesions.

In clinical practice, the tendency is to label any patient with a stiff, painful shoulder as a case of frozen shoulder. This should be resisted. Frozen shoulder is a specific condition that has a natural history of spontaneous resolution and requires a management pathway that is completely different from such distinct shoulder conditions as a rotator cuff tear or osteoarthritis.

Who gets it?

Frozen shoulder patients usually present in the sixth decade of life, and onset before the age of 40 is very uncommon.w4 The peak age is 56, and the condition occurs slightly more often in women than men.1 w4 In 6-17% of patients, the other shoulder becomes affected, usually within five years, and after the first has resolved.1 w4 The non-dominant shoulder is slightly more likely to be affected.1 w4

Summary points

True frozen shoulder is a clinical diagnosis

The three hallmarks of frozen shoulder are insidious shoulder stiffness; severe pain, even at night; and near complete loss of passive and active external rotation of the shoulder

Lab tests are normal

Frozen shoulder is rare under the age of 40; the peak age is 56

Frozen shoulder progresses through three clinical phases

It lasts about 30 months, but recovery can be accelerated by simple measures

Physiotherapy alone is of little benefit, although steroid injection is effective and best combined with physiotherapy

Refractory cases can be referred for manipulation under anaesthesia and, rarely, arthroscopic release

Nearly all patients recover, but normal range of movement may never return

Few attempts have been made to calculate the cumulative lifetime risk of frozen shoulder. In the Scandinavian population at risk, it has been estimated at a minimum of 2% per year.w4 w5 Recurrence is highly unusual.w6

Clinical presentation and examination

A patient with frozen shoulder traditionally progresses through three overlapping phases (box).2

When examining any joint, it is useful to apply the well known axiom of the late Alan Apley, a popular orthopaedic speaker and teacher: “Look, Feel, Move.”

Embedded Image

Arthroscope view of a shoulder with synovitis

Credit: CNRI/SPL

Look: On inspection, the arm is held by the side in adduction and internal rotation. Mild disuse atrophy of the deltoid and supraspinatus may be present.

Feel: On palpation, there is diffuse tenderness over the glenohumeral joint, and this extends to the trapezius and interscapular area owing to attempted splinting of the painful shoulder.

Move: In true frozen shoulder there is almost complete loss of external rotation. This is the pathognomonic sign of a frozen shoulder.1 2 w1-w3 Confirming that external rotation is impossible with active and passive movements is important. For example, if external rotation was easily possible with the help of the doctor, we would consider the diagnosis of a large rotator cuff tear, which would require completely different management. In frozen shoulder, all other movements of the joint are reduced, and if movement occurs this usually comes from the thoracoscapular joint.

Three phases of clinical presentation

Painful freezing phase

Duration 10-36 weeks. Pain and stiffness around the shoulder with no history of injury. A nagging constant pain is worse at night, with little response to non-steroidal anti-inflammatory drugs

Adhesive phase

Occurs at 4-12 months. The pain gradually subsides but stiffness remains. Pain is apparent only at the extremes of movement. Gross reduction of glenohumeral movements, with near total obliteration of external rotation

Resolution phase

Takes 12-42 months. Follows the adhesive phase with spontaneous improvement in the range of movement. Mean duration from onset of frozen shoulder to the greatest resolution is over 30 months

What's the natural history of frozen shoulder?

Although the natural history of frozen shoulder is for ultimate resolution, this may not be complete. Reeves, in a prospective study of 41 patients with 5-10 years' follow-up, found that 39% had full recovery, 54% had clinical limitation without functional disability, and 7% had functional limitation.2 Shaffer et al showed that 50% of his 61 patients with frozen shoulder had some degree of pain and stiffness an average of seven years after onset of the disease.3

Secondary frozen shoulder

Frozen shoulder can be a primary or idiopathic problem or it may be associated with another systemic illness. By far the most common association of a secondary frozen shoulder is diabetes mellitus.4 w7 w8 The incidence of frozen shoulder in diabetes patients is reported to be 10%-36%.46 The incidence in type 1 and type 2 diabetes is similar.5 Unfortunately, frozen shoulder in diabetes is often more severe and is more resistant to treatment.7 Moren-Hybbinette et al reported on the natural history of the diabetic painful stiff shoulder and found a restriction in the range of motion in 35 (65%) of 54 shoulders at a mean follow-up of 29 months.8

Bunker et al have shown an association with Dupuytren's disease in the hand, proposing that the contracting shoulder tissue itself represents a form of fibromatosis.6 w9 Much more rarely, secondary frozen shoulder may be associated with conditions such as hyperthyroidism,w10 hypothyroidism,w11 and hypoadrenalism.w12

Additional associations include Parkinson's disease,w13 cardiac disease, pulmonary disease, and stroke,w14-w16 although the pathological condition here may be different from idiopathic frozen shoulder. Clearly, in the case of stroke, shoulder stiffness may be simply the result of muscle spasticity in the shoulder region.

Frozen shoulder has also been reported subsequent to non-shoulder surgical procedures, such as cardiac surgery,w17 cardiac catheterisation through the brachial artery,w18 neurosurgery,w19 and radical neck dissection.w20

Laboratory investigations and radiology in frozen shoulder

There are few specific laboratory tests or radiological markers for frozen shoulder, and the diagnosis is essentially clinical. Immunological studies (such as human leucocyte antigen B27), C reactive protein, and erythrocyte sedimentation rate are all normalw21-w23 and would be measured only to exclude other conditions. Most orthopaedic surgeons would not investigate a frozen shoulder beyond a plain x ray. When plain radiographs of the frozen shoulder are taken they may well be reported as normal, although they may show periarticular osteopenia as a result of disuse.9 w24

Contrast technetium-99m diphosphonate bone scan shows an increased uptake on the affected side in 92% of patients compared with the opposite side or with controls.9 Arthrography shows characteristic findings of limitation of capacity of the shoulder joint (5-10 ml compared with 25-30 ml in the normal joint) and a small or non-existent dependent axillary fold.9 10 w24 However, in most units, arthrography is a historical investigation in frozen shoulder. Magnetic resonance imaging may show a slight thickening in the joint capsule and the coracohumeral ligament.w25


The aetiology of frozen shoulder remains unclear. The disease process particularly affects the anterosuperior joint capsule and the coracohumeral ligament.11 Arthroscopy shows a small joint with loss of the axillary fold and tight anterior capsule, mild or moderate synovitis, and no adhesions.12 w4 w26

Neviaser and Neviaser have described an arthroscopic four stage classification for the frozen shoulder,10 and Hannafin et alw5 have described a correlation between the arthroscopic stage, the clinical examination, and the histological appearance of the tissues.

Disagreement prevails about whether the underlying pathological process is an inflammatory condition,w5 w27 a fibrosing condition,6 or even an algo-neurodystrophic process.w28

Evidence shows a synovial inflammation with subsequent reactive capsular fibrosis. A dense matrix of type I and type III collagen is laid down by fibroblasts and myofibroblasts in the joint capsule. Subsequently, this tissue contracts.

Increased growth factors, cytokines, and expression of matrix metalloproteinases in capsular biopsy specimens obtained from patients with primary and secondary frozen shoulder indicate that these are involved in the inflammatory and fibrotic cascades seen in frozen shoulder.w27 w29

Cytokines and growth factors are involved in the initiation and termination of repair processes in musculoskeletal tissues through regulating fibroblasts, and the remodelling process is controlled by matrix metalloproteinases and their inhibitors.w29 w30 An association between frozen shoulder and Dupuytren's disease has been identified,6 13 and this may be related to matrix metalloproteinase inhibitors.w31

How should I treat it?

Educating patients helps to reduce frustration and encourages compliance. An explanation that the condition will spontaneously resolve and stiffness will greatly reduce helps. However, it is important to emphasise that the full range of motion may never recover. Ideally, the treatment of frozen shoulder should be tailored to the stage of the disease.

Treatment in the painful freezing phase

During the initial painful freezing stages, treatment is directed at pain relief. The patient is encouraged to use pain as a guide to limit activity, with all pain free activities allowed and all painful activities avoided.

It is traditional to give patients non-steroidal anti-inflammatory drugs (NSAIDs) if they can tolerate these. Where necessary these should be supplemented with other analgesics. There are, however, no randomised controlled trials that confirm the effectiveness of NSAIDs in the specific condition of frozen shoulder.


Dierks et al described a prospective study of 77 patients that compared exercise within the limits of pain with intensive physiotherapy in patients with frozen shoulder.14 They found better results with exercise performed within the limits of pain (64% reached near normal, painless shoulder movements at 12 months and 89% at 24 months) than with intensive physiotherapy (63% achieved a similar result at 24 months).

Steroid injection

Hazelman performed a meta-analysis on the use of intra-articular steroids and reported that the success of the treatment depends on the duration of symptoms–patients who receive the injection earlier in the course of the disease recover more quickly.15

Early treatment with a steroid injection into the intra-articular glenohumeral joint may reduce the synovitis, thus shortening the natural history of the disease.w5 De Jong et al have reported that the response to steroid injection is dose dependent.16

In a randomised placebo controlled trial, Carette et al compared the effectiveness of physiotherapy alone with a single intra-articular steroid injection given under x ray control.17 This study also looked at patients treated with physiotherapy and steroid injection in combination and a fourth, placebo group treated with a saline injection. The authors concluded that when used alone, supervised physiotherapy is of limited benefit, but that a single steroid injection in combination with physiotherapy is effective in reducing both pain and disability associated with frozen shoulder.

X ray control is not normally available for a joint injection in primary care. However, in a separate study, Van der Wind et al showed that steroid injection by a general practitioner to be more effective than physiotherapy alone at six weeks.18

Other treatment modalities

Oral steroids have been proposed as a treatment for frozen shoulder: Buchbinder et al19 described a double blind, randomised controlled trial on a series of 50 patients. In this study, oral steroids initially improved the frozen shoulder, but their effects did not last beyond six weeks. The adverse side effects of oral steroids are well documented, and they should not be regarded as routine treatment for this condition.

Suprascapular nerve blocks20 may be beneficial in terms of pain relief (but not movement), and repeated joint distension may improve movement.21 w32

Treatment during the adhesive phase

Intra-articular steroid injections are not indicated in the adhesive phase as the inflammatory stage of the disease has passed. More aggressive stretching exercises will be tolerated and should be the focus of treatment, with the aim of regaining the range of motion. Low load, prolonged stretches produce plastic elongation of tissues as opposed to the high tensile resistance seen with high load, brief stretches.22 w33

Manipulation under anaesthesia

For patients who are unable to tolerate the pain and disability associated with the condition, manipulation under anaesthesia23 24 w34 w35 is the most reliable way to improve the range of movement in a frozen shoulder. It is indicated if the functional disability persists in spite of adequate non-operative treatment for six months.10 24 w36 Manipulation under anaesthesia generally results in notable improvement in shoulder function and range of motion within three months.w34

Surgical release

More recently, arthroscopic release of the capsule has been advocated to allow a more controlled release of the contracted capsule than manipulation under anaesthesia.25 w37 This is required if manipulation fails to release the capsule, which is a common problem in frozen shoulder in diabetes.w37 w38 Arthroscopic release also avoids reported complications associated with manipulation, such as fracture of the humerus24 and iatrogenic, intra-articular shoulder lesions.w39 Arthroscopic release and synovectomy in the painful freezing phase of the disease may be effective in controlling the progression of the disease, if synovitis is an essential factor in the development of frozen shoulder.w5


Frozen shoulder is a common, sometimes painful, but ultimately self limiting, condition that is usually managed in the primary care setting with a combination of analgesics, injections, and physiotherapy. Formal investigations are usually normal, and the diagnosis is essentially clinical. Most cases can be managed in the primary care setting. Educating patients plays an important part in the management of the condition. A minority of patients require referral to an orthopaedic specialist, where manipulation under anaesthesia is the most common method of treatment. Arthroscopic surgical release has proved itself to be useful in refractory cases. Irrespective of the treatment given, a high proportion of patients with frozen shoulder do not regain a full range of motion.


  • Embedded Image Additional references w1-w39 are on Contributors: RD proposed the article and mapped out an overview and also provided the illustrations. SC wrote the first draft and recruited SM, a consultant surgeon with a special interest in shoulder surgery, who rewrote the first draft and served as senior technical adviser on the project as a whole. SC is guarantor.

  • Competing interests None.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
View Abstract