Frozen Shoulder: Causes, Recovery Time, and What Treatment Actually Changes

Key Takeaways: Frozen Shoulder

Frozen shoulder is a clinical diagnosis. It typically presents as slowly developing pain near the deltoid together with a disproportionately severe loss of passive external rotation, and routine X-rays have a low yield for other pathology.[1]
Diabetes is clearly associated with frozen shoulder. In a meta-analysis of six case–control studies (5,388 people), the odds of frozen shoulder were 3.69 times higher in people with diabetes, although most included studies were at high risk of bias.[2]
Thyroid disease is also associated. In one case–control study, hypothyroidism was diagnosed in 27.2% of frozen shoulder patients versus 10.7% of controls.[3]
“It always thaws on its own” is an oversimplification. A systematic review found no evidence that untreated frozen shoulder reliably passes through recovery phases to full resolution, and most improvement in trials happened early rather than late.[4]
Most people do well long term, but not everyone fully recovers. At a mean of 4.4 years after onset, 59% had normal or near-normal shoulders and 41% had ongoing symptoms, in most cases mild to moderate (35% of all shoulders) rather than severe (6%).[5]
A steroid injection into the joint has the most consistent short-term support. In one meta-analysis, its superiority appeared to last for as long as 6 months; the BESS guideline analysis found no difference after 3 months.[6][1]
Surgery is not clearly better at 12 months. In the UK FROST trial (503 patients), neither manipulation under anaesthesia nor arthroscopic capsular release was clinically superior to early structured physiotherapy with a steroid injection.[7]
In practice, the cause usually remains open. In my clinical experience, there is rarely a clear trigger, and even when diabetes or thyroid disease is present, a cause-and-effect relationship cannot be shown in an individual patient.
Rotator cuff problems are an important look-alike. In my clinical experience, frozen shoulder pain is often more diffuse than impingement pain, with restriction in several directions, and an ultrasound can help look for other causes such as impingement.
Steroid injection benefit varies. In my practice, injections are usually given under ultrasound guidance, but the relief is often temporary and varies between patients; trials have not shown a clear outcome difference between guided and blind injections.[1]
Activity within pain limits. I usually tell patients they can do everything within the limits of pain, which is in line with a non-randomised study in which exercise within pain limits did better than intensive physical therapy with passive stretching.[8]

Introduction: Frozen Shoulder

Frozen shoulder, also called adhesive capsulitis, is a painful and often disabling condition in which the capsule of the shoulder joint contracts and both active and passive movement become restricted, with loss of external rotation being especially characteristic.[2] It is generally described as presenting between the ages of 50 and 60 and rarely before 40.[2]

What makes frozen shoulder frustrating is not only the pain, but the time scale. Frozen shoulder is traditionally described as “freezing” and then “thawing” by itself. As discussed below, the evidence for complete recovery without treatment is weaker than this description suggests, and the treatment evidence is more nuanced than either “just wait” or “fix it with surgery.”

This article covers what frozen shoulder is, who gets it, what happens inside the joint, how it is diagnosed, how long it realistically lasts, and what the main treatments — injections, physiotherapy, hydrodilatation and surgery — have actually been shown to change.

In my clinical practice, I see frozen shoulder from time to time. Typically, the shoulder pain has started without any clear explanatory factor or injury. Some patients already recognise the picture, because their other shoulder has frozen before. From a clinical perspective, the diagnosis is both good and bad news for the patient, and both sides are worth discussing openly.

What Is Frozen Shoulder?

The classic description of frozen shoulder rests on three features: insidious shoulder stiffness, severe pain that is present even at night, and near-complete loss of both passive and active external rotation.[9] Routine blood tests are generally normal, and there are few specific laboratory or imaging markers, so the diagnosis remains essentially clinical.[9]

Frozen shoulder is often divided into three overlapping stages — painful, freezing and thawing. The 2025 British Elbow and Shoulder Society (BESS) guideline points out, however, that the stages overlap so much that their usefulness for treatment decisions is limited.[1]

How Common Is Frozen Shoulder, and Who Gets It?

Prevalence estimates vary between sources. The BESS guideline cites a prevalence of 2–5% and describes frozen shoulder as typically affecting women between 45 and 60, with a peak age of 56.[1] The authors of a 2023 BMJ Open review give a lower general-population figure of around 0.75%, and note that women (58%) are affected more often than men (42%) and that the other shoulder becomes involved in 6–17% of patients.[2]

Diabetes and Frozen Shoulder

The same review pooled six case–control studies including 5,388 people and found that people with diabetes had 3.69 times the odds of developing frozen shoulder (95% CI 2.99 to 4.56) compared with people without diabetes.[2] Two cohort studies pointed in the same direction, with hazard ratios of 1.32 and 1.67.[2]

This evidence has limits. Seven of the eight included studies were judged to be at high risk of bias, and unmeasured confounding was the main weakness, so the association is clear but causation has not been established.[2] The authors also noted that studies on whether glycaemic control affects the risk have produced conflicting results.[2]

Thyroid Disease and Frozen Shoulder

Thyroid disease has also been associated with frozen shoulder. In a case–control study of 93 frozen shoulder patients and 151 controls, a diagnosis of hypothyroidism was more common in the frozen shoulder group (27.2% vs. 10.7%), and frozen shoulder remained independently associated with hypothyroidism in multivariate analysis (odds ratio 3.1).[3] TSH levels were also higher in patients with bilateral or severe frozen shoulder.[3]

The BESS guideline additionally lists hypothyroidism, Dupuytren’s disease, Parkinson’s disease, osteoporosis, stroke, hyperlipidaemia, and previous cardiac or neurosurgery among the conditions that have been associated with frozen shoulder.[1]

From a clinical perspective, diabetes and thyroid disease are worth taking into account in a patient with frozen shoulder. In practice, however, if a patient’s diabetes or hypothyroidism is already known and treated, I see little point in making the patient feel responsible for the shoulder: even though the risk factors are recognised, a cause-and-effect relationship cannot be demonstrated in an individual patient. In my experience in Finnish occupational health care, patients are often screened with laboratory tests at regular intervals, so diabetes and thyroid disease are often already known. If a patient has not had laboratory tests, in my practice we usually arrange a basic laboratory panel, partly as a routine periodic check-up. This can reveal problems in these areas, although in my view it does not directly change the treatment of the frozen shoulder itself.

What Happens Inside a Frozen Shoulder?

The exact aetiology of frozen shoulder remains unclear.[2] In a classic histology study, 50 of 935 consecutive patients referred with shoulder pain met the criteria for primary frozen shoulder. Tissue from 12 of these patients, who underwent surgery after failing to improve with conservative treatment and manipulation, showed active fibroblast proliferation with partial transformation into myofibroblasts, laying down dense collagen in a pattern very similar to Dupuytren’s disease of the hand, with no inflammation and no synovial involvement in these specimens.[10] Because these were selected surgical cases, the findings may not apply to all patients or all stages of the condition.

Later research has also pointed to an inflammatory component. A 2020 scoping review of 48 original studies suggested that inflammatory cytokines — with transforming growth factor-β1 playing a prominent role — together with mechanical stress stimulate fibroblast proliferation and their differentiation into myofibroblasts, although the exact sequence of events remains uncertain.[11] The same review proposed that a state of low-grade inflammation, such as that associated with diabetes, cardiovascular disease and thyroid disorders, predisposes to frozen shoulder.[11]

For diabetes specifically, hypothesised mechanisms include glycation of collagen through advanced glycation end products and hyperglycaemia-driven inflammation, but these remain hypotheses rather than proven causes.[2]

Patients often ask what has caused their frozen shoulder. In my clinical experience, the honest answer is usually that there is no clear factor explaining why this particular patient has developed it at this particular time. In practice, I usually explain it to patients as an internal inflammatory process in the shoulder joint, and for many patients the exact cause remains open.

How Frozen Shoulder Is Diagnosed

According to the BESS guideline, frozen shoulder is diagnosed from the history and examination, and typically presents as slowly developing pain near the deltoid insertion combined with a disproportionately severe loss of passive external rotation.[1] The key distinction is between active and passive movement: if the examiner can easily move the arm into external rotation, a large rotator cuff tear becomes a more likely explanation than frozen shoulder.[9]

In my clinical experience, shoulder pain that begins without clear trauma is most often a rotator cuff problem, such as supraspinatus-related impingement, but sometimes it turns out to be frozen shoulder instead. Frozen shoulder often differs from impingement in that the pain is not so clearly linked to abduction; it can be more diffuse and felt in several directions of movement. Rotations may be tender, both abduction and flexion can be restricted and painful, and the range of motion is often clearly limited. A patient may, for example, have difficulty reaching behind the back. In my own practice, I often request an ultrasound in these situations, mainly to look for other causes such as supraspinatus impingement: if no such explanation is found, the radiologist, together with the clinical findings, can often consider adhesive capsulitis the most likely explanation. The diagnosis itself still rests on the history and examination.

Routine X-rays have a low yield for other pathology, such as osteoarthritis or tumours, but the guideline notes that in patients over 40 with pain and stiffness, an underlying tumour can occasionally be misdiagnosed as frozen shoulder.[1] Because routine blood tests are normal in frozen shoulder itself,[9] blood test results may be more useful for identifying associated conditions than for making the diagnosis.

In my clinical experience, some patients may also have had some kind of injury in the background, which can confuse the diagnosis. In unclear cases, an MRI of the shoulder may be taken, although it is not always necessary for making the diagnosis itself.

Severe pain, even at night, is one of the hallmarks of the condition, and Codman’s original description already noted difficulty sleeping on the affected side.[9]

How Long Does Frozen Shoulder Last?

The traditional teaching, first proposed in the 1940s, is that frozen shoulder is self-limiting.[4] The BESS guideline still describes it as a self-resolving condition and therefore frames treatment around improving pain and function as quickly as possible, through shared decision-making.[1]

The evidence for complete spontaneous recovery is thinner, however. A 2017 systematic review of studies with untreated comparison groups found low-quality evidence that no treatment produced some, but not complete, improvement in range of motion over 1 to 4 years of follow-up.[4] It found no evidence supporting progression through recovery phases to full resolution, and moderate-quality evidence from three randomised trials showed that most improvement occurred early rather than late.[4]

Long-term follow-up gives a realistic middle ground. In a study of 269 shoulders in 223 patients with primary frozen shoulder, followed for a mean of 4.4 years from symptom onset, 59% had normal or near-normal shoulders, 35% had mild to moderate persistent symptoms, and 6% had severe symptoms with pain and functional loss.[5] Pain was the most common complaint among those with ongoing symptoms.[5] Patients with the most severe symptoms at onset had the worst long-term outcomes, 20% reported symptoms in both shoulders, and there were no recurrent cases.[5]

In practical terms, most people recover substantially, but “it will go away on its own” is not a guarantee of a full return to normal.

From a clinical perspective, this is why the diagnosis is both good and bad news. The good news is that frozen shoulder is generally expected to improve over time, and in my clinical experience it rarely leads to surgery. The bad news is that it is often very long-lasting, and months of pain can mean a long period of suffering for the patient. In my clinical experience, patients nevertheless often find some relief in knowing that the condition is expected to settle on its own, even if, as the research above shows, not always completely.

Frozen Shoulder Treatment: What the Evidence Shows

In my practice, I usually arrange an orthopaedic consultation for these patients, and the orthopaedic surgeon makes the final decision on the treatment plan. In my experience, management is most often conservative, and I often refer the patient to physiotherapy. In some situations, a corticosteroid injection or manipulation under anaesthesia may be tried. Sometimes the radiologist may also inject corticosteroid into the shoulder.

Corticosteroid Injection

Of the non-surgical options, a steroid injection into the shoulder joint has the most consistent evidence. A 2020 JAMA Network Open meta-analysis of 65 studies (4,097 participants) found that intra-articular corticosteroid was the only intervention associated with both statistical and clinical superiority over other non-surgical treatments in the short term, and its superiority appeared to last for as long as 6 months.[6] The authors associated early injection in frozen shoulder of less than one year’s duration with better outcomes, and found that adding a home exercise programme and/or physiotherapy (electrotherapy or passive mobilisations) may be associated with added benefit.[6]

The BESS guideline reaches a similar but more cautious conclusion: compared with physiotherapy alone, low-volume steroid injections improved pain and range of motion for up to 3 months, with no difference after that point, although the long-term evidence is limited.[1] It found no outcome difference between ultrasound-guided and freehand injections.[1]

In my practice, I usually refer patients for an ultrasound-guided corticosteroid injection. The referral for the diagnostic ultrasound often already includes a possible ultrasound-guided injection into the shoulder joint, because in my clinical experience the injection reaches its target somewhat more reliably than when the doctor injects blind. I also give landmark-guided injections myself when ultrasound is not available or is not covered by the patient’s insurance or occupational health contract. From a clinical perspective, this is a question of accurate placement rather than proven outcome, since the guideline found no difference in clinical outcome between the two approaches.

The guideline also highlights an unresolved question. In placebo-controlled trials, a low-volume steroid injection improved external rotation at 3 months more than a saline or local anaesthetic injection, but there was no difference in pain or patient-reported outcomes; the authors note that even a placebo injection may have some therapeutic effect through distension of the capsule.[1]

In my clinical experience, the benefit from a corticosteroid injection also varies. The injection does not, as such, cure the condition, but some patients feel that it gives them temporary relief. In my view, it is hard to say how much of this is a placebo effect related to the injection itself and how much is genuine improvement, but because many patients experience at least short-term benefit, injections continue to be used.

For early, pain-predominant frozen shoulder, a 2024 network meta-analysis found that intra-articular corticosteroid injections improved pain and function at 12 weeks compared with physical therapy. Intra-articular platelet-rich plasma also showed benefit for both outcomes, whereas oral anti-inflammatory drugs and hyaluronate injections did not show a benefit.[12] Most included studies carried considerable risk of bias, so these findings should be read with caution.[12]

In my practice, I often prescribe anti-inflammatory painkillers early on, and many of my patients have found a long-acting option, such as etoricoxib (available in Finland), convenient. In my clinical experience, the benefit patients get from these medicines varies, which is in keeping with the limited trial evidence above. Sometimes the neck and shoulder region also becomes stiff as a secondary problem, and some patients take a muscle relaxant for that.

Physiotherapy and Exercise

Frozen shoulder is commonly treated with manual therapy and exercise, but the evidence for their independent effect is limited. A 2014 Cochrane review of 32 trials (1,836 participants) found moderate-quality evidence that six weeks of combined manual therapy and exercise probably produces less improvement at seven weeks than a glucocorticoid injection.[13] These figures come from one included trial: for treatment success (108 participants), 46% reported success with manual therapy and exercise, compared with 77% with injection.[13] The wider role of hands-on treatment alongside active rehabilitation is discussed in manual therapy for athletes.

The BESS guideline states that it is unknown whether supervised physiotherapy provides greater benefit than the natural history of frozen shoulder, because no randomised trials have compared the two.[1] It did find low-certainty evidence that physiotherapy after a steroid injection improves external rotation, pain and disability at 6 weeks, with no evidence of a difference in pain and disability at 6 months.[1]

Intensity may matter. In a prospective study of 77 patients, one group received intensive physical therapy including passive stretching and manual mobilisation, while the other received supportive therapy and exercises within pain limits (“supervised neglect”).[8] After 24 months, 89% of the supervised-neglect group had normal or near-normal painless function (Constant score ≥80), compared with 63% of the intensive group.[8] The study was not randomised, so the result should be interpreted with caution. The broader limits of stretching are covered in stretching for athletes.

Hydrodilatation (Hydrodistension)

Hydrodilatation, also called hydrodistension, is a high-volume (20 ml or more) steroid injection into the shoulder joint.[1] The BESS guideline supports it over physiotherapy or supportive therapy for short- to medium-term improvement in pain and abduction; in its meta-analysis, however, only the improvement in abduction at up to 3 months was statistically significant, while improvements in pain, disability and external rotation were not. It found insufficient evidence to comment on long-term outcomes.[1] Transient flushing and post-procedure pain were reported in 14% of patients in one trial.[1] A 2023 systematic review similarly concluded that hydrodilatation appears to provide a transient improvement in disability, but that its clinical relevance remains unclear.[14]

Surgery: Manipulation Under Anaesthesia and Capsular Release

Key evidence comes from UK FROST, which randomised 503 patients referred to secondary care with primary frozen shoulder, recruited from 35 UK hospital sites, to manipulation under anaesthesia (MUA) with a steroid injection, arthroscopic capsular release, or early structured physiotherapy with a steroid injection.[7] Early structured physiotherapy combined mobilisation techniques and a graduated home exercise programme, delivered over 12 sessions within up to 12 weeks.[7]

At 12 months, mean Oxford Shoulder Scores (0–48) were 38.3 after MUA, 40.3 after capsular release, and 37.2 after physiotherapy.[7] All between-group differences (1.05 to 3.06 points) were smaller than the predefined clinically important differences, so none of the three treatments was clinically superior.[7] Capsular release carried higher risks, with eight serious adverse events compared with two after MUA, while MUA was the most cost-effective option.[7]

The BESS guideline notes that 15% of the physiotherapy group in UK FROST required further treatment, and recommends shared decision-making before capsular release because its 12-month outcomes do not exceed the minimal clinically important difference compared with alternatives.[1]

Frozen Shoulder in Active People and Athletes

Clinical synthesis: the studies reviewed here were not conducted in athletes, so the evidence applies to active people by extension rather than directly. Two points are particularly relevant for those who train regularly.

First, frozen shoulder rarely presents before 40.[2] In a younger athlete, a stiff and painful shoulder is therefore less typical of frozen shoulder, and other causes are worth considering.

Second, the training mindset of “pushing through” may not suit a frozen shoulder. In the supervised-neglect study, exercise within pain limits was associated with better outcomes than intensive physical therapy with passive stretching and manual mobilisation, although the study was not randomised.[8]

Patients often ask whether they are allowed to move and exercise with a frozen shoulder. In practice, I usually tell them that they can do everything within the limits of pain. In my clinical experience, however, the shoulder is often painful enough that patients do not necessarily feel like doing sport anyway.

Conclusion: Frozen Shoulder

Frozen shoulder is a clinical diagnosis built on night pain, gradual stiffness and a marked loss of passive external rotation. Diabetes is clearly associated with it, with 3.69 times the odds in pooled case–control data, and thyroid disease is also linked.[2][3]

Most people improve substantially, but the idea that frozen shoulder always fully “thaws” is not well supported: 59% had normal or near-normal shoulders at a mean of 4.4 years, and most of the improvement tends to happen early.[5][4] A steroid injection into the joint has the most consistent short-term evidence, one non-randomised study favoured exercise within pain limits over intensive stretching and mobilisation, and surgery has not been shown to be clinically superior to structured physiotherapy with an injection at 12 months.[6][8][7] The choice between options is usually made together with a clinician, weighing pain, function, risks and waiting times.

From a clinical perspective, the diagnosis is both good and bad news. In my practice, most patients are managed conservatively, with physiotherapy, painkillers and sometimes an ultrasound-guided steroid injection, and surgery is rarely needed; the hardest part for patients is often how long the pain lasts. In my clinical experience, the cause usually remains open, and I usually tell patients that they can stay active within the limits of pain.

Frozen Shoulder: Causes, Recovery Time, and What Treatment Actually Changes

Table of Contents

Key Takeaways: Frozen Shoulder

Introduction: Frozen Shoulder

What Is Frozen Shoulder?