CLINICAL EVIDENCE REVIEW · PROGNOSIS
Frozen shoulder often improves substantially, but “self-limiting” is an incomplete prognosis. Recovery is variable, pain and function do not change on the same timetable, and some people retain measurable stiffness or disability for years.
Traditional freezing, frozen and thawing phases describe recognisable patterns but are not reliable calendar stages. A systematic review found no convincing evidence that untreated frozen shoulder consistently progresses through defined phases to complete resolution. Diabetes is strongly associated with developing frozen shoulder and may predict poorer outcomes, although prognosis evidence is limited in certainty. Thyroid disease is also associated in several studies, but estimates vary.
Why prognosis needs careful language
People usually ask two questions: “How long will this last?” and “Will my shoulder return to normal?” Research cannot provide a precise answer for an individual. Diagnostic criteria vary between studies, treatment is rarely absent, starting severity differs and outcomes are measured in incompatible ways. A study reporting normal daily function may still find a range deficit; another may define recovery through pain or a questionnaire.
The phrase “it will go away in 18 to 24 months” converts a broad historical description into a deadline. When that deadline passes, patients may believe that something has gone wrong or that they failed rehabilitation. A better prognosis combines reassurance - substantial improvement is common - with honest uncertainty about timing, residual symptoms and treatment needs.
The classic phase model: useful map, unreliable timetable
The freezing, frozen and thawing model remains useful when it helps clinicians explain changing priorities. It becomes harmful when stage is assigned solely from duration or when treatment is withheld because a calendar says recovery has not begun. Pain can improve before movement; function can improve through adaptation while passive range remains restricted; strength may lag behind both.
What natural-history research actually shows
The 2017 systematic review Natural history of frozen shoulder: fact or fiction? examined studies with untreated comparison groups and at least six months of follow-up. Only seven studies met inclusion criteria. Low-quality evidence suggested some but incomplete range improvement after one to four years, and no evidence supported a uniform progression through recovery phases to full resolution without treatment. In several datasets, improvement was greater earlier rather than being concentrated in a late “thawing” phase.
Long-term observational studies sometimes report more favourable outcomes. One cohort followed patients for two to 27 years and found that most recovered normal or near-normal function and movement without injection, manipulation or surgery. However, retrospective selection, missing records, treatment definitions and differing recovery thresholds limit direct comparison with the systematic review.
| Study feature | How it changes the result |
|---|---|
| Definition of frozen shoulder | Broader criteria can include other stiff or painful shoulder disorders with different trajectories. |
| Definition of recovery | Normal function, no pain and full measured range are not equivalent endpoints. |
| Follow-up length | A deficit at two years may become less important later; long follow-up also increases attrition. |
| Treatment exposure | “Natural history” is difficult to isolate when analgesia, exercise or other care occurs. |
| Starting population | Primary-care cases may differ from people referred to specialist services. |
| Missing participants | People who recover or remain unwell may be differently likely to return for follow-up. |
Recovery is multidimensional
One shoulder can be improving and still be restricted. Pain, sleep, movement, strength, confidence and participation should be tracked separately. An intervention may accelerate pain relief without changing final range; a person may regain independence by adapting tasks before external rotation returns; another may recover movement yet remain deconditioned or fearful of loading.
Conceptual display only: bar lengths illustrate different recovery domains and are not numerical outcome estimates.
How common is frozen shoulder?
General-population prevalence is frequently quoted as 2% to 5%, but this range should be treated as an approximation rather than a universal fact. Estimates depend on diagnostic definition, age, healthcare setting and whether studies measure current disease, lifetime experience or coded consultations. A Dutch general-practice estimate often cited in reviews reported a cumulative incidence around 2.4 per 1,000 people per year, but one healthcare system should not be assumed to represent the world.
Frozen shoulder is uncommon before 40 and is seen most often in midlife and later adulthood. Many samples include more women than men, although sex differences are not consistent enough to diagnose from demographics. Bilateral involvement can occur simultaneously or sequentially, and the opposite shoulder may become affected years later.
Age, diabetes and a typical movement pattern increase suspicion, but common risk factors cannot replace examination or exclude other causes of a painful stiff shoulder.
Diabetes: the most established association
Diabetes has the clearest and most consistently reported association with frozen shoulder. Meta-analytic evidence suggests people with diabetes have several-fold higher odds of developing the condition, while frozen shoulder prevalence within diabetic populations is substantially higher than in the general population. Exact estimates vary because studies use different diabetes and shoulder definitions.
Possible mechanisms include advanced glycation end-products, collagen cross-linking, microvascular change and altered inflammatory or fibrotic signalling. These mechanisms are plausible but do not allow a clinician to attribute an individual shoulder directly to glucose exposure.
A systematic review of diabetes as a prognostic factor found a general trend toward worse pain, movement and multidimensional outcomes, but certainty ranged from very low to moderate and many studies had high risk of bias. The practical implication is closer monitoring and realistic discussion, not therapeutic pessimism. People with diabetes still improve and should not be told that recovery is unlikely.
Thyroid disease and other associations
Several cohort and case-control studies report an association between thyroid disease - particularly hypothyroidism - and frozen shoulder, and the 2025 guideline recognises this evidence. Meta-analyses do not all reach the same conclusion, reflecting differences in populations and definitions. Routine testing should follow the person’s broader clinical context and local medical guidance rather than the shoulder diagnosis alone.
Associations have also been reported with dyslipidaemia, cardiovascular and cerebrovascular disease, Parkinson’s disease, Dupuytren’s disease and other metabolic or fibrotic conditions. Association does not establish causation. Detection bias, age, healthcare use and shared metabolic factors can produce or amplify relationships.
| Factor | Current interpretation | Clinical implication |
|---|---|---|
| Diabetes | Consistent association with onset; possible poorer outcomes, but prognostic certainty is limited. | Ask, monitor trajectory, discuss glucose effects when steroid is considered, and avoid deterministic language. |
| Thyroid disease | Association reported in several studies; pooled findings vary. | Consider medical context and symptoms rather than universal screening by shoulder clinicians. |
| Previous frozen shoulder | Contralateral disease is recognised and recurrence patterns vary. | Educate about new symptoms without creating expectation that the other shoulder must freeze. |
| Surgery, trauma or immobilisation | Important contexts for secondary stiffness. | Use the relevant post-operative or trauma diagnosis and precautions; do not automatically apply primary-frozen-shoulder evidence. |
| Age and sex | Most common in midlife; many cohorts include more women. | Useful background probability, never diagnostic criteria. |
| Other systemic associations | Observed relationships vary in strength and causal certainty. | Investigate according to symptoms, medical history and scope, not an indiscriminate panel. |
Primary and secondary frozen shoulder do not share one prognosis
Studies of primary frozen shoulder often exclude people with fracture, surgery, major cuff tears or arthritis. Applying those recovery estimates to secondary stiffness can mislead. Post-operative recovery depends on the original procedure, tissue healing, immobilisation, complications and surgical precautions. Post-traumatic stiffness may reflect bony injury, instability or soft-tissue damage in addition to capsular restriction.
Clinical records and patient information should state which population the prognosis describes. “Frozen shoulder usually improves” is more defensible when followed by “your recovery also depends on the reason it developed and the other findings in your shoulder.”
Factors that may influence the lived recovery
Biological factors are only part of prognosis. Sleep loss, work demands, caregiving, access to treatment, health literacy, distress, fear of movement and the ability to modify daily tasks can change disability. These factors do not mean symptoms are psychological; they influence how a painful, restricted shoulder affects life and how manageable a recovery plan becomes.
Monitoring recovery without creating false deadlines
Agree a small set of repeatable outcomes and review them over meaningful intervals. Week-to-week fluctuation is common and can obscure a longer trend. Use pain and sleep measures, patient-selected function, disability questionnaires where appropriate, active and passive movement measured consistently, and later strength or load tolerance.
| Observed course | Possible interpretation | Reasonable response |
|---|---|---|
| Pain improving; stiffness unchanged | Meaningful progress may be occurring before range recovery. | Progress activity and mobility according to irritability. |
| Function improving; measured range changes little | Adaptation, confidence or strength may be improving. | Continue to value patient-important gains; check measurement consistency. |
| Repeated flares with treatment | Dose may exceed current tolerance. | Reduce and rebuild dosage; reconsider purpose of each intervention. |
| No meaningful progress over a suitable review period | Slow natural course, barrier, competing pathology or incorrect diagnosis may contribute. | Review diagnosis, adherence, comorbidity, treatment options and referral need. |
| New weakness, neurological or systemic change | Not simply “a slow thaw.” | Prompt reassessment through the appropriate pathway. |
Communicating prognosis well
“It always resolves completely in two years.”
“Diabetes means your shoulder will not recover.”
“You are still frozen, so nothing can improve yet.”
“Substantial improvement is common, but timing and completeness vary.”
“Diabetes may make recovery more difficult on average, so we will monitor progress closely.”
“Pain, movement and function can change on different timelines.”
Uncertainty should not remove hope. It should improve planning: protect sleep and participation, choose treatments according to current goals and irritability, and establish review points for changing course.
Implications for professional practice
- Replace deadlines with ranges and review points. Recovery is common but not clockwork.
- Describe stages as overlapping patterns. Do not infer biology or treatment readiness from duration alone.
- Measure several recovery domains. Pain, sleep, movement, strength and participation may diverge.
- Use risk factors probabilistically. Diabetes and thyroid disease alter likelihood; they do not establish diagnosis or destiny.
- Distinguish primary from secondary stiffness. Prognosis and precautions depend on context.
- Reassess an atypical trajectory. New findings or deterioration deserve diagnostic review, not simply more time.
Selected evidence and guidance
- Natural history of frozen shoulder: fact or fiction? A systematic review.
- The natural history of idiopathic frozen shoulder: a 2- to 27-year follow-up study.
- Clinical Practice Guidelines for Diagnosis and Non-Surgical Treatment of Primary Frozen Shoulder (2025).
- Diabetes as a risk factor for the onset of frozen shoulder: systematic review and meta-analysis.
- Diabetes as a prognostic factor in frozen shoulder: systematic review.
- Association between frozen shoulder and thyroid diseases.
Evidence checked: August 2026. Review earlier if a major cohort, systematic review or guideline materially changes prognosis estimates.