CLINICAL EVIDENCE REVIEW · GUIDELINES
Guidelines translate research, clinical expertise, patient priorities and health-system constraints into recommendations. They can improve consistency, but they do not remove uncertainty, replace clinical reasoning or guarantee that one pathway transfers unchanged to every patient, profession and country.
Contemporary guidance broadly supports clinical diagnosis, selective imaging, education, activity modification, individualised rehabilitation and shared decision-making. Intra-articular corticosteroid injection can be considered for short-term pain and function. Hydrodilatation and operative procedures are selective escalation options. Many rehabilitation and procedural questions remain supported by heterogeneous or low-certainty evidence.
Why a recommendation is not simply “the evidence”
A systematic review estimates what happened across included studies. A guideline panel must go further. It considers evidence certainty, clinical importance, harms, values, feasibility, equity, acceptability, cost and applicability. Two responsible guidelines can examine overlapping evidence and produce different recommendations because their populations, systems and decision thresholds differ.
Recommendation language therefore needs interpretation. “Recommend” may indicate that most informed patients should receive an intervention in a defined context. “Consider,” “may offer” or “conditional” signals greater dependence on individual priorities, resources or uncertainty. The strength of the wording should be read alongside the certainty rating and rationale, not in isolation.
From study result to clinical action
Skipping a step creates predictable errors. A statistically significant study result can be promoted as mandatory care without checking clinical importance. A guideline recommendation can be copied into a service without confirming staff competence or access. A local pathway can become outdated after the evidence or available treatments change.
Understanding evidence certainty
GRADE and related systems commonly classify certainty as high, moderate, low or very low. Low certainty does not mean an intervention is ineffective; it means confidence in the estimated effect is limited and future research may change the conclusion. High certainty does not mean every patient will respond.
| Certainty | Meaning | Practical implication |
|---|---|---|
| High | Further research is unlikely to change confidence in the effect estimate substantially. | Still assess applicability, harms and patient preference. |
| Moderate | The true effect is probably close to the estimate but could differ meaningfully. | Explain likely benefit and remaining uncertainty. |
| Low | Confidence is limited; the true effect may be substantially different. | Use conditional decisions, close monitoring and proportionate claims. |
| Very low | The effect estimate is highly uncertain. | Avoid certainty language; prioritise safety, reversibility, cost and informed preference. |
Certainty may differ by outcome. Evidence for short-term pain relief may be stronger than evidence for long-term function or harm. A recommendation should not be summarised using only the most favourable outcome.
Current guideline landscape
The 2025 clinical practice guideline for diagnosis and non-surgical treatment of primary frozen shoulder synthesised diagnostic and treatment evidence into background questions and 16 key recommendations. The British Elbow & Shoulder Society patient care pathway addresses service-level management and escalation. Primary- and community-care shoulder guidelines add referral and safety context. UK FROST supplies influential comparative evidence for secondary-care interventions.
These documents overlap but do not serve identical purposes. A non-surgical guideline does not answer every operative question. A national pathway incorporates resource and referral assumptions that may not apply internationally. A trial comparing three secondary-care strategies does not establish diagnosis or first-contact management.
Diagnosis and non-surgical recommendations with evidence grading.
Roles, referral thresholds, options and escalation.
Effects, harms and costs within an eligible study population.
Areas of broad contemporary agreement
| Domain | Broad principle | Important qualification |
|---|---|---|
| Diagnosis | Primary frozen shoulder is principally a clinical diagnosis involving restriction of active and passive movement. | No single movement, test or demographic factor confirms it. |
| Imaging | Use imaging selectively to exclude alternatives or clarify atypical presentations. | Ultrasound and MRI findings are supplementary, not stand-alone proof. |
| Education | Explain the condition, uncertainty, likely course, activity and treatment options. | Reassurance should not become a promise of full recovery by a fixed date. |
| Rehabilitation | Offer individualised movement, exercise and activity guidance. | Evidence does not establish one optimal exercise, intensity or visit schedule. |
| Manual therapy | May be used as an adjunct to support pain, movement or active care. | Do not present a specific technique as a necessary mechanical cure. |
| Corticosteroid injection | Can be considered for shorter-term pain and functional benefit. | Benefits commonly narrow over time; account for risks and diabetes. |
| Hydrodilatation | May be considered within an informed escalation discussion. | Protocols and additional benefit over steroid injection vary. |
| Procedures | MUA or ACR may be discussed for persistent unacceptable disability. | UK FROST found no clinically important 12-month superiority; burdens differ. |
Where recommendations remain unsettled
Uncertainty is concentrated around rehabilitation dosage, manual-therapy technique, the value and timing of hydrodilatation, optimal injection details, stage-specific treatment and the sequence of escalation. Studies frequently use small samples, inconsistent diagnostic criteria, short follow-up and combined interventions that make the active component difficult to isolate.
Select by purpose, irritability, function and response.
Record range, force, volume, frequency and recovery rule.
Consider burden, trajectory, care tried, risk and preference.
Benefits, harms, cost and recovery demands differ.
Population, intervention, comparator and timeframe
Before applying a recommendation, identify who was studied. Primary frozen shoulder is not interchangeable with post-operative, post-traumatic or arthritic stiffness. Secondary-care trial participants are not necessarily representative of newly presenting primary-care patients. Diabetes may alter risk, response and safety considerations.
Then examine the comparator. An intervention may outperform no treatment yet offer no meaningful advantage over another active treatment. A corticosteroid injection can outperform physiotherapy for short-term pain without proving that rehabilitation has no role in longer-term function. A statistical difference at one follow-up does not establish a clinically important or durable benefit.
| Question | Example | Risk if ignored |
|---|---|---|
| Does the patient match the population? | Primary versus secondary stiffness; first contact versus specialist referral. | Indirect evidence is treated as directly applicable. |
| Is the intervention actually available and competently delivered? | Image-guided injection, rehabilitation, hydrodilatation or surgery. | A paper pathway promises care the service cannot safely provide. |
| What was the comparator? | Placebo, no treatment, steroid, exercise or surgery. | Relative benefit is described without context. |
| Which outcome and timeframe improved? | Short-term pain versus long-term function. | Temporary effects become claims of superior recovery. |
| What harms and burdens were measured? | Adverse events, cost, visits, time off work and post-procedure care. | Benefit is separated from the treatment burden. |
Shared decision-making is not an optional extra
When benefits are modest, certainty is limited or trade-offs differ, patient preference becomes more important, not less. Shared decision-making requires balanced options, likely timeframe, uncertainty, harms, practical burden and the option of supported waiting. It is not achieved by presenting one recommendation and asking for agreement.
From recommendation to an observable service standard
Implementation means specifying who does what, for whom, under which safeguards and how the outcome will be known. “Provide education” is not auditable until the service defines essential content and documentation. “Offer physiotherapy” is incomplete without access, dosage review and escalation arrangements.
| Recommendation | Observable action | Measure | Review trigger |
|---|---|---|---|
| Use clinical diagnosis with selective imaging. | Document supporting pattern, differentials and reason for imaging or non-imaging. | Record audit and inappropriate imaging rate. | Atypical course, new findings or high variation between clinicians. |
| Provide education and activity guidance. | Supply consistent information on condition, prognosis, options and safety-netting. | Patient understanding and documented shared plan. | Repeated misconceptions, low confidence or poor access. |
| Individualise rehabilitation. | Record functional goal, starting dose and symptom-response rule. | Pain, sleep, function, movement and adherence. | Repeated flare, no meaningful progress or changed presentation. |
| Consider injection. | Document target outcome, consent, diabetes plan and follow-up. | Target outcome, adverse effects and rehabilitation access. | No benefit, harm, repeat request or diagnostic uncertainty. |
| Escalate selectively. | Confirm diagnosis, previous care, disability, options, risk and preference. | Waiting time, decision quality, outcomes, harms and re-intervention. | Unwarranted variation, access inequity or complication pattern. |
Audit should include access and harm
A service can appear guideline-compliant while producing inequitable care. Audit should ask whether people with diabetes, language needs, disability, rural location or limited insurance receive the same opportunity for explanation and appropriate options. Waiting time can change irritability and treatment preference.
Outcome audit should include adverse events, treatment burden and patient-selected function alongside range and standard scores. A service that performs more procedures is not necessarily more effective; a service that avoids procedures is not necessarily patient-centred if severe disability is left unsupported.
Guideline disagreement is information
When documents disagree, identify the reason rather than choosing the recommendation that matches prior preference. Publication date, included evidence, panel composition, healthcare system, cost assumptions and scope can all explain divergence. A rehabilitation guideline and surgical pathway may emphasise different outcomes without being mutually exclusive.
Where no direct evidence exists, guidelines may rely on expert consensus. Consensus can be useful, particularly for safety and service organisation, but it should be labelled and updated when better evidence emerges.
Education, accreditation and evidence are different claims
Professional accreditation or continuing-education approval indicates that a course meets the relevant organisation’s educational and administrative requirements. It does not mean every intervention discussed is endorsed as effective, nor does it extend a learner’s legal scope. Educational content should distinguish established recommendation, research finding, expert interpretation and practical demonstration.
Commercial availability is also not evidence. A course, device, manual technique, injection or surgical service should not receive stronger claims because it is being sold. Conflicts of interest and material updates should be visible.
Editorial implementation for FrozenShoulder.com
Link to guidelines, systematic reviews and pivotal trials.
Separate recommendation strength from confidence in evidence.
Do not merge primary, secondary, post-operative and post-traumatic stiffness.
Include harms, burdens, indirectness and unresolved questions.
Reassess after major evidence, safety or pathway changes.
Paid status must not remove balanced claims or alternatives.
Implications for professional practice
- Read beyond the recommendation sentence. Check certainty, population, comparator, outcomes and timeframe.
- Distinguish statistical from clinical importance. Small average differences may not justify additional burden.
- Translate guidance into observable action. Define roles, documentation, follow-up and review triggers.
- Make uncertainty visible in consent. Conditional recommendations require patient priorities.
- Audit access, outcome, burden and harm. Process compliance alone is insufficient.
- Set an evidence-review date. A pathway becomes less reliable when no one owns its maintenance.
Selected guidance and pivotal evidence
- Clinical Practice Guidelines for Diagnosis and Non-Surgical Treatment of Primary Frozen Shoulder (2025).
- British Elbow & Shoulder Society patient care pathways and guidelines.
- Shoulder Pain Diagnosis, Treatment and Referral Guidelines for Primary, Community and Intermediate Care.
- UK FROST multicentre randomised clinical trial.
- Comparison of treatments for frozen shoulder: systematic review and meta-analysis.
- Natural history of frozen shoulder: fact or fiction? Systematic review.
Evidence checked: August 2026. Review earlier if a major guideline, safety communication or pivotal trial materially changes recommendations.