Image‐guided glucocorticoid injection versus injection without image guidance for shoulder pain

Abstract

Background

Despite widespread use, our 2012 Cochrane review did not confirm that use of imaging to guide glucocorticoid injection for people with shoulder pain improves its efficacy.

Objectives

To update our review and assess the benefits and harms of image‐guided glucocorticoid injection compared to non‐image‐guided injection for patients with shoulder pain.

Search methods

We updated the search of the Cochrane Central Register of Controlled Trials (CENTRAL, via Ovid), MEDLINE (Ovid), Embase (Ovid) and clinicaltrials.gov to 15 Feb 2021, and the World Health Organisation International Clinical Trials Registry Platform (http://www.who.int/trialsearch/Default.aspx) to 06 July 2020. We also screened reference lists of retrieved review articles and trials to identify potentially relevant studies.

Selection criteria

We included randomised or quasi‐randomised controlled trials that compared image‐guided glucocorticoid injection to injection without image guidance (either landmark‐guided or intramuscular) in patients with shoulder pain (rotator cuff disease, adhesive capsulitis or mixed or undefined shoulder pain). Major outcomes were pain, function, proportion of participants with treatment success, quality of life, adverse events, serious adverse events and withdrawals due to adverse events. Minor outcomes were shoulder range of motion and proportion of participants requiring surgery or additional injections. There were no restrictions on language or date of publication.

Data collection and analysis

We used standard methodologic procedures expected by Cochrane.

Main results

Nineteen trials were included (1035 participants). Fourteen trials included participants with rotator cuff disease, four with adhesive capsulitis, and one with mixed or undefined shoulder pain. Trial size varied from 28 to 256 participants, most participants were female, mean age ranged between 31 and 60 years, and mean symptom duration varied from 2 to 23 months.

Two trials were at low risk of bias for all criteria. The most notable sources of bias in the remaining trials included performance bias and detection bias.

Moderate‐certainty evidence (downgraded for bias) indicates that ultrasound‐guided injection probably provides little or no clinically important benefits compared with injection without guidance with respect to pain (15 trials) or function (14 trials) at three to six weeks follow‐up. It may not improve quality of life (2 trials, low‐certainty evidence, downgraded due to potential for bias and imprecision) and we are uncertain about the effect of ultrasound‐guided injection on participant‐rated treatment success due to very low‐certainty evidence (downgraded for bias, inconsistency and imprecision).

Mean pain (scale range 0 to 10, higher scores indicate more pain) was 3.1 points with injection without image guidance and 0.5 points better (0.2 points better to 0.8 points better; 1003 participants, 15 trials) with an ultrasound‐guided injection. This represents a slight difference for pain (0.5 to 1.0 points on a 0 to 10 scale). Mean function (scale range 0 to 100, higher scores indicate better function) was 68 points with injection without image guidance and 2.4 points better (0.2 points worse to 5.1 points better; 895 participants, 14 trials) with an ultrasound‐guided injection. Mean quality of life (scale range 0 to 100, higher scores indicate better quality of life) was 65 with injection without image guidance and 2.8 points better (0.7 worse to 6.4 better; 220 participants, 2 trials) with an ultrasound‐guided injection. In five trials (350 participants), 101/175 (or 606 per 1000) people in the ultrasound‐guided group reported treatment success compared with 68/175 (or 389 per 1000) people in the group injected without image guidance (RR 1.56 (95% CI 0.89 to 2.75)), an absolute difference of 22% more reported success (4% fewer to 62% more).

Low‐certainty evidence (downgraded for bias and imprecision) indicates that ultrasound‐guided injections may not reduce the risk of adverse events compared to injections without image guidance. In five trials (402 participants), 38/200 (or 181 per 1000) people in the ultrasound‐guided group reported adverse events compared with 51/202 (or 252 per 1000) in the non‐image‐guided injection group (RR 0.72 (95% CI 0.4 to 1.28)), an absolute difference of 7% fewer adverse events (15% fewer to 7% more). Five trials reported that there were no serious adverse events. The remaining trials did not report serious adverse events. One trial reported that 1/53 (or 19 per 1000) in the injection without image guidance group and 0/53 in the ultrasound‐guided group withdrew due to adverse events.

Sensitivity analyses indicate that the effects for pain and function may have been influenced by selection bias, and the effects for function may have been influenced by detection bias. The test for subgroup differences indicated there were unlikely to be differences in pain and function across different shoulder conditions.

Authors' conclusions

Our updated review does not support use of image guidance for injections in the shoulder. Moderate‐certainty evidence indicates that ultrasound‐guided injection in the treatment of shoulder pain probably provides little or no benefit over injection without imaging in terms of pain or function and low‐certainty evidence indicates there may be no difference in quality of life. We are uncertain if ultrasound‐guided injection improves participant‐rated treatment success, due to very low‐certainty evidence. Low‐certainty evidence also suggests ultrasound‐guided injection may not reduce the risk of adverse events compared with non‐image‐guided injection. No serious adverse events were reported in any trial.

The lack of significant benefit of image guidance over injection without image guidance to improve patient‐relevant outcomes or reduce harms, suggests that any added cost of image guidance appears unjustified.

Plain language summary

Image‐guided versus blind glucocorticoid injection for shoulder pain

Background

Shoulder pain is most often caused by rotator cuff disease, or adhesive capsulitis (‘frozen shoulder’). The rotator cuff is a group of tendons that holds the shoulder joint in place allowing people to lift their arm. Shoulder pain can be related to wear and tear or inflammation of the shoulder tendons, and pressure on the tendons by the overlying bone when lifting the arm up (impingement). Both conditions cause pain with movement and often pain during the night and sleeping on the affected side; adhesive capsulitis also causes shoulder stiffness.

Glucocorticoids injections can relieve shoulder pain but their effect usually wears off after six to eight weeks. Traditionally, injections are given using anatomic landmarks around the shoulder. Sometimes imaging techniques, such as ultrasound, are used to more accurately guide the injections into the shoulder. It is not known if image‐guided injection relieves shoulder pain more effectively than injections delivered without imaging.

Study characteristics

This Cochrane review is current to 15 February 2021. Nineteen trials (1035 participants) compared ultrasound‐guided injection to ‘blind’ injection. Fourteen trials included participants with rotator cuff disease, four with adhesive capsulitis, and one with mixed shoulder pain. Trials were performed in Korea, Taiwan, Iran, Turkey, Australia, Norway, Spain, Ireland, India and Switzerland. Most participants were female, with a mean age from 31 to 60 years, and mean symptom duration from 2 to 23 months. Six studies reported funding sources.

Key results

Compared to injection into the shoulder without image guidance, ultrasound‐guided injection resulted in little to no benefit at three to six weeks:

Pain (lower scores mean less pain)

Improved by 0.5 points more (0.2 more to 0.8 more) on a 0 to 10‐point scale. Differences of 0.5 to 1.0 points are considered slight or small and are unlikely to be clinically important.

‐ People who had ultrasound‐guided injection rated their pain as 2.6 points

‐ People who had injection without image guidance rated their pain as 3.1 points

Function (higher scores mean better function)

Improved by 2.4 points more (0.2 points worse to 5.1 points more) on a 0 to 100‐point scale. Differences below 10 points are considered slight or small and are unlikely to be clinically important.

‐ People who had ultrasound‐guided injection rated their function as 70.4 points

‐ People who had injection without image guidance rated their function as 68 points

Quality of life (higher scores mean better quality of life)

Improved by 2.8 points (0.7 worse to 6.4 better) on a 0 to 100‐point scale

‐ People who had ultrasound‐guided injection rated their quality of life as 67.8 points

‐ People who had injection without image guidance rated their quality of life as 65 points

Treatment success (defined as pain moderately or a great deal better)

22% more people rated their treatment a success (4% fewer to 62% more), or 22 more people out of 100.

‐ 61 out of 100 people reported treatment success with ultrasound‐guided injection

‐ 39 out of 100 people reported treatment success with injection without image guidance

Adverse events

7% fewer people (15% fewer to 7% more) had adverse events (post‐injection pain, facial redness and warmth) with ultrasound‐guided injection.

‐ 18 out of 100 people reported adverse events with ultrasound‐guided injection

‐ 25 out of 100 people reported adverse events with injection without image guidance

Serious adverse events

Five trials reported that there were no serious adverse events (like infection or nerve injury) with or without use of ultrasound guidance of the injection.

Withdrawals due to adverse events

One trial reported that 1/53 (or 19 out of 1000) people who received the injection without image guidance withdrew from the study due to adverse events, while no one (0/53) in the ultrasound‐guided injection group withdrew due to adverse events.

Quality of evidence

Low to moderate‐certainty evidence shows that in people with shoulder pain, ultrasound‐guided injection does not provide clinically important benefits in pain, function or quality of life compared with non‐image‐guided injection, nor does it reduce the risk of adverse events. These findings were consistent across different shoulder conditions. Further high‐quality research is unlikely to change the conclusions of this review.

Summary of findings

Summary of findings 1. Ultrasound‐guided injection compared to non‐image‐guided injection for shoulder pain.

Ultrasound‐guided injection compared to non‐image‐guided (landmark or intramuscular) injection for shoulder pain
Patient or population: Patients with shoulder pain Settings: International; clinic/hospital Intervention: Ultrasound‐guided injection Comparison: Landmark or intramuscular injection
Outcomes	Illustrative comparative risks* (95% CI)		Relative effect (95% CI)	No of Participants (studies)	Certainty of the evidence (GRADE)	Comments
	Assumed risk	Corresponding risk
	Landmark or intramuscular injection	Ultrasound‐guided injection
Overall Pain numerical rating scale 0 to 10 points (lower score indicates less pain) Follow‐up: > 3 weeks up to 6 weeks	3.1 points1	2.6 points (2.9 to 4.9)		1003 (15 studies)	⊕⊕⊕⊝ moderate2	Image‐guided injection probably results in little to no improvement in pain compared with blind injection. Mean difference in pain 0.5 points better with image‐guided injection (0.2 better to 0.8 points better).
Function various scales 0 to 100 points (higher score indicates better function) Follow‐up: > 3 weeks up to 6 weeks	68 points1	70.4 points (67.8 to 73.1)		895 (14 studies)	⊕⊕⊕⊝ moderate2	Image‐guided injection may have little or no effect on function compared with blind injection. Mean difference in function 2.4 points better with image guided injection (0.2 worse to 5.1 better).
Participant‐assessed success 50% improvement on numerical rating 0 to 10 scale for pain) Follow‐up: end of study (6 weeks up to 12 weeks)	389 per 1000	606 per 1000 (346 to 1000)	RR 1.56 (0.89 to 2.75)	350 (5 studies)	⊕⊝⊝⊝ verylow2,3,4	We are uncertain if image‐guided injection improves treatment success compared with blind injection. Absolute difference: 22% more reported success (4% fewer to 62% more); relative difference: 56% more reported success (11% fewer to 175% more).
Quality of life SF‐36 MCS 0 to 100 points (higher score indicates better quality of life) Follow‐up: > 3 weeks up to 6 weeks	65 points1	67.8 points (64.3 to 71.5)		220 (2 studies)	⊕⊕⊝⊝ low2,4	Image‐guided injection may have little or no effect on quality of life compared with blind injection. Mean difference in quality of life 2.8 points better with image‐guided injection (0.7 worse to 6.4 better).
Number of adverse events Follow‐up: end of study	252 per 1000	181 per 1000 (100 to 323)	RR 0.72 (0.40 to 1.28)	402 (5 studies)	⊕⊕⊝⊝ low2,4	Image‐guided injection may have little or no effect on adverse events compared with blind injection. Absolute difference of 7% fewer adverse events (15% fewer to 7% more); relative difference was 28% fewer events (60% fewer to 28% more).
Serious adverse events Not reported	See comment	See comment	Not estimable	See comment	See comment	Five trials (Ekeberg 2009; Naredo 2004; Saeed 2014; Ucuncu 2009; Zufferey 2012) reported that there were no serious side effects. The remaining trials did not report the incidence of serious adverse events.
Number of withdrawals due to adverse events Follow‐up: end of study	See comment	See comment	Not estimable	See comment	See comment	Only one trial reported withdrawals due to adverse events (Ekeberg 2009): 1/53 in the control group withdrew and received an additional local steroid injection at 2 weeks and 0/53 in the image‐guided group withdrew due to adverse events.
*The basis for the assumed risk (e.g. the median control group risk across studies) is provided in footnotes. The corresponding risk (and its 95% confidence interval) is based on the assumed risk in the comparison group and the relative effect of the intervention (and its 95% CI). CI: Confidence interval; RR: Risk ratio;
GRADE Working Group grades of evidence High quality: Further research is very unlikely to change our confidence in the estimate of effect. Moderate quality: Further research is likely to have an important impact on our confidence in the estimate of effect and may change the estimate. Low quality: Further research is very likely to have an important impact on our confidence in the estimate of effect and is likely to change the estimate. Very low quality: We are very uncertain about the estimate.

¹Mean pain and function in non‐image‐guided group in Ekeberg 2009 (3.1 points on a 0 to 10 scale for pain; 68 points on a 0 to 100 scale for function); mean QoL in non‐image‐guided group in Hsieh 2013.

² Downgraded one level for risk of bias.

³Downgraded one level for inconsistency: I² = 77% for treatment success was difficult to interpret with only five trials but two showed no apparent difference between groups; two favoured image‐guided injection. Although I² was 79% for pain, Naredo 2004 seemed to drive quite a bit of the inconsistency (I² = 64% once it was removed); reporting a large benefit compared to other studies.

⁴Downgraded one level for imprecision due to low event rates, or data from single small study only

Background

This review is one in a series of reviews aiming to determine the benefit and safety of common interventions for shoulder pain. This review is an update of a previous Cochrane review (Bloom 2012).

Description of the condition

Shoulder pain is common, with a point

prevalence of 7% to 26% in the general population (Luime 2004). Shoulder pain accounts for 1.3% of all general practice encounters in Australia, being third only to back and knee pain as musculoskeletal reasons for primary care consultation (Britt 2016). Shoulder pain can also substantially reduce a person's quality of life (MacDermid 2004; Taylor 2005).

While shoulder disorders are labelled and defined in diverse and often conflicting ways (Cools 2017), the most common underlying cause of shoulder pain is rotator cuff disorders, estimated to account for between 65% and 85% of cases depending upon the setting and age of the study population (Ostör 2005; Vecchio 1995). Rotator cuff disease is an umbrella term that encompasses all symptomatic disorders of the rotator cuff, regardless of mechanism (inflammatory, degenerative or acute injury), or precise anatomical location (e.g. supraspinatus tendon versus subacromial bursa (Buchbinder 1996)). It includes diagnostic labels such as rotator cuff tendinopathy or tendinitis, impingement or subacromial syndrome, partial and complete rotator cuff tears, calcific tendinitis, painful arc syndrome and subacromial bursitis. People with symptomatic rotator cuff disease present with shoulder pain, often described as pain in the upper outer arm (Karjalainen 2019a). The pain is aggravated by overhead activities and is often worse at night and lying on the affected side, leading to disrupted sleep. It is accompanied by loss of function and often significant disability. A painful arc (as the arm is passively abducted away from the body, pain occurs between 60° and 120°) is nearly always present.

Adhesive capsulitis (also termed frozen shoulder, painful stiff shoulder or periarthritis) is the next most common cause of shoulder pain. Based on presentations to Dutch general practice, its cumulative incidence is reported to be 2.4 per 1000 people per year (95% confidence interval (CI) 1.9 to 2.9) (Van der Windt 1995). It is characterised by spontaneous onset of pain and progressive restriction of movement of the shoulder. Like rotator cuff disease, disability from the condition also results in restriction of activities of daily living, work and leisure (Codman 1934; Neviaser 1945, Reeves 1975). While specific diagnostic criteria for the condition are lacking, clinical trials of adhesive capsulitis have usually indicated that restricted movement must be present (Green 1998, Schellingerhout 2008). Glenohumeral osteoarthritis accounts for about 2 to 5 percent of shoulder pain in the adult population, its prevalence increases with age and it has a female preponderance (Meislin 2005). It is the most frequent site of osteoarthritis after the knees, hands, hips and ankle and subtalar joints. Osteoarthritis in this joint is most commonly primary but can be secondary to other conditions including trauma, massive rotator cuff tears, inflammatory arthritis and avascular necrosis.

Although there are many accepted forms of conservative therapy for shoulder pain, evidence of their efficacy is not well established. Recent systematic reviews of randomised controlled trials investigating conservative (e.g. manual therapy, exercise, electrotherapy) and operative (e.g. subacromial decompression, rotator cuff tear repair) treatments for shoulder pain often conclude that there is very little evidence to support or refute the efficacy of these commonly used interventions (Page 2014a; Page 2014b; Page 2016a; Page 2016b; Karjalainen 2019a; Karjalainen 2019b).

Description of the intervention

Glucocorticoid injections, which act to reduce inflammation, are commonly used in routine care to treat various forms of shoulder pain including both rotator cuff disease and adhesive capsulitis. Glucocorticoid injections are often recommended as standard treatment in clinical practice guidelines for shoulder pain (American Academy of Orthopedic Surgeons 2011). Traditionally, injection has been performed in the doctor's office at the time of the consultation, using anatomical landmarks to guide needle placement into the subacromial space and/or glenohumeral joint (shoulder joint). With the advent of new imaging modalities such as ultrasound, more people are referred for injections using image guidance.

Glucocorticoid injections provide significant short‐term benefits for patients with shoulder pain. For example, a systematic review of placebo‐controlled trials reported a relative risk for improvement in rotator cuff tendonitis following subacromial glucocorticoid injection of 3.08 (95% CI 1.94 to 4.87) and the number of patients needed to treat for an additional beneficial outcome (based on the pooled relative risk) of 3.3 (95% CI 1.8 to 7.7) (Arroll 2005). Benefits were maintained for up to nine months. The same review also found that subacromial glucocorticoid injections are probably more effective than NSAID medication (Arroll 2005).

Between 30 and 80% of subacromial injections given as a blind injection are reported to reach the subacromial bursa or space (Eustace 1997; Henkus 2006; Partington 1998), although in expert hands, blinded injection may not differ in accuracy to ultrasound‐guided injection (Rutten 2007). For glenohumeral injections, the reported accuracy of blinded injections varies from 27% (Porat 2008) to 99% (Sethi 2005), with one study reporting greater accuracy with an anterior approach (95% accuracy compared with 50% accuracy using a posterior approach) (White 1996).

Despite this, the importance of the accuracy of needle placement with respect to patient outcome remains questionable. Based upon moderate evidence from five trials (290 participants), our previous review (Bloom 2012) was unable to establish that ultrasound‐guided glucocorticoid injections for shoulder pain is superior to either landmark‐guided or intramuscular injections for improving pain, function, and shoulder range of motion. Adverse events were also just as likely with either approach.

How the intervention might work

Glucocorticoid injections are potent anti‐inflammatories and have both systemic and local effects (Ekeberg 2009; Pekarek 2011). The onset, duration, and local effect depends upon the anti‐inflammatory potency of the glucocorticoid, its solubility (depot effect), and the dose given. Based on their potency, an injection may be short, intermediate or long‐acting (Pekarek 2011).

A variety of imaging methods have been used to better localise needle placement for glucocorticoid injection into the subacromial bursa, subacromial space or glenohumeral joint. Potential advantages of image guidance might be increased safety (avoidance of important neurovascular structures), decreased discomfort, and diagnostic value in terms of response to accurate anatomic administration. Ultrasound imaging emits no radiation and can be used to visualise subcutaneous body structures including tendons, muscles and joints. Computerised tomography (CT) scans or magnetic resonance imaging (MRI) scans may also be used to visualise the structures around the shoulder. A series of images, usually X‐rays, can be taken after injection (using a small amount of contrast, i.e. an arthrogram) to ensure that a needle is placed in the correct position within the joint.

In contrast to anatomical landmark‐guided injection, which in many instances can be performed by trained general practitioners or specialists such as rheumatologists and orthopaedic surgeons, most image‐guided procedures are performed by radiologists in a radiology service. While still in the minority, some clinical specialists are now learning these techniques and may have the necessary equipment to perform image‐guided injections in their office.

Why it is important to do this review

It is important to know whether image‐guided injection improves outcomes for people with shoulder pain. Any added benefit in patient outcome achieved by the image‐guided approach will also need to be considered in light of any delay in receiving image‐guided treatment and the added expense of the imaging modality used. In Australia, there has been more than a 3.8‐fold increase in the number of image‐guided injections since 2000 from 15,495 services in 2000‐01 (78 per 100,000 population) to 58,116 services in 2017‐8 (231 per 100,00 population) (Medicare Australia 2018). While the site of injection is not specified, most were likely to be for shoulder pain. This has been accompanied by a substantial increase in health care costs. For example, in the 2014/2015 financial year alone, the total benefits paid through the Medical Benefits Scheme for ultrasound‐guided injection at any site was almost AU$27.5 million (Morrisroe 2018). Due to the lack of clinical justification for the use of ultrasound guidance and its significant added cost, the Australian Rheumatology Association recommends against the use of ultrasound guidance to perform injections into the subacromial space (Morrisroe 2018).

Whether image‐guided glucocorticoid injection is an effective therapeutic tool for improving outcomes for people with shoulder pain is yet to be established. Since our previous review (Bloom 2012), several new trials comparing the clinical effects of ultrasound‐guided glucocorticoid injection to injection without image guidance have been published. It is therefore important to update the evidence on the importance of the accuracy of needle placement with respect to outcome.

Objectives

To update our review and assess the benefits and harms of image‐guided glucocorticoid injection (e.g. injection guided by ultrasound or other imaging modality) into the subacromial bursa or space or glenohumeral joint compared with injection given without image guidance (i.e. relies on anatomical landmarks or systemic intramuscular injection) for people with shoulder pain.

Methods

Criteria for considering studies for this review

Types of studies

We included randomised controlled trials (RCTs) and controlled clinical trials (CCTs) (that used quasi‐randomised methods to allocate participants, for example, by date of birth, hospital record number, or alternation). There were no restrictions on language or date of publication.

Types of participants

Trials that included participants with rotator cuff disease, adhesive capsulitis or glenohumeral osteoarthritis of any age were included. We also included trials with participants of unspecified shoulder pain provided that the inclusion/exclusion criteria were compatible with a diagnosis of either rotator cuff disease or adhesive capsulitis. We excluded trials that included participants with a recent history of significant injury or fracture, and systemic inflammatory conditions such as rheumatoid arthritis or polymyalgia rheumatica.

Types of interventions

Trials comparing image‐guided (ultrasound, arthrogram or MRI) subacromial and/or glenohumeral injection to injection given without imaging, i.e. injection relying on anatomical landmarks or systemic intramuscular injection of glucocorticoids were included. We included systemic intramuscular injection as glucorticoid injection is not accurately delivered into the purported site of pathology. Trials that only considered accuracy of needle placement were excluded.

Trials that provided participants with additional care (e.g. physical therapy, exercise, and medication) were included, provided both treatment arms received the same additional care.

Types of outcome measures

Major outcomes

• Overall pain (mean or mean change) measured by visual analog scales, numerical or categorical rating scales.

• Function (mean or mean change, as reported in the trials), measured by the Shoulder Pain and Disability Index (SPADI); Constant score; Shoulder Disability Questionnaire; Disabilities of the Arm, Shoulder and Hand (DASH), or any other function scale, as outlined below (Data extraction and management).

• Participant‐reported global assessment of treatment success (e.g. proportion of participants with significant overall improvement, or proportion reporting treatment success).

• Quality of life as measured by generic measures (such as components of the SF‐36, EQ‐5D or disease‐specific tools).

• Number of participants experiencing any adverse events.

• Number of participants experiencing serious adverse events.

• Number of participants withdrawing due to adverse events.

Minor outcomes

• Number of participants requiring additional treatments.

• Range of motion (active measures preferred over passive measures: shoulder abduction, flexion, and external rotation). External rotation range of motion is only clinically relevant for adhesive capsulitis so was not extracted from studies including participants with rotator cuff disease.

Time points

We extracted outcomes up to 3 weeks; > 3 weeks to 6 weeks (primary time point); > 6 weeks to 3 months; > 3 months to 6 months; and > 6 months. If a trial reported outcomes at more than one time point within these subgroups (e.g. four weeks and five weeks), we extracted the later time point (five weeks).

Search methods for identification of studies

Electronic searches

We searched the following electronic databases, unrestricted by date or language up to 15 February 2021:

• Cochrane Central Register of Controlled Trials (CENTRAL, via Ovid EBM Reviews) (Appendix 1);

• MEDLINE (Ovid) (Appendix 2);

• Embase (Ovid) (Appendix 3).

We searched the following trial registries to identify ongoing trials: ClinicalTrials.gov up to 15 February 2021 and the World Health Organisation International Clinical Trials Registry Platform (http://www.who.int/trialsearch/Default.aspx) up to 06 July 2020 (we could not access the portal after this date as the WHO ICTRP was not available) (Appendix 4).

Although this is an update of the previously published version, we ran the searches from database inception to 15 February 2021 as several MeSH terms were updated in the original search strategy which was run in June 2011.

Searching other resources

We also screened reference lists of retrieved review articles and trials to identify potentially relevant studies.

Data collection and analysis

Selection of studies

Two review authors (JZ, AR) independently selected the trials to be included based on title and abstract screening. Articles selected by at least one of the review authors were retrieved in full text for closer examination. The review authors were not blinded to the journal or authors. Disagreement about inclusion or exclusion of individual studies was resolved by a third review author (RJ or RB).

Data extraction and management

The same two review authors (JZ, AR) independently extracted the following data from the included trials and entered the data in RevMan 5:

1) trial characteristics including size and location of the trial, notable possible author conflicts of interests, and source of funding;

2) characteristics of the study population including age, and characteristics of shoulder pain including diagnosis criteria, and disease duration, baseline measures in treatment and control group for pain, function, quality of life;

3) characteristics of the therapy in all trial arms including type and dose of glucocorticoid therapy, site of injection and the method of anatomic or image‐guided location of the needle;

4) risk of bias domains (as outlined in Assessment of risk of bias in included studies, below);

5) outcome measures ‐ mean and standard deviation (SD) for continuous outcomes, and number of events for dichotomous outcomes (as outlined in Types of outcome measures).

If additional data were required, we contacted the trial authors to obtain this. Where data were imputed or calculated (e.g. standard deviations calculated from standard errors, P values, or confidence intervals, or imputed from graphs, or from standard deviations in other trials) we reported this in the Characteristics of included studies table. Any disagreements were resolved by consensus and/or arbitration by a third review author (RJ).

There may be multiple outcome results available in the trial reports (e.g. from multiple scales, time points and analyses). To prevent selective inclusion of data based on the results (Page 2013), we used the following decision rules to select data from trials:

Where trials did not include a measure of overall pain but included one or more other measures of pain, for the purpose of pooling data for the primary analysis of pain, we combined overall pain with other types of pain in the following hierarchy: unspecified pain, pain at rest or night, pain with activity or daytime pain.

Where trialists reported pain on more than one scale, we extracted data on the scale that was highest on the following list:

• visual analog scale;

• numerical rating scale;

• categorical rating scale;

• pain subscale of an overall function scale such as the Shoulder Pain and Disability Index (SPADI), Constant score, or other scale;

• any other measure of pain.

Where trialists reported outcome data for more than one function scale, we extracted data on the scale that was highest on the following list, based on the most commonly used scores used in trials of interventions for shoulder pain (Page 2015):

• Constant‐Murley Score (Constant 1987);

• Shoulder Pain and Disability Index (SPADI) (Roach 1991);

• Oxford Shoulder Score (OSS);

• American Shoulder and Elbow Surgeons Standardized Form (ASES‐SF);

• University of California Los Angeles (UCLA) Shoulder Score;

• Disabilities of the Arm, Shoulder and Hand (DASH);

• Shoulder Disability Questionnaire (SDQ);

• Croft Shoulder Disability Questionnaire (Croft 1994); or

• any other shoulder‐specific function scale.

For continuous outcomes, if trialists reported end of treatment mean scores that were adjusted for baseline scores (e.g. analysis of covariance adjusted for baseline score) along with either final values and change from baseline values for the same continuous outcome, we extracted adjusted mean score preferentially, over final values, and over change scores.

Where trialists reported data analysed based on the intention‐to‐treat (ITT) sample and another sample (e.g. per‐protocol, as‐treated), we extracted ITT‐analysed data.

If we had identified, cross‐over RCTs, we would have extracted data from the first period only.

Assessment of risk of bias in included studies

Two review authors (JZ, AR) independently assessed the risk of bias of each included trial and resolved any disagreements by consensus, or consultation with a third review author (RJ), where necessary.

We assessed the following methodological domains, as recommended by Cochrane (Higgins 2017):

• sequence generation;

• allocation sequence concealment;

• blinding of participants, personnel;

• blinding of outcome assessors: we considered blinding of assessors of self‐reported subjective outcomes (pain, function, success, quality of life) separately from assessors of more objective outcomes (such as range of motion, number who had additional surgery);

• incomplete outcome data;

• selective outcome reporting; and

• other potential threats to validity, such as unit of analysis issues, inappropriate or unequal application of co‐interventions across treatment groups.

We graded each potential source of bias as high, low or unclear risk, and provided a quote from the study report together with a justification for our judgment in the risk of bias table. We summarised the risk of bias judgements across different studies for each of the domains listed. We considered blinding separately for different key outcomes, where necessary. Where information on risk of bias related to unpublished data or correspondence with a trialist, we noted this in the risk of bias table.
When considering treatment effects, we took into account the risk of bias for the studies that contributed to that outcome. We presented the figures generated by the risk of bias tool to provide summary assessments of the risk of bias.

Measures of treatment effect

When possible, the analyses were based on intention‐to‐treat data (outcomes provided for every randomised participant) from the individual trials. For each trial, we presented outcome data as point estimates with mean and standard deviation for continuous outcomes and risk ratios (RRs) with corresponding 95% confidence intervals for dichotomous outcomes. Where possible, for continuous outcomes, we extracted end of treatment scores, rather than change from baseline scores.

Where trials used different measures for the same outcome or concept, we used the most common outcome measure as an index outcome measure (Guyatt 2013). We transformed means and standard deviations (SDs) of other outcome measures to the scale of the index instrument and pooled the data using MD as the summary estimate, according to the methods of Thorlund 2011. For pain, we assumed the zero to 10 numerical rating scale (zero represented no pain) as the index instrument, assumed the 10 cm visual analog scale (VAS) was comparable, and transformed zero to 9 (Ekeberg 2009) to a zero to 10 scale. The trials reported various functional measures (including the Constant score, Shoulder Pain And Disability Index, Shoulder Function Assessment, and others). We assumed these were comparable, and that the zero to 100 Constant score (where 100 is best function) was the index measure of overall shoulder function. We transformed the zero to 70 Shoulder Function Assessment (e.g. Naredo 2004) to a zero to 100 scale, and reversed the direction of the Shoulder Pain And Disability Index shoulder disability score (e.g. used in Ekeberg 2009 and Hsieh 2013) and the Shoulder Disability Questionnaire (e.g. used in Dogu 2012).

For dichotomous outcomes, we calculated the absolute difference from the difference in the risks between the intervention and control group, as calculated in GRADEpro GDT (GRADEpro GDT 2015), and expressed as a percentage. We calculated the relative percent change as the RR minus 1 and expressed this as a percentage.

Rather than using minimal clinically important differences (MCIDs) which are not well defined for shoulder pain, we judged the magnitude of the effect based upon between‐mean group differences for continuous measures based upon a similar approach used in the American College of Physicians 2017 guidelines for low back pain (Chou 2017). For pain measured on a 0 to 10‐point scale, a difference of 0.5 to 1.0 points was considered slight to small, a difference of > 1 to 2 points was considered moderate and a difference of > 2 points was considered large. Similarly, for function and health‐related quality of life measured on a 0 to 100‐point scale, a difference of 5 to 10 points was considered slight/small, a difference of > 10 to 20 points was considered moderate and a difference of > 20 points was considered large.

Unit of analysis issues

The unit of analysis was the participant. For trials that injected both shoulders, we planned to extract data once for each participant, if the trials had adjusted for the unit of analysis, and presented data that way. However, since two trials (randomised by participant) injected both shoulders (Cole 2016; Saeed 2014) in a few participants, and reported the results for the number of shoulders, we had to extract data for the number of shoulders.

Where multiple trial arms were reported in a single trial, we included only the relevant arms for our comparison but have reported that there were multiple trial arms in the Characteristics of included studies table.

Dealing with missing data

We contacted trial authors to obtain data that were missing from the trial reports. For dichotomous outcomes, we used the number randomised as the denominator, making the assumption that any participants missing at the end of treatment did not have a positive outcome. For continuous outcomes with no standard deviation reported, we calculated standard deviations if possible from standard errors, P values, or confidence intervals.

If no measures of variance were reported and standard deviation could not be calculated, we planned to impute standard deviations from other studies in the same meta‐analysis, using the average of the other standard deviations available provided only a small proportion of studies comprising the meta‐analysis had missing data.

Assessment of heterogeneity

We assessed included trials for clinical similarity in terms of participants and interventions and comparators. For studies judged as clinically similar, we quantified the possible magnitude of inconsistency (i.e. heterogeneity) across studies, using the I² statistic with a rough guide for interpretation as follows: 0% to 40% might not be important; 30% to 60% may represent moderate heterogeneity; 50% to 90% may represent substantial heterogeneity; 75% to 100% may represent considerable heterogeneity (Deeks 2008).

If we had identified cases of considerable heterogeneity (defined as I² ≥ 75%), we would explore the data further by comparing the characteristics of individual studies and performing subgroup analyses.

Assessment of reporting biases

In order to determine whether outcome reporting bias was present, we checked a priori trial protocols against published reports of trial results (checked if all planned outcomes had results reported).

We compared the fixed‐effect estimate against the random‐effects model to assess the possible presence of small sample bias in the published literature (i.e. in which the intervention effect is more beneficial in smaller studies). In the presence of small sample bias, the random‐effects estimate of the intervention is more beneficial than the fixed‐effect estimate (Sterne 2017). If we were able to pool more than 10 trials, we performed formal statistical tests to investigate funnel plot asymmetry to detect the possibility of publication bias (Sterne 2017).

Data synthesis

For clinically similar studies, we pooled outcomes in a meta analysis using the random‐effects model as a default, based on the assumption that clinical diversity is likely to exist, and that different studies are estimating different intervention effects.

Subgroup analysis and investigation of heterogeneity

To explain the heterogeneity between the results of the included studies, we planned to assess if different underlying shoulder disorders impacted on pain and function. To do this, we presented outcomes separately by diagnosis (rotator cuff disease versus adhesive capsulitis, versus mixed or undefined shoulder pain). We used the formal test for subgroup interactions in Review Manager (RevMan 2014) and exercised caution in the interpretation of subgroup analyses, as advised in section 9.6 of the Cochrane Handbook (Deeks 2017).

Sensitivity analysis

We performed a sensitivity analysis to investigate the robustness of the treatment effect (on pain and function) to the presence of selection and detection biases. To do this, we removed trials that reported inadequate or unclear allocation concealment (at risk of selection bias) and trials that lacked participant blinding (at risk of detection bias) from meta‐analysis to see if this changed the overall treatment effect.

Post hoc, we assessed the robustness of the pain outcome at the primary time point to the inclusion of the two trials with a potential unit of analysis issue. These trials included a small proportion of participants with bilateral shoulder injections that were not adjusted for in their trial analyses.

Summary of findings and assessment of the certainty of the evidence

We presented the major outcomes (pain, function, global assessment of success, quality of life, withdrawals, adverse events, serious adverse events) in a summary of findings (SoF) table which summarised the quality of evidence, the magnitude of effect of the interventions examined, and the sum of available data on the outcomes, as recommended by Cochrane (Schünemann 2017a). The summary of findings table includes an overall grading of the evidence related to each of the major outcomes, using the GRADE approach (Schünemann 2017b).

All authors (JZ, AR, RJ, RB) independently assessed the quality of the evidence. We used the five GRADE considerations (study limitations, consistency of effect, imprecision, indirectness and publication bias) to assess the quality of a body of evidence as it related to the studies which contributed data to the meta‐analyses for each of the major outcomes. We used methods and recommendations described in sections 8.5, 8.7 and chapters 11 and 13 (section 13.5) of the Cochrane Handbook for Systematic Reviews of Interventions (Higgins 2017; Schünemann 2017a ) using GRADEpro software (GRADEpro GDT 2015). We justified all decisions to downgrade the quality of studies using footnotes and made comments to aid the reader's understanding of the review, where necessary.

Results

Description of studies

Results of the search

The search identified 2152 records from the following databases: MEDLINE (633) Embase (592) CENTRAL (757) WHO ICTRP website (49) and Clinicaltrials.gov (121) (see flow chart in Figure 1). There were 1608 records after removal of duplicates, of which 24 were assessed in full text and fourteen new trials were included (Akbari 2020; Azadvari 2020; Bhayana 2018; Cho 2010; Cho 2021; Cole 2016; Dogu 2012; Haghighat 2015; Hsieh 2013; Lee 2015; Raeissadat 2017; Roddy 2020; Saeed 2014; Zufferey 2012) in addition to the five included trials (n = 290 participants) from the previous version of the review (Chen 2006; Ekeberg 2009; Lee 2009; Naredo 2004; Ucuncu 2009). In total, we included 19 trials in this review update.

1.

Study flow diagram.

Four trials are awaiting assessment (Cinar 2018; IRCT2017021524621N6; Moore 2018; Pierce 2018) (see Table of Characteristics of studies awaiting classification). No ongoing trials were identified.

Included studies

We have provided a complete description of the 19 included trials (n = 1035 participants) in the Characteristics of included studies table. Since our previous review (Bloom 2012), 14 additional trials (n = 873 additional participants) were included.

Trial design, setting and characteristics

Three of the five trials included in our previous review were RCTs (Ekeberg 2009; Naredo 2004; Ucuncu 2009). Of the two categorised as CCTs, one (reported to be an RCT) alternatively assigned participants to treatment (Lee 2009), while the other did not specify how participants were allocated to treatment group (Chen 2006). All 14 of the new trials were stated to be RCTs although five failed to report their method of randomisation (Bhayana 2018; Haghighat 2015; Naredo 2004; Roddy 2020; Ucuncu 2009).

Trials were conducted in 11 countries: four trials were conducted in Korea (Cho 2010; Cho 2021; Lee 2009; Lee 2015), two in Taiwan (Chen 2006; Hsieh 2013), three in Iran (Akbari 2020; Haghighat 2015; Raeissadat 2017), three in Turkey (Akbari 2020; Dogu 2012; Ucuncu 2009), and one each in Australia (Cole 2016), Norway (Ekeberg 2009), Spain (Naredo 2004), Ireland (Saeed 2014), India (Bhayana 2018), Switzerland (Zufferey 2012), and the United Kingdom (Roddy 2020).

Twelve trials were conducted in outpatient clinics (Akbari 2020; Azadvari 2020; Bhayana 2018; Cho 2021; Ekeberg 2009; Haghighat 2015; Hsieh 2013; Lee 2009; Roddy 2020; Saeed 2014; Ucuncu 2009; Zufferey 2012), two in a hospital (no further details specified) (Lee 2015; Raeissadat 2017), and one in an orthopaedic research institute (Cole 2016). Four trials did not specify the setting of the trial (Chen 2006; Cho 2010; Dogu 2012; Naredo 2004).

Funding source was reported in six trials (Akbari 2020; Azadvari 2020; Bhayana 2018; Cho 2021; Ekeberg 2009; Hsieh 2013). Akbari 2020 reported funding from the Baskent University Research Fund. Bhayana 2018 reported there was no funding for the study. Ekeberg 2009 declared funding from the University of Oslo. Hsieh 2013 declared funding from Shin Kong Wu Ho‐Su Memorial Hospital. Roddy 2020 reported funding from Arthritis Research UK Primary Care Centre and the National Institute for Health Research.

Trial participants

The mean age of participants in 18 trials varied between 31 (Azadvari 2020) and 60 (Cho 2010) years; the mean age of all participants was 53 years. The percentage of female participants in 17 trials varied between 33% (Chen 2006) and 73% (Ucuncu 2009); the mean percentage of females was 55%. Azadvari 2020 did not report the gender distribution of the sample. Saeed 2014 did not report the mean age or gender distribution of the sample.

Symptom duration was reported in 15 trials. Twelve trials reported mean symptom duration (Azadvari 2020; Cho 2021; Cole 2016; Dogu 2012; Haghighat 2015; Hsieh 2013; Lee 2009; Naredo 2004; Raeissadat 2017; Saeed 2014; Ucuncu 2009; Zufferey 2012); this varied between 1.8 months (Haghighat 2015) and 23 months (Azadvari 2020). Chen 2006 reported that symptom duration ranged from two to 10 months. Ekeberg 2009 and Roddy 2020 reported symptom duration as categories. For Ekeberg 2009, symptom duration was > 24 months for 26% in the ultrasound‐guided group and 23% in the ‘systemic’ group. For Roddy 2020, symptom duration was > 12 months for 41% in the ultrasound‐guided group and 40% in the ‘blind’ group.

Nine studies relied on clinical features alone for confirming eligibility into the trial (Azadvari 2020; Bhayana 2018; Cho 2021; Ekeberg 2009; Haghighat 2015; Hsieh 2013; Naredo 2004; Raeissadat 2017; Roddy 2020), while two relied on ultrasound (Chen 2006; Cho 2010), three relied on ultrasound and radiography (Cole 2016; Lee 2009; Lee 2015), and two relied on MRI (Akbari 2020; Dogu 2012). Three trials did not report how a diagnosis was made (Saeed 2014; Ucuncu 2009; Zufferey 2012).

Fourteen trials included participants that could be categorised as having rotator cuff disease although the diagnostic labels varied (Akbari 2020; Azadvari 2020; Bhayana 2018; Chen 2006; Cho 2010; Cole 2016; Dogu 2012; Ekeberg 2009; Haghighat 2015; Hsieh 2013; Naredo 2004; Roddy 2020; Saeed 2014; Ucuncu 2009). Bhayana 2018 included 60 participants with rotator cuff syndrome. Three trials included participants with 'subacromial bursitis' (Chen 2006: n = 40; Cho 2010: n = 28 with either 'subacromial' or 'subdeltoid bursitis'; and Hsieh 2013: n = 96 with 'chronic subacromial bursitis'). Seven trials included participants with 'subacromial impingement syndrome' (Akbari 2020: n = 29; Azadvari 2020: n = 30; Cole 2016: n = 51 (56 shoulders); Dogu 2012: n = 46; Haghighat 2015: n = 40; Roddy 2020: n = 128; Saeed 2014: n = 100 (125 shoulders)). Ekeberg 2009 included 106 participants with rotator cuff disease; Naredo 2004 included 41 participants with 'periarticular' disorders; and Ucuncu 2009 included 60 participants with 'soft‐tissue lesions' of the shoulder.

Four trials included participants with adhesive capsulitis (Cho 2021; Lee 2009; Lee 2015; Raeissadat 2017). Cho 2021 included 90 participants, Lee 2009 included 43 participants, Lee 2015 included 77 participants and Raeissadat 2017 included 41 participants. One trial (Zufferey 2012) included 70 participants with acute shoulder pain and did not provide a more specific diagnostic label. No trials involving people with glenohumeral osteoarthritis were included in this review but one trial is awaiting assessment as it is currently only available as a pre‐print in bioRxiv and has not yet been published in a peer‐reviewed journal (Moore 2018).

Interventions

Details of the interventions in each trial are presented in the Characteristics of included studies table and Table 2.

1. Characteristics of interventions in included trials.

Study ID	Study population	Ultrasound‐guided injection	Type of ultrasound	Landmark‐guided injection	Co‐interventions	Adverse events
Akbari 2020 (Turkey)	Subacromial impingement syndrome confirmed by magnetic resonance imaging (MRI), symptoms for more than three months, increased pain on shoulder abduction, painful restriction of active flexion and/or abduction of the shoulder with more restriction on passive range of motion, and a positive Hawkins‐Kennedy impingement sign	Summary Injection of 1 mL of 40 mg methylprednisolone and 4 mL procaine 2% into the subacromial bursa performed by a radiologist with 10 years of experience in musculoskeletal radiology Procedure The patient was seated and the shoulder was internally rotated with the ipsilateral hand positioned on the hip in the modified Crass position. Methylprednisolone acetate 40 mg in 1 mL and procaine 2% 4 mL were prepared in a 5 mL syringe. The anterolateral aspect of the shoulder was cleaned using 10% povidone iodine solution. The US probe was placed on the anterolateral aspect of the shoulder and the subacromial bursa was visualised. A 21‐gauge needle was used to enter the anteromedial aspect of the shoulder under continuous US guidance. Once the bevel of the needle was visualised in the subacromial bursa, the solution was injected.	2014 model Siemens Acuson S2000 (Siemens Healthcare, Erlangen, Germany) and a 9 MHz linear probe	Summary Injection of 1 mL of 40 mg methylprednisolone and 4 mL procaine 2% into the subacromial space performed by a physiatrist with more than 10 years of experience in the field Procedure The injection was performed using a standard posterolateral approach and an aseptic technique. The patient was seated upright with the arms resting comfortably at the side. The distal, lateral, and posterior edges of the acromion were palpated and the needle was inserted just inferiorly to the posterolateral edge of the acromion and directed towards the opposite nipple. Aspiration was performed to ensure that the needle was not in a blood vessel prior to administration of the drug. Methylprednisolone acetate 40 mg in 1 mL and procaine 2% 4 mL were injected slowly using a 21‐gauge needle 1 cm into the subacromial space.	Not reported	None in either group
Azadvari 2020 (Iran)	Subacromial impingement in paraplegic patients with spinal cord injury (below T6). Shoulder pain visual analog (VAS) score higher than 4 and positive Neer test and Hawkins‐Kennedy test	Summary Injection of 1 mL methylprednisolone and 2 mL lidocaine into the subacromial‐subdeltoid bursa performed by a physiatrist. Procedure The injection was performed with lateral approach. The patients were at sitting position, they put their hand on iliac crest by internal rotation of shoulder joint. The probe was placed on supraspinatus muscle in longitudinal plane. Acromion was found by ultrasound guidance. Then the probe was derived to the inferior part so the subacromial‐subdeltoid bursa was determined. A hypodermic needle (gauge‐22) was entered beneath the probe in in‐plane manner in upward medial direction until the needle tip entered the bursa and its dilatation was observed during injection. Injection in subacromial space was performed by a physiatrist.	Not reported	Summary Injection of 1 mL methylprednisolone and 2 mL lidocaine into the subacromial space performed by a physiatrist Procedure Participants received lateral approached blind glucocorticoid injection guided by anatomic landmarks under sterile condition. The injection site was marked and then sterilised. Lateral edge of the acromion was touched by hand and the needle with gauge of 22–23 was guided 2–3 cm beneath the lateral edge of the acromion toward the medial side (slightly superior). Under resistance‐free condition, 1 cc depo‐medrol (Methylprednisolone, Merck, Germany) was injected along with 2 cc lidocaine.	Exercise therapy was used in both groups after seven days. The exercises included range of motion, posterior capsule stretching and isometric strengthening exercises presented as an instruction.	Not reported
Bhayana 2018 (India)	Rotator cuff syndrome with shoulder pain and < 50% reduction in one direction of range of motion	Summary Injection of 2 mL (40 mg/mL) methylprednisolone and 2 mL of 1% lignocaine into subacromial bursa performed by one experienced radiologist Procedure The patient was seated with the affected arm in hyperextension and internal rotation with the elbow bent and back of the hand resting against the lower back. A 10 mL syringe connected to a 5 cm, 21‐gauge needle was prepared with 2 mL of 40 mg/mL methylprednisolone acetate suspension mixed and 2 mL of 1% lignocaine and was inserted parallel to the transducer in a semi‐oblique plane from the anterior side of the shoulder. The bevelled side facing the transducer, the needle was advanced under the real time ultrasound assistance till the tip of the needle appeared in the subacromial bursa. Rotation of the bevelled side of the needle by 180 degrees was done for confirmation of the tip positioning and the bulging of subacromial bursa ascertained in the real time imaging.	Phillips HD 7 US machine with 7–10 MHz multi frequency broad band transducer	Summary Injection of 2 mL (40 mg/mL) methylprednisolone and 2 mL of 1% lignocaine into subacromial space by one orthopaedic surgeon Procedure A 10 mL syringe connected to 5 cm, 21‐gauge needle was prepared with 2 mL of 40 mg/mL methylprednisolone acetate suspension mixed, 2 mL lignocaine and 2 mL of radio opaque non‐ionic contrast media iohexol. Access to the subacromial space was achieved using a lateral approach. The needle was inserted just inferior to the midlateral aspect of the acromion, with the needle angled slightly cephalad, passing through the deltoid muscle, and directed medially and slightly anterior to the subacromial bursa. Care was taken to avoid injection directly into the tendons of the rotator cuff. Correct placement of the drug in the subacromial subdeltoid space was confirmed by the fluoroscopic assessment under the C arm within 10 minutes of the injection. Accuracy of correct instillation of methylprednisolone‐ lignocaine‐iohexol combination was assessed by a senior radiologist.	Both groups received 5‐day course of antibiotics for prevention of injection‐induced subacromial bursitis, and home‐based shoulder rehabilitation programme consisting of shoulder abduction and pendulum exercises.	None in either group
Chen 2006 (Taiwan)	Ultrasound‐confirmed subacromial bursitis with shoulder pain and a painful arc	Summary Injection of 1 mL betamethasone and 1 mL 1% lignocaine into subacromial bursa by one physician experienced using ultrasound probes Procedure The patient was seated upright, with the back well supported and arms behind the back with elbows bent. The needle was inserted into the subacromial bursa under ultrasound guidance. Aspiration of the effusion was done first before injecting the steroid‐lidocaine suspension into the subacromial bursa. Advancement of the needle to the lesion site was observed as continuous and real‐time images.	LOGIQ 9 machine with 10L probe (4‐10 MHz) (General Electronic Company, Milwaukee, WI)	Summary Injection of 1 mL betamethasone and 1 mL 1% lignocaine into subacromial bursa by one physician experienced in peripheral joint and soft‐tissue injections. Procedure The patient was seated upright, with the back well supported and arms behind the back with elbows bent. The acromion was palpated by the thumb and the needle inserted in a horizontal approach. The needle was adjusted in different depths and angles to try to aspirate the effusion. Exaggerated needle‐adjusting motions were prevented to avoid pain and discomfort that may be experienced by the patients. If no effusion could be aspirated, the physician injected the suspension into the subacromial bursa.	Not reported	Not reported
Cho 2010 (Korea)	Ultrasound‐confirmed subacromial or subdeltoid bursitis with shoulder pain and positive findings on Hawkin's test (i.e. reproduction of pain)	Summary Injection of 5 mL of 0.5% lidocaine and 5 mL triamcinolone acetate into subacromial or subdeltoid bursa by a single researcher Procedure After the patient was seated in a chair, the deformed shoulder was extended, the elbow joint was flexed, and the modified crass position was adopted. The transducer was placed in front of the acromion, the needle was inserted in parallel with the probe to position the tip to the bursa and real‐time ultrasound was used to confirm that the tip of the needle was in the bursa and solution was then injected into the bursa.	Philips Newand 13 MHz linear transducer	Summary Injection of 5 mL of 0.5% lidocaine and 5 mL triamcinolone acetate into subacromial or subdeltoid bursa by the same researcher Procedure A posterior approach was used.	Not reported	None in either group
Cho 2021 (Korea)	Primary frozen shoulder with pain and limitation of passive motion of greater than 30 degrees in two or more planes of movement	Summary Injection of 40 mg of triamcinolone acetonide, 4 mL of 1% lidocaine, 4 mL of normal saline, and 3 mL of water soluble unionised contrast into the glenohumeral joint by a single specialist with 15 years of experience in the field Procedure A posterior approach was used for the injection in both groups, with the patient in the semi‐lateral decubitus position on the unaffected side and 45° anterior tilting of affected side. For the US group, the needle was advanced laterally to medially with visualisation of its shaft using a linear 5‐ to 12‐MHz probe (HD15 ultrasound system; Philips, Bothell, Washington, USA), and reaching the glenohumeral joint space between the posterior aspect of the humeral head and the glenoid labrum.	Linear 5‐ to 12‐MHz probe (HD15 ultrasound system; Philips, Bothell, Washington, USA)	Summary Injection of 40 mg of triamcinolone acetonide, 4 mL of 1% lidocaine, 4 mL of normal saline, and 3 mL of water soluble unionised contrast into the glenohumeral joint by a single specialist with 15 years of experience in the field Procedure A posterior approach was used for the injection in both groups, with the patient in the semi‐lateral decubitus position on the unaffected side and 45° anterior tilting of affected side. For the blind group, the needle was inserted 2 cm inferior to the posterolateral margin of the acromion and directed anteriorly towards the coracoid process. After the needle contacted the humeral head, it was slightly withdrawn and the injection was administered. For the blind group, the US probe was placed just under the acromion without visualisation, to blind the patient to the allocation group.	All patients were instructed to use a home‐based exercise programme to increase ROM and were allowed to perform an exercise three times a day (15 minutes each round). The programme included pendulum exercises, wall‐climbing exercises, and gentle ROM exercises with a bar. During this programme, patients were asked to refrain from provoking post‐mobilisation soreness with self‐feedback. Patients were forbidden from having acupuncture or receiving additional injections from other hospitals.	None in either group
Cole 2016 (Australia)	Subacromial impingement syndrome with shoulder pain during overhead activities and clinical signs of impingement (either in internal rotation or external rotation)	Summary Injection of 1 mL of 40 mg/mL methylprednisolone acetate and 5 mL of 1% lidocaine hydrochloride into subacromial bursa by one surgeon with over 10 years of experience performing subacromial injections Procedure With the patient in an upright sitting position, a 22‐gauge needle was placed directly underneath the midpoint of the lateral acromion. The needle was directed anteriorly and cephaladly toward the subacromial bursa (as guided by the ultrasound image) until the tip of the needle was seen in the bursa.	General Electric Logiq E9 machine with a 6‐ to 15‐MHz linear transducer and 50 x 10–mm footprint	Summary Injection of 1 mL of 40 mg/mL methylprednisolone acetate and 5 mL of 1% lidocaine hydrochloride into subacromial bursa by the same surgeon that performed the ultrasound‐guided injections Procedure Blind injections were performed with the patient in the same upright sitting position with the ultrasound probe on the acromion to keep the patients blinded to the treatment group. An image could be seen on the screen by the patient, but it was not possible for this to be used by the surgeon to guide the injection, as the way that he was positioned behind the patient meant that he did not have direct vision of the screen. A posterior approach was used, and the needle was inserted 1 cm medially and inferiorly to the posterolateral corner of the acromion and directed cephaladly, anteriorly, and medially toward the subacromial bursa.	Participants in both groups were not restricted from the use of any analgesic or anti‐inflammatory medication or other treatments such as physical therapy.	None in either group
Dogu 2012 (Turkey)	Subacromial impingement syndrome with shoulder pain and at least two positive results in the provocative tests (Neer, Hawkins, and Jobe’s ‘‘Empty Can’’). Diagnosis of subacromial impingement syndrome was confirmed by MRI.	Summary Injection of 1 mL of 5 mg/mL betamethasone dipropionate, 9 mL of 10 mg/mL prilocaine hydrochloride and 0.02 mL of 0.01 mmol gadolinium diethylenetriaminepentaacetic acid directly underneath the mid‐lateral aspect of the acromion, directed anteriorly and cephaladly by a musculoskeletal radiologist Procedure The patient was seated in the upright position. Sterile gel was applied to the probe. The probe was held in one hand, and the syringe with the mixture was held in the other hand. With the patient in an upright sitting position, a 21‐gauge needle was placed under the probe, and a lateral injection was made directly underneath the mid‐lateral aspect of the acromion, directed anteriorly and cephaladly.	Aplio XV machine (Toshiba, Tokyo, Japan) with a 7.5‐ to 14‐MHz linear probe	Summary Injection of 1 mL of 5 mg/mL betamethasone dipropionate, 9 mL of 10 mg/mL prilocaine hydrochloride and 0.02 mL of 0.01 mmol gadolinium diethylenetriaminepentaacetic acid into subacromial space by a physiatrist Procedure Patient seated in upright position. A posterior approach was used. Under the guidance of a Toshiba Power Vision ultrasonograph, gel was applied to a 7.5‐ to 14‐MHz linear probe, and it was placed on the trapezius muscle. A 21‐gauge needle was located 1 cm posteriorly and inferiorly to the border of the acromion and directed cephaladly, anteriorly, and medially toward the subacromial space.	Participants in both groups could use only 2000 mg paracetamol daily; no participant received physical therapy.	None in either group
Ekeberg 2009 (Norway)	Rotator cuff disease with shoulder pain, < 50% reduced range of movement in no more than one direction of external rotation, internal rotation or abduction, and a positive Hawkins‐Kennedy impingement sign	Summary Injection of 2 mL (20 mg) triamcinolone and 5 mL (10mg/mL) lidocaine to subacromial bursa and 4 mL (10mg/mL) lidocaine to upper gluteal region by one physician Procedure To improve blinding of participants by inducing a temporary pain relief (impingement test) and mask possible post‐injection pain, participants received an injection of local anaesthetic in the shoulder and the gluteal region. The participant was seated with the arm rotated internally behind the back, elbow bent, and the back of the hand resting against the lower back. The physician used the ultrasound probe to visualise the insertion of the supraspinatus tendon and the subacromial bursa on the longitudinal axis, taking care that the contents of the syringes were never shown to the participants. The physician used the anterior approach with a 0.8 × 50 mm intra‐muscular needle for the subacromial injections, perforating the skin and tracking the needle in real time until it reached the subacromial bursa. Participants received an injection of 2 mL (10 mg/mL) triamcinolone (Kenacort‐T, Bristol‐Myers Squibb) and 5 mL (10 mg/mL) lidocaine hydrochloride (Xylocaine, AstraZeneca) to the subacromial bursa and an intramuscular injection of 4 mL (10 mg/mL) lidocaine hydrochloride to the upper gluteal region.	Commercial ultrasound equipment (Medison 128 BWprime, Medison Co, Seoul, Korea) with a 5‐9 MHz linear transducer	Summary Injection of 5 mL (10mg/mL) lidocaine to subacromial bursa and 2 mL (20 mg) triamcinolone and 2 mL (10 mg/mL) lidocaine to upper gluteal region by same physician Procedure To improve blinding of participants by inducing a temporary pain relief (impingement test) and mask possible post‐injection pain, participants received an injection of local anaesthetic in the shoulder and the gluteal region. The participant was seated with the arm rotated internally behind the back, elbow bent, and the back of the hand resting against the lower back. The physician used the ultrasound probe to visualise the insertion of the supraspinatus tendon and the subacromial bursa on the longitudinal axis, taking care that the contents of the syringes were never shown to the participants. The physician used the anterior approach with a 0.8 × 50 mm intra‐muscular needle for the subacromial injections, perforating the skin and tracking the needle in real time until it reached the subacromial bursa. Participants received an injection of 5 mL (10 mg/mL) lidocaine hydrochloride to the subacromial bursa and an intramuscular injection of 2 mL (10 mg/mL) triamcinolone and 2 mL (10 mg/mL) lidocaine hydrochloride to the upper gluteal region.	Participants in both groups were allowed to use analgesics, and to continue any physiotherapy programme that they were attending at baseline.	There were no serious adverse events in either group. 1/53 participants in the ‘local’ group and 4/53 in the ‘systemic’ group reported post‐injection pain in the shoulder. Nine participants from both groups reported mild adverse effects such as facial redness, dizziness, and feeling of warmth, however, per group data were not given.
Haghighat 2015 (Iran)	Subacromial impingement syndrome with shoulder pain, pain in abduction (or painful restriction of glenohumeral mobility) and positive Neer and Hawkins tests	Summary Injection of 40 mg methylprednisolone with 1 cc lidocaine 2% into subacromial bursa; personnel not specified Procedure A lateral approach using the ultrasound equipment with high frequency linear transducer for guidance was adopted. The ultrasound probe was positioned parallel to the long axis of the supraspinatus. The skin was punctured at a distance of about 2–3 cm from the probe in order to avoid contact between the needle and the probe. As soon as the needle had penetrated the subcutaneous tissues, its progress was real‐time monitored on the US image. The needle was visualised as a hyperechoic structure with posterior comet‐tail artefact. Progress of the needle until the tip had entered the subacromial bursa was observed. When the tip of the needle appeared to be inside the subacromial bursa, a small amount of liquid was injected to confirm the correct position (fluid passing from the needle tip into the bursa). The injection could be visualised in real‐time as a spreading "cloud" of hyperechoic echoes inside the bursa.	Ultrasound equipment with high frequency linear transducer. No further details provided	Summary Injection of 40 mg methylprednisolone with 1 cc lidocaine 2% into subacromial bursa; personnel not specified Procedure A posterior approach was used. The needle was placed approximately 2 cm below the posterior‐lateral aspect of the acromion. The needle was guided forward and slightly to the left under the acromion.	Not reported	Not reported
Hsieh 2013 (Taiwan)	Subacromial bursitis with shoulder pain, ≥ 4/10 pain during abduction on a VAS, a painful arc, and > 40% reduction in pain during shoulder abduction at end range after injection of 3 mL of 1% lidocaine into the subacromial bursa	Summary Injection of 0.5 mL (5 mg/mL) dexamethasone suspension and 3 mL lidocaine (10mg/ml) into subdeltoid/subacromial bursa by one senior physician who was a board‐certified rheumatologist, physiatrist, and ultrasonographer in musculoskeletal medicine Procedure After the sterilisation of the skin on the lateral side of the affected arm, a 21‐gauge needle was inserted into the subacromial bursa under ultrasound guidance. Any effusion, if present, was aspirated prior to the injection.	LOGIQ P5 machine (General Electronic Company, Milwaukee, WI) with 5‐ to 12‐MHz linear array transducer	Summary Injection of 0.5 mL (5 mg/mL) dexamethasone suspension and 3 mL lidocaine (10 mg/mL) into subacromial bursa by the same senior physician Procedure The physician injected the subacromial bursa according to Cyriax’s method, where he/she first localised the lateral edge of the acromion. He/she then asked the patient to relax the affected arm, and at 1–2 cm beneath the middle point of the lateral edge of the acromion, inserted a 21‐gauge needle medially and in a slightly cranial direction into the bursa. If the needle encountered resistance, either the coracoacromial ligament or the capsulotendinous structures had been contacted, and the needle position was slightly adjusted.	Not reported	Not reported
Lee 2009 (South Korea)	Adhesive capsulitis with pain and limited range of movement. Ultrasound used to rule out rotator cuff disease.	Summary Injection of 0.5 mL (20 mg) triamcinolone and 1.5 mL 2% lidocaine and 3 mL normal saline into glenohumeral joint by one physician with two years’ experience in ultrasound‐guided shoulder injections Procedure The posterior approach was used. Patients were seated, with the affected shoulder bent and adducted. The ultrasound probe touched the lower part of the acromion sideways, and a long needle (23G, 7 cm) was inserted to the posterior articular surface of the humeral head and positioned inside the articular capsule. The expansion of the articular capsule was checked while the fluid was being injected.	LOGIQ 5a machine with a 10‐MHz linear probe	Summary Injection of 0.5 mL (20 mg) triamcinolone and 1.5 mL 2% lidocaine and 3 mL normal saline into articular capsule by one physician with 7 years experience in anatomic landmark‐guided shoulder injections Procedure The posterior approach was used. Patients were seated, with the affected shoulder bent and adducted. The acromion of the scapula was palpated and the needle inserted 1 cm inferior to the tip of the acromion. The needle was directed toward the coracoid process and advanced into the articular capsule, after which the drug was injected.	Participants in both groups were taught exercises for increasing joint ROM including stretching forward and bending down to a desk, Codman exercise, and a wall‐climbing exercise with the fingers. They were instructed to practice these at home. Participants were checked and encouraged to keep up with the exercises every time they visited the hospital. All participants had 5 weekly injections of 25 mg sodium hyaluronate.	Not reported
Lee 2015 (South Korea)	Shoulder stiffness on the basis of < 100 degrees flexion or < 30 degrees external rotation or internal rotation at a level below the first lumbar spine junction. Radiography and ultrasonography were used to rule out rotator cuff disease.	Summary Injection of 40 mg triamcinolone acetonide and 2 mL of 2% lidocaine into glenohumeral joint by one experienced senior surgeon with more than 5 years of experience in performance of ultrasound‐guided injection at the shoulder joint Procedure The patient was supine with the arm by the side. After skin and transducer preparation with alcohol and povidone, the 21‐gauge 1.5‐inch needle on a 3 mL syringe with triamcinolone acetonide 40 mg and 2 mL of 2% lidocaine was inserted at the level of the coracoids, from lateral to medial, aimed at the medial border of the humeral head. When the needle made contact with the articular cartilage of the humeral head, the needle was tilted to position the point of the needle in the articular cavity. The intra‐articular position of the needle and fluid infiltration in the shoulder joint was confirmed by ultrasonography.	High‐resolution transducer with 12 MHz linear array	Summary Injection of 40 mg triamcinolone acetonide and 2 mL of 2% lidocaine into glenohumeral joint by the same senior surgeon Procedure With the patient in sitting, the sulcus between the lateral tip of the coracoid and the humeral head was palpated. Then, a 21‐gauge 1.5‐inch needle on a 3 mL syringe with triamcinolone acetonide 40 mg and 2 mL of 2% lidocaine was inserted in a slightly cephalad direction at the anterior‐lateral tip of the coracoid. The needle was slowly advanced while infiltrating the local anaesthetic until resistance was lost, indicating intra‐articular position. The injection was performed slowly.	Both groups received a home‐based exercise programme for rehabilitation involving: pendulum circumduction and passive shoulder exercises for self‐stretching in forward flexion, external rotation and internal rotation, pulley exercise, isometric exercise, thera‐band exercises, scapula stabilisation exercises, and strengthening using dumbbells.	None in either group
Naredo 2004 (Spain)	‘Periarticular disorders’ (impingement syndrome rotator cuff lesions, subacromial‐subdeltoid bursitis and/or biceps tendon abnormalities) diagnosed by clinical history and examination	Summary Injection of 20 mg triamcinolone into either subacromial‐subdeltoid bursa, biceps tendon sheath or rotator cuff calcifications according to ultrasound findings by single rheumatologists experienced in ultrasound Procedure The transducer and the patient’s skin were sterilised with alcohol. Sterile gel was applied to the probe. The transducer was held in one hand and the syringe with glucocorticoid in the other hand. The needle was placed under the probe and its route was visualised in real‐time by US, as a hyper‐reflective line, often with reverberation, from the skin to the target. Injection was directed into SA‐SD bursa or biceps tendon sheath when increased fluid was detected. When there was effusion in both the SA‐SD bursa and the biceps tendon sheath, injection was directed into the SA‐SD bursa due to its easier approach. Peri and intralesional injection was performed when rotator cuff calcifications were found. Perilesional injection directed into the SA‐SD bursa was performed when tendon lesions without increased fluid were detected, avoiding direct injection into the rotator cuff tendons.	Commercial equipment (Sonoline, Prima, Siemens, Seattle, WA, USA) using a 7.5 MHz linear phased array transducer	Summary Injection of 20 mg triamcinolone into subacromial space by same rheumatologist Procedure A 21‐gauge needle was used. The patient’s skin was sterilised with alcohol. Access to the subacromial space was achieved with a lateral approach, inserting the needle under the anterolateral aspect of the acromion process, passing it through the deltoid muscle, and directing it medially and slightly anterior to the subacromial‐subdeltoid bursa, with care taken to avoid injection directly into the tendons of the rotator cuff. Immediately after blind injection, an ultrasound examination was performed to search for steroid deposit as hyperechoic foci or lines, with or without acoustic shadowing.	Participants in both groups with loss of shoulder ROM were instructed to start a home physical therapy programme consisting of pendulum exercises and slow shoulder abduction. No restriction was placed on the participant’s ability to work or to use their shoulder as tolerated, or to take nonsteroidal anti‐inflammatory drugs. Participants did not receive physical therapy during the follow‐up period.	There were no serious adverse events in either group. 1/20 participants in the ‘blind’ injection group had a mild post‐injection adverse effect (not specified).
Raeissadat 2017 (Iran)	Adhesive capsulitis on the basis of findings from the history and physical examination	Summary Injection of 1 cm3 lidocaine 1%, then, 3 cm3 water soluble un‐ionised contrast with 1 cm3 distilled water, and finally, 1 cm3 triamcinolone 40 mg/cm3 with 1 cm3 lidocaine 1% into glenohumeral joint by one physical medicine and rehabilitation specialist with 15 years’ experience in his field Procedure The patient was seated and the affected hand was resting on the thigh. A glenohumeral joint injection was given using the posterior short axis approach, and the needle was inserted in the plane relative to the ultrasound probe and medial to the posterior aspect of the head of the humerus.	Alpinion E‐cube 7 ultrasound device with linear 3–12‐MHz probe	Summary Injection of 1 cm3 lidocaine 1%, then, 3 cm3 water soluble un‐ionised contrast with 1 cm3 distilled water, and finally, 1 cm3 triamcinolone 40 mg/cm3 with 1 cm3 lidocaine 1% into glenohumeral joint by the same specialist Procedure A glenohumeral joint injection was given using a posterior approach. A 25‐gauge needle was inserted 2.5 cm lower than the posterolateral aspect of the acromion. If the physician felt that the needle was not accurately inserted, he was not allowed to withdraw and re‐enter again.	After the injection, all participants in both groups received naproxen tablet 500 mg twice daily for 5 days and were told to perform Codman’s exercises.	None in either group
Roddy 2020 (United Kingdom)	Subacromial impingement syndrome with pain in deltoid insertion area, positive Neer and Hawkins Kennedy tests, and pain on shoulder abduction	Summary Injection of methylprednisolone 40 mg and 1 mL 1% lidocaine into the bursa by one of nine clinicians with extensive clinical experience performing ultrasound‐guided injections or completed an accredited course on ultrasound‐guided subacromial injections Procedure The skin and transducer were cleaned with chlorhexidine 0.5% solution and sterile gel applied to the transducer. The participant sat with the shoulder internally rotated and the ipsilateral hand on the buttock to maximise visibility of and access to the subacromial bursa. The transducer was placed anterolaterally, the hypoechoic subacromial bursa visualised, and a 21‐G needle inserted under real‐time ultrasound guidance until the needle‐tip entered the bursa. A commercially available premixed solution of methylprednisolone 40 mg and 1 mL 1% lidocaine was injected into the bursa.	LOGIQ e system with a 12 MHz transducer	Summary Injection of methylprednisolone 40 mg and 1 mL 1% lidocaine into the bursa by one of eight clinicians (different to those performing the ultrasound‐guided injections) with extensive clinical experience performing subacromial injections, and having attended a half‐day injection protocol workshop Procedure The participant sat with the arm hanging with the elbow bent and forearm resting on the lap. The skin was cleaned with chlorhexidine solution 0.5%. A 21‐G needle was inserted through the deltoid under the acromion process laterally. The same premixed solution of methylprednisolone 40 mg and 1 mL 1% lidocaine was injected.	None	Ultrasound‐guided injection and physiotherapist‐led exercise (n = 64): Serious adverse event: No. and nature of events: 1/64 ‐ pyelonephritis Other adverse event: No. and nature of events: 17/64 ‐ shoulder pain > 3 days following injection No. and nature of events: 16/64 ‐ discomfort during the injection or local skin changes, presyncope, nausea or flushing following the injection Ultrasound‐guided injection and exercise leaflet (n = 64): Other adverse event: No. and nature of events: 17/64 ‐ shoulder pain > 3 days following injection No. and nature of events: 16/64 ‐ discomfort during the injection or local skin changes, presyncope, nausea or flushing following the injection Unguided injection and physiotherapist‐led exercise (n = 64): Other adverse event: No. and nature of events: 20/64 ‐ shoulder pain > 3 days following injection No. and nature of events: 17/64 ‐ discomfort during the injection or local skin changes, presyncope, nausea or flushing following the injection Unguided injection and exercise leaflet (n = 64): Other adverse event: No. and nature of events: 20/64 ‐ shoulder pain > 3 days following injection No. and nature of events: 17/64 ‐ discomfort during the injection or local skin changes, presyncope, nausea or flushing following the injection
Saeed 2014 (Ireland)	Shoulder impingement pain with shoulder pain	Summary Injection of 40 mg of methylprednisolone acetate with 4 mL of lidocaine hydrochloride into either subacromial‐subdeltoid bursa, acromio‐clavicular joint, biceps tendon or other pathology according to ultrasound findings by one experienced physician. Procedure A two‐handed technique was used with the transducer held in one hand and the syringe with 21‐G needle in the other hand. The needle was directed in real time by ultrasound from the skin to the target (e.g. subdeltoid bursa, acromio‐clavicular joint or biceps pathology). In those shoulders where ultrasound demonstrated more than one pathology, the pathology most consistent with clinical examination was injected. When there was effusion in both the subacromial‐subdeltoid bursa and the biceps tendon sheath, injection was directed into the subacromial‐subdeltoid bursa. For confirmed rotator cuff pathology and inpatients with clinical impingement but normal shoulder, ultrasound subacromial‐subdeltoid bursa injection was performed. When ultrasound confirmed the principal pathology as acromioclavicular joint inflammation or biceps tendon inflammation, ultrasound‐guided injection of the principal pathological structure was performed.	Acuson Sequoia 512 ultrasound systems (Siemens, CA, USA) using 8L‐RS MHz linear phased‐array transducer	Summary Injection of 40 mg of methylprednisolone acetate with 4 mL of lidocaine hydrochloride into subacromial‐subdeltoid bursa by the same physician Procedure A standard technique was performed using a 21‐G needle via a lateral approach to inject the subacromial‐subdeltoid bursa.	Participants in both groups with the loss of shoulder range of movement were given post‐injection instructions in pendulum exercises and slow shoulder abduction. No restriction was placed on participants' ability to work or to use their shoulder as tolerated, or intake of NSAIDs. Participants did not receive physical therapy during the follow‐up period.	Apart from mild shoulder pain in a small number of cases (group not specified) within a few hours of shoulder injection, no serious adverse event was observed.
Ucuncu 2009 (Turkey)	‘Soft‐tissue lesions’ (acromioclavicular degeneration, rotator cuff lesions (rupture, partial rupture, tendinosis, impingement, calcification), fluid accumulation or partial rupture of biceps tendon and bursitis (subdeltoid, subacromial)) that had not responded satisfactorily to at least 1 month of nonsteroidal anti‐inflammatory drugs	Summary Injection of 1 mL (40 mg) triamcinolone and 1 mL 1% lidocaine into site of observed pathology. Injections were administered ‘perilesionally’ and ‘intralesionally’; personnel not specified. Procedure The injector was directed to the site of observed pathology beneath the probe by imaging a hyper‐reflective line.	Commercial equipment (Esaote, Mylab 60, Italy) using a 6 to 18MHz linear phased array transducer	Summary Injection of 1 mL (40 mg) triamcinolone and 1 mL 1% lidocaine into subacromial region; personnel not specified Procedure Injections were administered with a lateral entry approach to the sub‐acromial region.	Participants in both groups with loss of shoulder ROM performed a home exercise programme consisting of shoulder abduction and pendulum exercises. No limit was imposed on nonsteroidal anti‐inflammatory consumption or use of the shoulder. Participants did not receive physical therapy during the follow‐up period.	There were no serious adverse events in either group. 1/30 participants in the ultrasound‐guided injection group had mild post‐injection pain. In the ‘blind’ injection group, 5/30 had a slight increase in pain and 1/30 had skin peeling after the injection.
Zufferey 2012 (Switzerland)	Acute shoulder pain that had not responded satisfactorily to nonsteroidal anti‐inflammatory drugs or physiotherapy	Summary Injection of 2 mL of Diprophos® (7 mg of a mixture made up of one‐third soluble and two‐thirds long‐acting bethamethasone) into site of observed pathology by two rheumatologists experienced in musculoskeletal ultrasonography Procedure Steroid injection under ultrasound guidance was performed according to the ultrasound diagnosis within two days of the clinical evaluation. The injection was directed into the location assessed by ultrasound to be the cause of shoulder pain.	5‐9 MHz probe using a Philips HD11 machine in two centres and Easote mylab25 gold machine in one centre	Summary Injection of 2 mL of Diprophos® (7 mg of a mixture made up of one‐third soluble and two‐thirds long‐acting bethamethasone) into subacromial bursa; personnel not specified Procedure Blind injections were directed at the subacromial bursa.	Rescue analgesia was permitted for both groups and participants recorded their intake of nonsteroidal anti‐inflammatory drugs and paracetamol until the end of the study. New infiltrations after two weeks were tolerated but considered a poor result.	There were no serious adverse events in either group. 3/32 in the ultrasound‐guided group and 3/35 in the ‘blind’ injection group required a second local injection after 2 or 6 weeks. Four patients reported flushing and one diabetic patient a transient hyperglycaemia, however, there were no details of which groups these participants belonged to.

MHz = megahertz; mL = millilitre; mg = milligrams; cm = centimetre; T6 = sixth thoracic vertebrae; cc = cubic centimetre; mmol = millimole; ROM = range of motion; SA‐SD bursa = subacromial‐subdeltoid bursa; 21‐G = 21 gauge; NSAIDs = non‐steroidal anti‐inflammatory drugs

Image guidance in all trials utilised ultrasound. In four trials, ultrasound was used to direct the injection to the site of observed pathology (Naredo 2004; Saeed 2014; Ucuncu 2009; Zufferey 2012), while in the other trials the ultrasound‐guided injection was into either the subacromial bursa (Akbari 2020; Azadvari 2020; Bhayana 2018; Chen 2006; Cho 2010; Cole 2016; Dogu 2012; Ekeberg 2009; Haghighat 2015; Hsieh 2013; Roddy 2020) or glenohumeral joint (Cho 2021; Lee 2009; Lee 2015; Raeissadat 2017).

The landmark‐guided injections in the comparator groups were into the subacromial space (Akbari 2020; Azadvari 2020; Bhayana 2018; Chen 2006; Cho 2010; Cole 2016; Dogu 2012; Haghighat 2015; Hsieh 2013; Naredo 2004; Roddy 2020; Saeed 2014; Ucuncu 2009; Zufferey 2012) or glenohumeral joint (Cho 2021; Lee 2009, Lee 2015 and Raeissadat 2017), except for one trial that compared the ultrasound‐guided injection into the subacromial bursa to an intramuscular injection in the upper gluteal region (Ekeberg 2009).

The dose and type of glucocorticoid and local anaesthetic used varied between studies. Three studies used 20 mg of triamcinolone (Ekeberg 2009; Lee 2009; Naredo 2004) and four trials used 40 mg of triamcinolone (Cho 2021; Lee 2015; Raeissadat 2017; Ucuncu 2009). Five studies used 40 mg of methylprednisolone (Akbari 2020; Cole 2016; Haghighat 2015; Roddy 2020; Saeed 2014) and one study used 80 mg of methylprednisolone (Bhayana 2018). One study used 7 mg of betamethasone (Zufferey 2012), one study used 1 mL of methylprednisolone (Azadvari 2020), one study used 1 mL of betamethasone (dosage not specified) (Chen 2006) and one study used 5 mg of betamethasone with 0.02 mL of 0.01 mmol gadolinium diethylenetriaminepentaacetic (Dogu 2012). One study used 2.5 mg of dexamethasone (Hsieh 2013) and one study used 5 mL of triamcinolone (dosage not specified) (Cho 2010).

Co‐interventions:

Co‐interventions were reported in 12 trials. Nine trials gave all participants a home‐exercise programme (Azadvari 2020; Bhayana 2018; Cho 2021; Lee 2009; Lee 2015; Naredo 2004; Raeissadat 2017; Saeed 2014; Ucuncu 2009), eight allowed participants to have analgesia as required (Bhayana 2018; Cole 2016; Dogu 2012; Ekeberg 2009; Naredo 2004; Saeed 2014; Ucuncu 2009; Zufferey 2012), one trial allowed physical therapy as required (Cole 2016), one allowed participants to continue with physical therapy if they were receiving it at baseline (Ekeberg 2009), one encouraged participants to take a naproxen tablet 500 mg twice daily for five days (Raeissadat 2017), and one provided all participants with antibiotics to prevent infection (Bhayana 2018). All participants in Lee 2009 also received intra‐articular injections of 25 mg low molecular weight hyaluronic acid weekly for five weeks.

Outcomes

Major outcomes

Pain

Seventeen trials assessed pain. Pain was assessed on a 10 cm VAS in 12 trials (Akbari 2020; Azadvari 2020; Cho 2010; Cho 2021; Dogu 2012; Haghighat 2015; Hsieh 2013; Lee 2009; Lee 2015; Raeissadat 2017; Saeed 2014; Ucuncu 2009), 100 cm VAS in three trials (Bhayana 2018; Cole 2016; Naredo 2004), a zero to 10‐point numerical rating scale (NRS) in two trials (Roddy 2020; Zufferey 2012) and one to nine‐point ordinal scale in one trial (Ekeberg 2009). Three trials measured pain at rest and with activity (Dogu 2012; Ekeberg 2009; Zufferey 2012) and three during sleep (Cole 2016; Dogu 2012; Naredo 2004). The following pain measures were only reported in single trials: pain in daytime and just before sleep (Lee 2009), pain in the previous week (Naredo 2004), pain during Hawkin’s impingement test (Cho 2010), pain with overhead activities (Cole 2016), current pain (Roddy 2020) and the frequency of pain during activity, during sleep and extreme pain (Cole 2016). Nine trials did not specify the context in which pain was measured (Akbari 2020; Azadvari 2020; Bhayana 2018;Cho 2021; Haghighat 2015; Hsieh 2013; Lee 2015; Raeissadat 2017; Saeed 2014; Ucuncu 2009).

Function

Eighteen trials assessed function. Six trials measured function using the Constant score (Akbari 2020; Azadvari 2020; Bhayana 2018; Dogu 2012; Ucuncu 2009; Zufferey 2012). Zufferey 2012 excluded the last item on strength and Dogu 2012 reported scores separately for the following domains: range of motion, general pain and activities of daily living. Four studies used the Shoulder Pain And Disability Index (SPADI) (Ekeberg 2009; Haghighat 2015; Hsieh 2013; Roddy 2020), two used the Shoulder Disability Questionnaire (SDQ) (Dogu 2012; Hsieh 2013) and three used the American Shoulder and Elbow Surgeons (ASES) score (Cho 2021; Cole 2016; Lee 2015). The following measures of function were only reported in single studies: Western Ontario Rotator Cuff Index (Ekeberg 2009), Shoulder Function Assessment (SFA) scale (Naredo 2004), Disability of Arm Shoulder and Hand (DASH) scale (Akbari 2020), subjective shoulder value (Cho 2021), functional activities of the shoulder (Lee 2009), Oxford questionnaire (Raeissadat 2017), University of California‐Los Angeles (UCLA) shoulder rating scale (Cho 2010), and Shoulder Function Tests (SFTs) score (Saeed 2014).

 Zufferey 2012 did not report data on function at 12 weeks.

Participant‐rated treatment success

Five trials included a dichotomous measure of treatment success (Cole 2016; Naredo 2004; Roddy 2020; Ucuncu 2009; Zufferey 2012). Cole 2016 measured the number of participants with 50% improvement in pain during internal and external rotation impingement tests. Naredo 2004 measured the number of participants with 50% improvement in pain and SFA score. Roddy 2020 measured the number of participants who had completely recovered or felt much better. Ucuncu 2009 and Zufferey 2012 measured the number of participants with 50% improvement in pain scores.

Health‐related quality of life

Two trials measured health‐related quality of life using the Short‐Form Health Survey (SF‐36) Mental Component Score (Hsieh 2013; Roddy 2020). Roddy 2020 also measured the SF‐36 Physical Component Score. Azadvari 2020 used the BREF questionnaire (no further description was reported; we assumed it was the World Health Organization WHOQOL‐BREF quality of life assessment).

Adverse events

Twelve trials reported on adverse events (Akbari 2020; Bhayana 2018; Cho 2021; Dogu 2012; Ekeberg 2009; Lee 2015; Naredo 2004; Raeissadat 2017; Roddy 2020; Saeed 2014; Ucuncu 2009; Zufferey 2012). Of these, five trials did not specify how they assessed adverse events but reported them in the results (Akbari 2020; Bhayana 2018; Cho 2021; Lee 2015; Raeissadat 2017).

Serious adverse events

Five trials (Ekeberg 2009; Naredo 2004; Saeed 2014; Ucuncu 2009; Zufferey 2012) reported that there were no serious side effects. The remaining trials did not report the incidence of serious adverse events.

Withdrawals due to adverse events

One trial reported withdrawals due to adverse events (Ekeberg 2009).

Minor outcomes

Additional treatment or surgery

Three trials reported whether participants received additional injections (Cole 2016; Ekeberg 2009; Zufferey 2012).

One trial reported the number of participants that went on to undergo surgery at the end of the study (Cole 2016).

Range of motion

Sixteen out of 19 trials included at least one measure of shoulder range of motion (excluding Azadvari 2020; Dogu 2012; Roddy 2020). Bhayana 2018, Naredo 2004 and Saeed 2014 measured range of motion but did not report their results. Saeed 2014 did not report what movements were measured. Akbari 2020 did not report SD for their measures of range of motion.

Flexion

Ten trials measured range of shoulder flexion. Active flexion was measured in six trials (Akbari 2020; Cho 2010; Ekeberg 2009; Hsieh 2013; Naredo 2004; Ucuncu 2009) and passive flexion in six (Cho 2021; Cole 2016; Hsieh 2013; Lee 2009; Naredo 2004; Ucuncu 2009). Haghighat 2015, Lee 2015 and Raeissadat 2017 did not specify whether flexion was measured actively or passively.

Abduction

Ten trials measured range of shoulder abduction. Active abduction was measured in six trials (Akbari 2020; Ekeberg 2009; Hsieh 2013; Naredo 2004; Ucuncu 2009; Zufferey 2012) and passive abduction in seven (Cho 2021; Cole 2016; Hsieh 2013; Lee 2009; Naredo 2004; Ucuncu 2009; Zufferey 2012). Chen 2006, Haghighat 2015 and Raeissadat 2017 did not specify whether abduction was measured actively or passively.

External rotation

Nine trials measured shoulder external range of motion. Active external rotation was measured in four trials (Cho 2010; Hsieh 2013; Naredo 2004; Zufferey 2012) and passive external rotation in six (Cho 2021; Cole 2016; Hsieh 2013; Lee 2009; Naredo 2004; Zufferey 2012). Haghighat 2015, Lee 2015 and Raeissadat 2017 did not specify whether external rotation was measured actively or passively. Lee 2015 assessed external rotation with the arm by the side and at 90 degrees of abduction. We only considered shoulder external range of motion measures relevant for the three trials including participants with adhesive capsulitis (Lee 2009; Lee 2015; Raeissadat 2017).

Excluded studies

We excluded 17 studies in our original review after full‐text review and excluded an additional six studies for this update (see Characteristics of excluded studies table). For the original 17 studies, seven studies were excluded because they were not randomised (Alfredson 2006; Bamji 2004; Esenyel 2003; Gutierrez 2004; Koes 2009; McCormack 2009; Yi 2006); two studies did not include the participants of interest (Cohen 2009; Sibbitt 2009); four studies did not include the interventions of interest (Buchbinder 2004; Henkus 2006; Kang 2008; Widiastuti‐Samekto 2004); two studies did not include the comparator of interest (Chavez‐Lopez 2009; Tveita 2008); and one study only considered accuracy of needle placement (Rutten 2007). One study was excluded because data were not presented separately for the treatment groups of interest (Valtonen 1978).

Of the six trials excluded from the update, two were not randomised (Hashiuchi 2010; Micu 2013); two did not compare image‐guided to non‐image‐guided injection (Kim 2016; Kim 2017); one did not include participants of interest (Zhang 2011 included participants with biceps brachii tendinitis); and one did not include the intervention of interest (Graber 1997).

Risk of bias in included studies

The assessment of each domain of risk of bias for the included trials is summarised in the Characteristics of included studies table. The results of the risk of bias assessment are also presented graphically in Figure 2.

2.

Risk of bias summary: review authors' judgements about each risk of bias item for each included study.

Only two trials met all low risk of bias criteria (Cho 2021; Ekeberg 2009). The most notable sources of bias in the remaining trials included lack of blinding of participants and personnel (performance bias) which affected all but three trials (Cole 2016; Dogu 2012; Lee 2009) and lack of blinding of outcome assessment for self‐reported outcomes (detection bias) which affected all but four trials (Chen 2006; Cole 2016; Dogu 2012; Lee 2009).

Allocation

Two trials reported adequate random sequence generation and allocation concealment, and were deemed to have low risk of selection bias (Cho 2021; Ekeberg 2009). Two trials had inadequate random sequence generation and allocation concealment, and were deemed to have high risk of bias (Chen 2006; Lee 2009). Five trials were judged to have unclear risk of selection bias due to failure to report their method of randomisation (Bhayana 2018; Haghighat 2015; Naredo 2004; Roddy 2020; Ucuncu 2009) and thirteen trials were judged to have unclear risk of selection bias due to failure to report their method of allocation concealment (Bhayana 2018; Cho 2010; Cole 2016; Dogu 2012; Haghighat 2015; Hsieh 2013; Lee 2015; Naredo 2004; Raeissadat 2017; Roddy 2020; Saeed 2014; Ucuncu 2009; Zufferey 2012).

Blinding

Only five trials reported adequate blinding of participants and personnel, and were deemed to have low risk of performance bias (Cho 2021; Cole 2016; Dogu 2012; Ekeberg 2009; Lee 2009). Blinding of participants was achieved in three trials by applying the ultrasound probe to the shoulder in both treatment groups but without using it as an aid to injection in the non‐image‐guided groups. In Ekeberg 2009, all participants received an ultrasound‐guided shoulder injection as well as an upper gluteal injection but the interventions could have been either short‐acting local anaesthetic alone or combined with glucocorticoid.

Fourteen trials were judged to have high risk of performance bias due to inadequate blinding (Akbari 2020; Azadvari 2020; Bhayana 2018; Chen 2006; Cho 2010; Haghighat 2015; Hsieh 2013; Lee 2015; Naredo 2004; Raeissadat 2017; Roddy 2020; Saeed 2014; Ucuncu 2009; Zufferey 2012).

Only five trials had adequate blinding for self‐reported outcomes (Cho 2021; Cole 2016; Dogu 2012; Ekeberg 2009; Lee 2009) and were considered to be at low risk of detection bias for these outcomes. An additional trial that did not blind participants did not include self‐reported outcomes so was also considered at low risk (Chen 2006). Six trials were judged to have adequate blinding for assessor‐reported outcomes (Chen 2006; Cho 2021; Cole 2016; Dogu 2012; Ekeberg 2009; Lee 2009) and were deemed to have low risk of detection bias for these outcomes.

Incomplete outcome data

Risk of attrition bias was low in 15 trials (Akbari 2020; Azadvari 2020; Bhayana 2018; Cho 2010; Cho 2021; Cole 2016; Dogu 2012; Ekeberg 2009; Hsieh 2013; Lee 2009; Lee 2015; Raeissadat 2017; Roddy 2020; Saeed 2014; Ucuncu 2009).

In Azadvari 2020, Bhayana 2018, Cho 2010, Cole 2016, Dogu 2012, Ucuncu 2009 and Raeissadat 2017, there was no loss to follow‐up. In Akbari 2020, all 14 participants in the ultrasound‐guided injection group completed follow‐up, while one of 15 (7%) participants in the non‐image‐guided injection group was lost at final follow‐up. In Cho 2021, 4/45 (9%) in the ultrasound‐guided injection group and 6/45 (13%) in the non‐image‐guided injection group were lost at final follow‐up. In Ekeberg 2009, all 53 participants who received ultrasound‐guided injection completed follow‐up, while 2/53 (4%) in the group who received systemic glucocorticoid were lost to follow‐up at six weeks. In Hsieh 2013, 2/48 (4%) in both the ultrasound‐guided and non‐image‐guided injection groups were lost to follow‐up at one month follow‐up. In Lee 2009, 2/22 (9%) in the non‐image‐guided injection group and 1/21 (5%) in the ultrasound‐guided injection group were lost to follow‐up at six weeks, although the number of participants with available data at one week (main end point) was unknown. In Lee 2015, 7/45 (15%) in the ultrasound‐guided injection group and 6/45 (13%) in the non‐image‐guided injection group were lost at final follow‐up. In Roddy 2020, 8/64 (13%) in the ultrasound‐guided injection group and 7/64 (11%) in the non‐image‐guided injection group were lost at 12 months follow‐up. In Saeed 2014, 11/50 (22%) in the non‐image‐guided injection group and 9/50 (18%) in the ultrasound‐guided injection group did not complete the 12‐week assessment and were excluded from the study after the six‐week assessments; all withdrawals were due to the need for a repeat injection or referral for surgery.

Risk of attrition bias was unclear in four trials (Chen 2006; Haghighat 2015; Naredo 2004; Zufferey 2012). Three trials did not provide any information on the number of participants with available data at any time point (Chen 2006; Haghighat 2015; Naredo 2004). In Zufferey 2012, 1/36 (3%) and 7/36 (19%) in the non‐image‐guided injection group and 2/34 (6%) and 7/34 (21%) in the ultrasound‐guided injection group were lost to follow‐up at six weeks (primary end point) and 12 weeks, respectively.

Selective reporting

Risk of reporting bias was low in seven trials (Cho 2021; Cole 2016; Dogu 2012; Ekeberg 2009; Haghighat 2015; Raeissadat 2017; Ucuncu 2009) as all proposed outcomes were reported in the results. Risk of reporting bias was high in two trials (Naredo 2004; Saeed 2014) as not all measured shoulder movements were reported in the results. Further, we could not include data for function from one trial (Saeed 2014) as we were unable to obtain an SD. Risk of reporting bias was unclear in ten trials (Akbari 2020; Azadvari 2020; Bhayana 2018; Chen 2006; Cho 2010; Hsieh 2013; Lee 2009; Lee 2015; Roddy 2020; Zufferey 2012). All of them other than Roddy 2020 were not prospectively registered. Roddy 2020 did not report on all outcomes in the published paper that were to be measured according to the protocol. Lee 2009 stated that means and SDs were reported for continuous outcomes, but the SDs seemed very small in comparison with the reported SDs in other trials, raising the possibility that these were standard errors.

Other potential sources of bias

All but one trial was deemed to be at low risk of other potential sources of bias (Haghighat 2015). In Haghighat 2015, baseline function was higher (worse) in the ultrasound‐guided injection group, which may have biased the result in favour of this intervention.

Effects of interventions

See: Table 1

See: Table 1 for efficacy of ultrasound‐guided injection versus non‐image‐guided injection for shoulder pain.

Benefits

Overall pain

We pooled data from 18 trials. Nine trials reported pain up to three weeks (Azadvari 2020; Bhayana 2018; Cho 2010; Cho 2021; Ekeberg 2009; Hsieh 2013; Lee 2009; Raeissadat 2017; Zufferey 2012), 15 from > three weeks to six weeks (Akbari 2020; Bhayana 2018; Cho 2021; Cole 2016; Dogu 2012; Ekeberg 2009; Haghighat 2015; Hsieh 2013; Lee 2009; Naredo 2004; Raeissadat 2017; Roddy 2020; Saeed 2014; Ucuncu 2009; Zufferey 2012), seven from > six weeks to three months (Azadvari 2020; Bhayana 2018; Cho 2021; Lee 2009; Lee 2015; Saeed 2014; Zufferey 2012), two trials from > three months to six months (Lee 2015; Roddy 2020) and two trials at > six months (Lee 2015; Roddy 2020). Azadvari reported that mean (SD) pain was 0.0 (0.0) at two months follow‐up, and thus we could not include these data in our > six weeks to three months analysis. Statistical heterogeneity was low to considerable (I² = 91% up to three weeks, I² = 74% from > three weeks to six weeks, I² = 0% from > six weeks to three months, I²= 0% from > three months to six months, and I² = 0% at > six months).

Moderate‐certainty evidence (downgraded for risk of bias) indicates that injection using ultrasound guidance probably has little to no effect on pain at > three weeks to six weeks when compared to injection performed without image guidance (Analysis 1.1; Table 1). Mean pain (0 to 10, lower is better) was 3.1 points with injection not guided by imaging and 0.5 points better (lower) (95% CI 0.2 to 0.8 points better); 15 studies, 1003 participants) with an ultrasound‐guided injection. Repeating this analysis, but excluding Lee 2009, based upon uncertainty about whether SEs rather than SDs were reported, did not appreciably alter the results (mean difference ‐0.6 points lower (95% CI 0.9 points lower to 0.2 points lower)). A funnel plot for data from > three weeks to six weeks did not indicate publication bias (Figure 3).

1.1. Analysis.

Comparison 1: Ultrasound‐guided injection versus anatomic landmark‐guided or systemic injection, Outcome 1: Overall pain (or daytime, activity‐related or unspecified)

3.

Funnel plot of comparison: 1 Ultrasound‐guided injection versus non‐image‐guided injection, outcome: 1.1 Overall pain (or daytime, activity‐related or unspecified).

Up to three weeks, mean pain was 2.7 with a non‐image‐guided injection and 0.7 points better (95% CI 0.0 better to 1.3 better; 554 participants) with an ultrasound‐guided injection. From > six weeks to three months, mean pain was 2.8 with a non‐image‐guided injection and 0.0 points better (95% CI 0.1 worse to 0.2 points better; 454 participants) with an ultrasound‐guided injection. From > three months to six months, mean pain was 3.4 with a non‐image‐guided injection and 0.6 points better (95% CI 0.3 points worse to 1.5 points better; 205 participants) with an ultrasound‐guided injection. At > six months, mean pain was 3.2 with a non‐image‐guided injection and 0.2 points better (95% 1.1 points worse to 0.7 points better; 205 participants) with an ultrasound‐guided injection.

Function

We pooled data from 17 trials. Eight trials reported function up to three weeks (Azadvari 2020; Bhayana 2018; Cho 2010; Cho 2021; Ekeberg 2009; Hsieh 2013; Lee 2009; Raeissadat 2017), 14 from > three weeks to six weeks (Akbari 2020; Bhayana 2018; Cho 2021; Cole 2016; Dogu 2012; Ekeberg 2009; Haghighat 2015; Hsieh 2013; Lee 2009; Naredo 2004; Raeissadat 2017; Roddy 2020; Ucuncu 2009; Zufferey 2012), four from > six weeks to three months (Azadvari 2020; Bhayana 2018; Cho 2021; Lee 2015), two from > three months to six months (Lee 2015; Roddy 2020) and two from > six months (Lee 2015; Roddy 2020). Statistical heterogeneity was low to considerable (I² = 92% up to three weeks, I² = 61% from three weeks to six weeks, I² = 97% from six weeks to three months, I² = 0 from three months to six months, and I² = 0 at > six months).

Moderate‐certainty evidence (downgraded for risk of bias) indicates that ultrasound‐guided injection probably has little to no effect on function at > three to six weeks when compared to injection that does not use image guidance (Analysis 1.2; Table 1). Mean function was 68 points on a 0 to 100‐point scale (a higher score is better) with an injection not guided by imaging and 2.4 points better (95% CI 0.2 points worse to 5.1 points better; 895 participants) with an ultrasound‐guided injection. A funnel plot for data from > three weeks to six weeks did not indicate publication bias (Figure 4).

1.2. Analysis.

Comparison 1: Ultrasound‐guided injection versus anatomic landmark‐guided or systemic injection, Outcome 2: Function

4.

Funnel plot of comparison: 1 Ultrasound‐guided injection versus non‐image‐guided injection, outcome: 1.2 Function.

Up to three weeks, mean function was 72 points with non‐image‐guided injection and 5.5 points better (95% CI 0.3 points worse to 11.2 points better; 487 participants) with an ultrasound‐guided injection. From > six weeks to three months, mean function was 77.4 points with a non‐image‐guided injection and 9.2 points better (95% CI 5.8 points worse to 24.1 points better; 257 participants) with an ultrasound‐guided injection. Removing data from Azadvari 2020 reduced the I² to 0 and changed the effect estimate to 3.7 points better (95% CI: 0.5 points better to 6.9 points better; 227 participants). From > three months to six months, mean function was 73.5 points with non‐image‐guided injection and 1.1 points worse (95% 9.3 points worse to 7.2 points better; 205 participants) with an ultrasound‐guided injection. From > six months, mean pain was 72.8 with a non‐image‐guided injection and 3.4 points better (95% CI 4.7 points worse to 11.4 points better; 205 participants) with an ultrasound‐guided injection.

Participant‐rated treatment success

We pooled data from five trials that measured participant‐rated treatment success using a dichotomous outcome (Cole 2016; Naredo 2004; Roddy 2020; Ucuncu 2009; Zufferey 2012) (Analysis 1.3). Statistical heterogeneity was substantial (I² = 77%). Ultrasound‐guided injection may reduce, increase or have little to no effect on participant‐rated treatment success (at end of study) when compared to non‐image guided injection as the evidence is very uncertain (very low‐quality evidence, downgraded for risk of bias, inconsistency and imprecision), although the 95% CI includes a possible benefit. One hundred and one out of 175 (60.6%) in the ultrasound‐guided group reported treatment success compared with 68/175 (38.9%) in the group who received a non‐image‐guided injection; RR 1.56, 95% CI: 0.89 to 2.75, an absolute difference of 22% (4% fewer to 62% more) and relative difference of 56% (11% fewer to 175% more).

1.3. Analysis.

Comparison 1: Ultrasound‐guided injection versus anatomic landmark‐guided or systemic injection, Outcome 3: Treatment success (50% pain improvement on VAS)

Health‐related quality of life

Three trials measured health‐related quality of life (Azadvari 2020; Hsieh 2013; Roddy 2020).Two trials measured health‐related quality of life up to three weeks (Azadvari 2020; Hsieh 2013), two from > three weeks to six weeks (Hsieh 2013; Roddy 2020), one from > six weeks to three months (Azadvari 2020), and one from > three months to six months and > six months (Roddy 2020). Statistical heterogeneity was low (I² = 0% from three weeks to six weeks).

Low‐certainty evidence (downgraded for risk of bias and imprecision) indicates that ultrasound‐guided injections may have little to no effect on quality of life at > three weeks to six weeks when compared to injections performed without image guidance (Analysis 1.4; Table 1). Mean quality of life was 65.0 points on a 0 to 100 scale (higher is better quality of life) with a non‐image‐guided injection and 2.8 points better (95% CI 0.7 worse to 6.4 points better; 220 participants) with an ultrasound‐guided injection.

1.4. Analysis.

Comparison 1: Ultrasound‐guided injection versus anatomic landmark‐guided or systemic injection, Outcome 4: Quality of life

Up to three weeks, mean quality of life was 67.0 points with a non‐image‐guided injection and 4.2 points better (95% CI 1.0 point worse to 9.3 points better; 122 participants) with an ultrasound‐guided injection. From > six weeks to three months, mean quality of life was 64.0 points with a non‐image‐guided injection and 4.6 points better (95% CI 1.3 points worse to 10.5 points better; 30 participants) with an ultrasound‐guided injection. From > three months to six months, mean quality of life was 64 points with a non‐image‐guided injection and 1.6 points better (95% CI 2.5 points worse to 5.7 points better; 128 participants) with an ultrasound‐guided injection. From > six months, mean quality of life was 64 points with a non‐image‐guided injection and 0.7 points worse (95% CI 4.9 points worse to 3.5 points better; 128 participants) with an ultrasound‐guided injection.

Harms

Adverse events

Twelve studies reported on adverse events (Akbari 2020; Bhayana 2018; Cho 2021; Dogu 2012; Ekeberg 2009; Lee 2015; Naredo 2004; Raeissadat 2017; Roddy 2020; Saeed 2014; Ucuncu 2009; Zufferey 2012). Seven trials reported that there were no adverse events (Akbari 2020; Bhayana 2018; Cho 2021; Dogu 2012; Lee 2015; Raeissadat 2017; Saeed 2014). We pooled data from the remaining five studies that reported numbers of adverse events (Ekeberg 2009; Naredo 2004; Roddy 2020; Ucuncu 2009; Zufferey 2012). There was low statistical heterogeneity (I² = 15%). Low‐certainty evidence (downgraded for risk of bias and imprecision ‐ low event rates) indicates that ultrasound‐guided injection may not reduce the risk of adverse events compared to non‐image‐guided injection (Analysis 1.5). Thirty‐eight out of 200 (18%) in the ultrasound‐guided group reported adverse events compared with 51/202 (25%) in the non‐image‐guided injection group; RR 0.72 (95% CI 0.4 to 1.28), an absolute difference of 7% fewer adverse events (15% fewer to 7% more) and a relative difference of 28% fewer events (60% fewer to 28% more).

1.5. Analysis.

Comparison 1: Ultrasound‐guided injection versus anatomic landmark‐guided or systemic injection, Outcome 5: Number of adverse events

Withdrawals due to adverse events

Only one trial reported withdrawals due to adverse events (Ekeberg 2009): 1/53 in the systemic injection group withdrew (and received an additional local injection at two weeks), and 0/53 in the ultrasound‐guided group withdrew (Analysis 1.6).

1.6. Analysis.

Comparison 1: Ultrasound‐guided injection versus anatomic landmark‐guided or systemic injection, Outcome 6: Withdrawals due to adverse events

Serious adverse events

Five trials (Ekeberg 2009; Naredo 2004; Saeed 2014; Ucuncu 2009; Zufferey 2012) reported that there were no serious side effects. One trial (Roddy 2020) reported that there was one case of pyelonephritis (1/64; 2%) in the ultrasound‐guided injection group. The remaining trials did not report the incidence of serious adverse events.

Minor outcomes

Additional injections

We pooled data from three trials that reported whether participants received additional injections (Cole 2016; Ekeberg 2009; Zufferey 2012). Statistical heterogeneity was low (I² = 9%). Low‐certainty evidence (downgraded for risk of bias and imprecision) indicates that ultrasound‐guided injection probably does not reduce the need for additional injections compared to non‐image‐guided injection (Analysis 1.7). Three out of 113 (2.7%) in the ultrasound‐guided group required an additional injection compared with 8/116 (6.9%) in the non‐image‐guided injection group; RR 0.56, 95% CI: 0.14 to 2.15, an absolute difference of 3.2% fewer additional injections with ultrasound‐guided injection (6.0% fewer to 8.0% more) and relative difference of 44% fewer (86% fewer to 115% more).

1.7. Analysis.

Comparison 1: Ultrasound‐guided injection versus anatomic landmark‐guided or systemic injection, Outcome 7: Additional injections or surgery

Surgery

One trial (Cole 2016) reported that 4/28 (14.3%) participants who received an ultrasound‐guided injection needed surgery, compared to 6/28 (21.4%) participants who received a non‐image‐guided injection. Low‐certainty evidence (downgraded for risk of bias and imprecision) indicates that ultrasound‐guided injection probably does not reduce the need for surgery compared to non‐image‐guided injection (Analysis 1.7); RR 0.67, 95% CI: 0.21 to 2.11, an absolute difference of 7.2% fewer surgeries with ultrasound‐guided injection (17.5% fewer to 18.7% more) and relative difference of 33% (79% fewer to 111% more).

Range of shoulder flexion

We pooled data from 10 trials. Five trials reported shoulder flexion up to three weeks (Cho 2010; Cho 2021; Ekeberg 2009; Lee 2009; Raeissadat 2017); eight trials reported this outcome from > three weeks to six weeks (Akbari 2020; Cho 2021; Cole 2016; Ekeberg 2009; Haghighat 2015; Lee 2009; Raeissadat 2017; Ucuncu 2009), two trials reported this outcome from > six weeks to three months (Cho 2021; Lee 2015); one trial reported this outcome from > three months to six months (Lee 2015); and one trial reported this outcome at > six months (Lee 2015). Statistical heterogeneity was substantial to low (I² = 64% up to three weeks, I² = 27% from three weeks to six weeks, and I² = 0% from six weeks to three months).

Low‐certainty evidence (downgraded for risk of bias and indirectness) indicates that ultrasound‐guided injection provides a negligible improvement in shoulder flexion at > three to six weeks when compared to non‐image‐guided injection (Analysis 1.8). Mean shoulder flexion was 140.5 degrees with a non‐image‐guided injection and 1.1 degrees better (95% CI 2.3 degrees worse to 4.5 degrees better; 461 participants) with an ultrasound‐guided injection.

1.8. Analysis.

Comparison 1: Ultrasound‐guided injection versus anatomic landmark‐guided or systemic injection, Outcome 8: Range of flexion

Up to three weeks, mean shoulder flexion was 141.6 degrees with a non‐image‐guided injection and 3.5 degrees better (95% CI 1.6 degrees worse to 8.5 degrees better, 305 participants; normal range of flexion 180 degrees) with an ultrasound‐guided injection. From six weeks to three months, mean shoulder flexion was 136.9 degrees with a non‐image‐guided injection and 3.7 degrees better (95% 1.42 worse to 8.82 better; 167 participants) with an ultrasound‐guided injection. From > three months to six months, mean shoulder flexion was 142.7 degrees with a non‐image‐guided injection and 0.5 degrees better (95% 5.5 degrees worse to 6.5 degrees better; 77 participants) with an ultrasound‐guided injection. At > six months, mean shoulder flexion was 142.2 degrees with a non‐image‐guided injection and 4.0 degrees better (95% 4.5 degrees worse to 12.5 degrees better; 77 participants) with an ultrasound‐guided injection.

Range of shoulder abduction

We pooled data from 10 trials. Five trials reported shoulder abduction range of motion up to three weeks (Chen 2006; Cho 2021; Ekeberg 2009; Lee 2009; Raeissadat 2017), nine from three weeks to six weeks (Akbari 2020; Cho 2021; Cole 2016; Ekeberg 2009; Haghighat 2015; Lee 2009; Raeissadat 2017; Ucuncu 2009; Zufferey 2012), and one from six weeks to three months (Cho 2021). Statistical heterogeneity was substantial (I² = 87% up to three weeks, I² = 60% from three weeks to six weeks).

Low‐certainty evidence (downgraded due to risk of bias and indirectness) indicates that ultrasound‐guided injection may provide a slight improvement in shoulder abduction at > three to six weeks when compared to non‐image‐guided injection (Analysis 1.9). From > three to six weeks, mean shoulder abduction was 116.3 degrees with a non‐image‐guided injection and 6.9 degrees better (95% CI 1.5 degrees better to 12.2 degrees better; 528 participants) with an ultrasound‐guided injection.

1.9. Analysis.

Comparison 1: Ultrasound‐guided injection versus anatomic landmark‐guided or systemic injection, Outcome 9: Range of abduction

Up to three weeks, mean shoulder abduction was 122 degrees with a non‐image‐guided injection and 12.6 degrees better (95% CI 2.6 to 22.5 degrees better, 317 participants) with an ultrasound‐guided injection. From > six weeks to three months, mean shoulder abduction was 1.1 degrees better (95% CI 8.5 degrees worse to 10.7 degrees better; 90 participants) with an ultrasound‐guided injection.

Range of shoulder external rotation

We pooled data from eight trials which assessed shoulder external rotation range of motion. Four trials reported shoulder external rotation up to three weeks (Cho 2010; Cho 2021; Lee 2009; Raeissadat 2017), six from > three to six weeks (Cho 2021; Cole 2016; Haghighat 2015; Lee 2009; Raeissadat 2017; Zufferey 2012), two from > six weeks to three months (Cho 2021; Lee 2015), one from > three months to six months (Lee 2015) and one at > six months (Lee 2015). Statistical heterogeneity was substantial to low (I² = 71% up to three weeks, I² = 19% from three weeks to six weeks, and I² = 0% from six weeks to three months).

Low‐certainty evidence (downgraded due to risk of bias and indirectness) indicates that ultrasound‐guided injection may provide no improvement in shoulder external rotation at > three to six weeks when compared to non‐image‐guided injection (Analysis 1.10). From > three weeks to six weeks, mean shoulder abduction was 68.8 degrees with a non‐image‐guided injection and 0.7 degrees worse (95% CI 3.5 degrees worse to 2.0 degree better; 334 participants) with an ultrasound‐guided injection.

1.10. Analysis.

Comparison 1: Ultrasound‐guided injection versus anatomic landmark‐guided or systemic injection, Outcome 10: Range of external rotation

Up to three weeks, mean shoulder external rotation was 53.2 degrees with a non‐image‐guided injection and 1.0 degrees better (95% CI 3.9 degrees worse to 5.9 degrees better; 199 participants) with an ultrasound‐guided injection. From > six weeks to three months, mean shoulder external rotation was 77.6 degrees with a non‐image‐guided injection and 3.9 degrees better (95% 0.4 degrees worse to 8.2 degrees better; 167 participants) with an ultrasound‐guided injection. From > three months to six months, mean shoulder external rotation was 79.2 degrees with a non‐image‐guided injection and 3.2 degrees better (95% 8.1 degrees worse to 14.5 degrees better; 77 participants) with an ultrasound‐guided injection. From > six months, mean shoulder external rotation was 87.1 degrees with a non‐image‐guided injection and 2.7 degrees worse (95% 13.3 degrees worse to 7.9 degrees better; 77 participants) with an ultrasound‐guided injection.

Subgroup analysis

See Analysis 2.1 and Analysis 2.2. Fourteen trials included participants that could be categorised as having rotator cuff disease (Akbari 2020; Azadvari 2020; Bhayana 2018; Chen 2006; Cho 2010; Cole 2016; Dogu 2012; Ekeberg 2009; Haghighat 2015; Hsieh 2013; Naredo 2004; Roddy 2020; Saeed 2014; Ucuncu 2009). Four trials included participants with adhesive capsulitis (Cho 2021; Lee 2009; Lee 2015; Raeissadat 2017). One trial (Zufferey 2012) included participants with acute shoulder pain and did not provide a more specific diagnostic label. Stratifying the results for pain and function at > three to six weeks (primary time point) by shoulder condition did not appreciably alter the results; the test for subgroup differences indicated there were unlikely to be differences in pain and function across different shoulder conditions.

2.1. Analysis.

Comparison 2: Subgroup analyses (stratified by shoulder condition), Outcome 1: Overall pain

2.2. Analysis.

Comparison 2: Subgroup analyses (stratified by shoulder condition), Outcome 2: Function

Pain

We pooled data from 12 trials of participants with rotator cuff disease (Akbari 2020; Azadvari 2020; Bhayana 2018; Cole 2016; Dogu 2012; Ekeberg 2009; Haghighat 2015; Hsieh 2013; Naredo 2004; Roddy 2020; Saeed 2014; Ucuncu 2009), three trials of participants with adhesive capsulitis (Cho 2021; Lee 2009; Raeissadat 2017) and one trial of participants with mixed or undefined shoulder pain (Zufferey 2012). Statistical heterogeneity was low to substantial (I² = 79% for rotator cuff disease, I² = 0% for adhesive capsulitis). In participants with rotator cuff diease, mean pain was 3.1 points with a non‐image‐guided injection and 0.6 points better (95% CI 0.1 to 1.05 points better; 777 participants) with an ultrasound‐guided injection; in participants with adhesive capsulitis, mean function was 1.1 points with a non‐image‐guided injection and 0.2 points better (95% CI 0.0 points to 0.4 points better; 189 participants) with an ultrasound‐guided injection; and in participants with mixed or undefined shoulder pain, mean function was 4.2 points with a non‐image‐guided injection and 1.2 points better (95% CI 0.1 points worse to 2.5 points better; 65 participants) with an ultrasound‐guided injection.

Function

We pooled data from 11 trials of participants with rotator cuff disease (Akbari 2020; Azadvari 2020; Bhayana 2018; Cole 2016; Dogu 2012; Ekeberg 2009; Haghighat 2015; Hsieh 2013; Naredo 2004; Roddy 2020 ; Ucuncu 2009), three trials of participants with adhesive capsulitis (Cho 2021; Lee 2009; Raeissadat 2017) and one trial of participants with mixed or undefined shoulder pain (Zufferey 2012). Statistical heterogeneity was substantial to low (I² = 94% for rotator cuff disease, I² = 0% for adhesive capsulitis).

In participants with rotator cuff disease, mean function was 68 points with a non‐image‐guided injection and 5.1 points better (95% CI 3.2 points worse to 13.4 points better; 687 participants) with an ultrasound‐guided injection; in participants with adhesive capsulitis, mean function was 92.8 points with a non‐image‐guided injection and 1.1 points better (95% CI 1.2 points worse to 3.3 points better; 171 participants) with an ultrasound‐guided injection; and in participants with mixed or undefined shoulder pain, mean function was 47 points with a non‐image‐guided injection and 4.0 points better (95% CI 0.1 to 8.0 points better; 65 participants) with an ultrasound‐guided injection.

Sensitivity analysis

We performed a sensitivity analysis to investigate the robustness of pain and function (at > three to six weeks) to the presence of selection and detection biases.

Selection bias

Four trials were judged to be at low risk of selection bias (Akbari 2020; Azadvari 2020; Cho 2021; Ekeberg 2009).

Estimates for pain and function including all trials with eligible data may have been slightly influenced by the presence of selection bias. Mean pain was 3.1 with a non‐image‐guided injection and 0.3 points better (95% CI 0.2 points worse to 0.8 points better, 224 participants) with an ultrasound‐guided injection when pooling the three trials with low risk of selection bias who provided eligible data at this time point (Akbari 2020; Cho 2021; Ekeberg 2009). There was no statistical heterogeneity (I² = 0%). Mean function was 68 with a non‐image‐guided injection and 0.8 points better (95% CI 7.1 points worse to 8.6 points better, 124 participants) with an ultrasound‐guided injection when pooling the three trials with low risk of selection bias that provided eligible data at this time point (Akbari 2020; Cho 2021; Ekeberg 2009). Statistical heterogeneity was moderate (I² = 55%).

However, the results of these sensitivity analyses are in keeping with the results obtained by pooling all trials with eligible data irrespective of the risk of selection bias which found that image guidance probably provides little or no benefit in pain or function over injection without imaging.

Detection bias

Five trials were judged to be at low risk of detection bias for participant‐reported outcomes and could be pooled for pain and function at > three to six weeks (Cho 2021; Cole 2016; Dogu 2012; Ekeberg 2009; Lee 2009) . Statistical heterogeneity might not be important for pain (I² = 19%) and there was no statistical heterogeneity for function (I² = 0%).

Estimates for pain and function do not appear to have been influenced by the presence of detection bias. Mean pain was 3.1 (on a scale of 0 to 10, higher score indicates more pain) with non‐image‐guided injection and 0.3 points better (95% CI 0.1 to 0.5 points better; 338 participants) with an ultrasound‐guided injection when pooling the five trials with low risk of detection bias. This was not substantively different to the analysis that included all trials.

Mean function was 68 points (on a scale of 0 to 100, higher score indicates better function) with a non‐image‐guided injection and 2.4 points worse (95% CI 5.0 points worse to 0.1 points better; 338 participants) with an ultrasound‐guided injection when pooling the five trials with low risk of detection bias. This was also not substantively different to the analysis that included all trials.

Unit of analysis issues

We performed a post hoc sensitivity analysis to assess the robustness of pain against the potential unit of analysis issue from two trials that randomised by participant, but injected both shoulders (Cole 2016; Saeed 2014).

We included unadjusted data from these trials, as adjusted data were not available. Removal of data from these studies from the analysis of pain at the primary time point confirmed that the analysis was robust to this potential unit of analysis issue: the effect estimates changed little from an MD (95% CI) of ‐0.5 (‐0.2 to ‐0.8) including all trials with eligible data to ‐0.5 (‐0.1 to ‐0.9) excluding these two trials.

Discussion

Summary of main results

Nineteen trials involving 1035 participants compared ultrasound‐guided injections with injections performed without image guidance for shoulder pain. Compared with non‐image‐guided injections, moderate‐certainty evidence (downgraded for risk of bias) indicates that ultrasound‐guided injections probably provides little important improvement in pain and function at six weeks (Table 1). Subgroup analyses stratifying results by shoulder condition (i.e. rotator cuff disease, adhesive capsulitis and mixed shoulder pain), did not appreciably alter these estimates. Ultrasound‐guided injection may have little to no effect on quality of life (low‐certainty evidence, downgraded for risk of bias and imprecision), and it is uncertain if there is any difference in participant‐rated treatment success due to very low‐certainty evidence (downgraded for risk of bias, inconsistency and imprecision). There may be little to no difference in the risk of adverse events (low‐certainty evidence, downgraded for risk of bias and imprecision) between ultrasound‐guided and non‐image‐guided injection. Serious adverse events and withdrawals due to adverse events were rarely reported and thus risk estimates could not be calculated.

Furthermore, compared with injections performed without image guidance, ultrasound‐guided injection may provide little to no effect on shoulder flexion, abduction and external rotation (low‐certainty evidence, downgraded for risk of bias and indirectness), and ultrasound‐guided injection may not reduce the need for additional injections or surgery (low‐certainty evidence, downgraded for risk of bias and imprecision).

Overall, current evidence does not confirm an advantage of ultrasound‐guided injections for improving patient‐relevant outcomes, such as pain and function, over injections that do not use image guidance for shoulder pain.

Overall completeness and applicability of evidence

The 19 trials included in this review compared ultrasound‐guided injection with anatomic landmark injection (n = 18 trials) or systemic intramuscular injection (n = 1 trial; Ekeberg 2009) for shoulder pain. Trials were conducted in 11 countries. Fourteen trials included participants with rotator cuff disease (Bhayana 2018; Chen 2006; Cho 2010; Cole 2016; Dogu 2012; Ekeberg 2009; Haghighat 2015; Hsieh 2013; Naredo 2004; Saeed 2014; Ucuncu 2009), four included participants with adhesive capsulitis (Cho 2021; Lee 2009; Lee 2015; Raeissadat 2017), and one did not define their study population beyond 'shoulder pain' (Zufferey 2012). Nine trials relied upon clinical features alone for diagnosis (Bhayana 2018; Ekeberg 2009; Haghighat 2015; Hsieh 2013; Naredo 2004; Raeissadat 2017; Azadvari 2020; Cho 2010; Roddy 2020), while the remaining trials either used MRI or US or did not report how inclusion criteria were determined. Injections typically targeted the subacromial bursa for rotator cuff disease or the glenohumeral joint for adhesive capsulitis and glenohumeral osteoarthritis.

Most trials excluded potential participants who had received previous glucocorticoid injections in the shoulder; five at any time (Bhayana 2018; Cho 2010; Dogu 2012; Naredo 2004; Ucuncu 2009), five within the last three months (Cho 2021; Haghighat 2015; Hsieh 2013; Raeissadat 2017; Zufferey 2012), and two within the preceding month (Ekeberg 2009; Saeed 2014). The other seven trials did not specify whether or not participants had previously received glucocorticoid injections for shoulder pain (Akbari 2020; Azadvari 2020; Chen 2006; Cole 2016; Lee 2009; Lee 2015; Roddy 2020). Most trials also excluded full‐thickness rotator cuff tears and participants who had previous shoulder surgery.

Participants were similar in terms of mean age (ranging from 31 (Azadvari 2020) to 60 (Cho 2010) years), baseline pain and function (except participants in Raeissadat 2017 who reported substantially lower function), and gender ratio (percent female ranged from 22% (Lee 2015) to 73% (Ucuncu 2009)). Although the mean duration of symptoms varied across trials (from 2 (Haghighat 2015) to 23 (Azadvari 2020)) months, we considered the study populations in all trials to be representative of patients seen in routine care. The synthesis of this review can therefore be applied to similar patients in clinical practice.

Measurement of pain varied across trials from pain at rest or sleep, to pain during overhead activities or during a pain provocation test (e.g. Hawkin's impingement test). We preferred to extract overall pain data when available, but some trials only reported pain with activity and others did not specify the framing of the question by which they assessed pain. Measurement of function also varied across trials. Constant score, our primary measure of function, was assessed in five trials (Azadvari 2020; Bhayana 2018; Dogu 2012; Ucuncu 2009; Zufferey 2012), SPADI in four (Ekeberg 2009; Haghighat 2015; Hsieh 2013; Roddy 2020), ASES in three (Cho 2021; Cole 2016; Lee 2015), and SDQ in two (Dogu 2012; Hsieh 2013). Other measures of function were not used in more than one trial.

Seven trials followed participants up to 12 weeks (Akbari 2020; Azadvari 2020; Bhayana 2018; Cho 2021; Lee 2015; Saeed 2014; Zufferey 2012) and one trial up to 6 months (Lee 2015). However, since the benefits of glucocorticoid injection are immediate and generally only last for six to eight weeks (Buchbinder 2003), longer trials are unnecessary to determine the comparative effectiveness of different forms of administration of glucocorticoid injection.

Six trials (Akbari 2020; Ekeberg 2009; Naredo 2004; Saeed 2014; Ucuncu 2009; Zufferey 2012) reported that there were no serious adverse events.

Four unpublished studies (Cinar 2018; IRCT2017021524621N6; Moore 2018; Pierce 2018) were identified. The authors of two studies indicated that they were in the process of publishing the results (IRCT2017021524621N6; Pierce 2018) and we were not able to contact the authors of Cinar 2018. One trial has been published as a pre‐print in bioRxiv but has not yet been published in a peer‐reviewed journal (Moore 2018). It reported that ultrasound‐guided intra‐articular injection into the glenohumeral joint space did not result in greater pain relief at two weeks or six months compared with landmark‐guided injection in participants with painful shoulder due to osteoarthritis.

It is unclear if availability of the results from the other completed trials will substantially change the conclusions of the review, but it may increase the certainty of the effect estimates, by reducing heterogeneity around the pain and function estimates, and reducing imprecision around estimates for treatment success, quality of life and adverse events.

Quality of the evidence

For the major outcomes, evidence for pain and function was downgraded once to moderate‐certainty due to possible selection, performance and detection biases. We considered downgrading evidence for inconsistency; however, the statistical heterogeneity was driven largely by a single trial that appears to show a larger benefit in favour of image‐guided injection. We downgraded twice to low‐certainty evidence for quality of life and adverse events due to bias and imprecision (the 95% CIs around the effect estimate did not rule in or out a clinically important effect around quality of life, and the number of adverse events was low). Participant‐rated treatment success was downgraded to very low‐certainty due to bias, imprecision, and possible inconsistency (although with only four trials contributing to this outcome, it is harder to interpret the I² value).

The minor outcomes, additional treatments and range of motion were downgraded twice to low‐certainty evidence, due to potential bias, and imprecision (additional treatments) as the 95% CIs around the effect estimate did not rule in or out clinically important effects; and indirectness (range of motion). Evidence for shoulder range of motion was downgraded for indirectness as it is essentially a surrogate for function and presumably a less relevant outcome for participants. Only four studies reported data for range of motion, and it is possible there was reporting bias.

We identified four completed studies with unpublished results, however, as authors of three studies indicated that they planned to publish the results and results for one study is at pre‐publication in a peer‐reviewed journal, we did not downgrade for publication bias. Funnel plots for pain and function at the 6‐week follow‐up time also indicate little evidence of publication bias (Figure 3 and Figure 4). Estimates for pain may have been influenced by the presence of selection bias since the slight/small benefit seen in the main analysis disappeared once we restricted the analysis to studies at low risk of selection bias. Restricting the analysis to studies at low risk of selection bias did not appreciably alter the effect estimate for function and detection bias did not appreciably alter the effect estimates for pain or function.

Low‐certainty evidence shows that ultrasound‐guided injection for shoulder pain probably does not reduce the need for additional treatment, nor reduce the risk of adverse events. Evidence was downgraded for imprecision and risk of performance, detection and attrition bias for these two outcomes. Evidence was not downgraded for indirectness and publication bias for the same reasons listed above.

Potential biases in the review process

We believe that the search likely identified all relevant published studies. A thorough search strategy was devised and all major databases were searched for relevant studies with no language restrictions applied. Two review authors assessed the trials for inclusion in the review and risk of bias, with a third reviewer adjudicating if there was any discrepancy.

Apart from the risk of bias of the included trials, the biggest limitation of the review was that the trials varied in their outcome measures. Measures of both pain and function varied across trials. There were 12 measures of pain (assessed on either 10 cm VAS 100 cm VAS, 0 to 10‐point NRS, or 1‐9 ordinal scale) and 10 measures of function assessed across trials. For 'overall' pain, one of our prespecified primary outcomes, we elected to pool results for pain with activity, daytime pain and unspecified type of pain. For function, we similarly pooled different measures that may not necessarily be measuring the exact same construct.

One trial (Lee 2009) had very small standard deviations. We speculated that the authors had inadvertently published standard errors instead of standard deviations, but this could not be confirmed from the authors.

Another limitation is that we do not know what a meaningful improvement in all outcomes with image‐ versus non‐image‐guided injection for people with shoulder pain might be. Instead of considering the minimal clinically important difference (MCID), we elected to instead describe the magnitude of the mean between‐group differences for pain and function as slight/small, moderate or large. Across all time points, the mean between‐group differences were either slight/small or not different and, in general, the 95% CI also excluded clinically important benefits (large CI were seen when there was substantial statistical heterogeneity).

Agreements and disagreements with other studies or reviews

Another systematic review of image‐guided injection versus non‐image‐guided injection for shoulder pain has been published since the publication of our original review (Aly 2015). Aly 2015 included seven of the 14 trials included in our review (Chen 2006; Dogu 2012; Hsieh 2013; Lee 2009; Naredo 2004; Ucuncu 2009; Zufferey 2012). In contrast to our conclusions, they concluded that patients who received ultrasound‐guided injections had significantly greater improvements in shoulder pain and function at six weeks compared to those who received non‐image‐guided injections. There were, however, several issues with their approach. Notably, the authors did not report the overall quality of the evidence nor acknowledge that the results should be interpreted cautiously in view of the limited number of studies and small sample sizes. The authors' assessment of risk of bias for these trials also differed from our own; they pooled change scores instead of final scores (which appears to have exaggerated their findings for function at six weeks), and they excluded Ekeberg 2009, a high‐quality, adequately powered trial that fulfilled all criteria of low risk of bias in our assessment and failed to demonstrate a difference in efficacy between local ultrasound‐guided injection and intramuscular injection in the upper gluteal buttock region.

Use of ultrasound by clinicians as well as radiologists for both diagnostic and therapeutic purposes for a variety of musculoskeletal indications is expanding rapidly in many settings. There are distinct advantages of ultrasound guidance, for example, when diagnostic aspiration into joints that are deep and difficult to locate using landmark‐guidance are necessary, or for injection of radioisotopes where accuracy is paramount (Iagnocco 2010). There continues to be debate, however, in the literature regarding whether or not accuracy of glucocorticoid injections is a prerequisite for efficacy for the treatment of joint and soft tissue pathology (Hall 2004; Iagnocco 2010), with studies continuing to produce conflicting results (Hartung 2010; Micu 2010; Sibbitt 2009). As outlined by Iagnocco 2010, Naredo 2004 and Ekeberg 2009, the mechanism of local glucocorticoid action is not well understood. Nevertheless, our review provides moderate‐certainty evidence that in patients with shoulder pain, ultrasound‐guided injection does not confer an advantage over landmark or systemic intramuscular injection for improving patient‐relevant outcomes, such as pain and function.

Authors' conclusions

Implications for practice.

Based upon 19 trials, our updated review was unable to establish any important advantage in terms of pain, function, quality of life, treatment success, safety, or shoulder range of motion of ultrasound‐guided glucocorticoid injection for shoulder pain over injection without image guidance. The lack of any added benefit of ultrasound‐guided injection into the subacromial bursa (for rotator cuff disease) or glenohumeral joint (for adhesive capsulitis) suggests that the benefits of glucocorticoids may arise independent of accuracy of needle placement.

The lack of significant benefit of image guidance over injection without image guidance, to improve patient‐relevant outcomes or reduce harms, suggests that the added cost of image guidance to patients and the healthcare system appears unjustified. Patients should be informed about the moderate‐ to low‐certainty evidence for lack of important added benefits of using imaging to guide injections for shoulder pain.

Implications for research.

Our conclusions are based upon moderate‐certainty evidence for pain and function, and low‐certainty evidence for patient‐reported treatment success, quality of life, adverse events, range of motion and requirement for additional treatment. If further comparative effectiveness trials to determine whether or not ultrasound guidance to direct glucocorticoid injection into the putative site of local pathology in the shoulder is superior to non‐image‐guided injection are contemplated, these should consider the sample size required to change our conclusions, especially for adverse events where there is insufficient evidence to be certain about the safety of either approach.

In addition to adequate randomisation and treatment allocation concealment, participant blinding is a necessary requirement. Trials should also measure valid outcomes of importance to patients such as those that have been proposed by the Outcome Measures in Rheumatology (OMERACT) Shoulder Working Group (Ramiro 2019). The core domain set, which has been endorsed by OMERACT, is based upon a systematic review of domains measured in shoulder trials (Page 2018), an international Delphi involving patients, clinicians and researchers (Page 2016c), and a qualitative evidence synthesis of the experiences of people with shoulder disorders (Page 2019). It includes four ‘mandatory’ trial domains: pain, function, patient global ‐ shoulder and adverse events including death; and four ‘important but optional’ domains: participation (recreation/work), sleep, emotional well‐being and condition‐specific pathophysiological manifestations (Ramiro 2019).

What's new

Date	Event	Description
2 September 2021	Amended	Typo in abstract corrected

History

Protocol first published: Issue 5, 2011
Review first published: Issue 8, 2012

Date	Event	Description
18 May 2021	New search has been performed	Search for the review updated 15 February 2021: 14 new studies (873 additional participants) were included (Akbari 2020; Azadvari 2020; Bhayana 2018; Cho 2010; Cho 2021; Cole 2016; Dogu 2012; Haghighat 2015; Hsieh 2013; Lee 2015; Raeissadat 2017; Roddy 2020; Saeed 2014; Zufferey 2012) in addition to the five included trials (n = 290 participants) from the previous version of the review (Chen 2006; Ekeberg 2009; Lee 2009; Naredo 2004; Ucuncu 2009).
18 May 2021	New citation required but conclusions have not changed	New authors were added for this version of the review, methods were updated in accordance with current Cochrane recommendations including presentation of 7 major outcomes in the summary of findings table, and nomination of primary time point (up to 6 weeks follow‐up). Conclusions were similar in this update: Low to moderate‐certainty evidence shows that ultrasound‐guided injection does not provide clinically important benefits in pain, function or quality of life compared with ‘blind’ injections, nor does it increase the risk of adverse events, in people with shoulder pain. Further studies are unlikely to change these conclusions.

Appendices

Appendix 1. CENTRAL search strategy

Database: EBM Reviews ‐ Cochrane Central Register of Controlled Trials <January 2021>

Search Strategy:

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐

1 Shoulder/ (550)

2 Rotator Cuff/ (336)

3 1 or 2 (854)

4 Calcium/ (3451)

5 exp bursitis/ (374)

6 4 or 5 (3825)

7 3 and 6 (25)

8 Shoulder Pain/ (919)

9 Shoulder Impingement Syndrome/ (353)

10 (rotator cuff or supraspinatus or infraspinatus or subscapular$ or teres).tw. (2349)

11 ((shoulder$ or subacromial or rotator cuff) adj5 (tendon$ or tendin$ or bursitis or calcium or calcif$ or impinge$ or tear$ or pain$ or periarthritis or stiff$)).tw. (5585)

12 Shoulder Joint/ (748)

13 (glenohumeral adj5 (joint or capsule)).tw. (253)

14 (adhesive capsuliti$ or frozen shoulder$).tw. (640)

15 or/7‐14 (7132)

16 exp Adrenal Cortex Hormones/ (27688)

17 (adrenal hormone$ or steroid$ or corticosteroid$ or corticoid$ or glucocorticoid$ or sub‐acromial or subacromial or hydroxycorticosteroid$).tw. (47844)

18 (triamcinolone or methylprednisolone or hydrocortisone or predniso$ or cortisone or dexamethasone or betamethasone).tw. (33048)

19 or/16‐18 (78875)

20 exp Injections/ (22337)

21 inject$.tw. (90172)

22 20 or 21 (101579)

23 19 and 22 (10869)

24 15 and 23 (757)

Appendix 2. MEDLINE search strategy

Database: Ovid MEDLINE(R) <1946 to February 15, 2021>

Search Strategy:

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐

1 Shoulder/ (13264)

2 Rotator Cuff/ (6690)

3 1 or 2 (19195)

4 calcium/ (269579)

5 exp bursitis/ (4832)

6 4 or 5 (274400)

7 3 and 6 (776)

8 shoulder pain/ (5002)

9 shoulder impingement syndrome/ (1798)

10 rotator cuff injuries/ (6083)

11 (rotator cuff or supraspinatus or infraspinatus or subscapular$ or teres).tw. (15951)

12 ((shoulder$ or subacromial or rotator cuff) adj5 (tendon$ or tendin$ or bursitis or calcium or calcif$ or impinge$ or tear$ or pain$ or periarthritis or stiff$)).tw. (17038)

13 shoulder joint/ (19714)

14 (glenohumeral adj5 (joint or capsule)).tw. (2488)

15 (adhesive capsuliti$ or frozen shoulder$).tw. (1482)

16 or/7‐15 (40299)

17 Adrenal Cortex Hormones/ (65507)

18 (adrenal hormone$ or steroid$ or corticosteroid$ or corticoid$ or glucocorticoid$ or sub‐acromial or subacromial or hydroxycorticosteroid$).tw. (352115)

19 (triamcinolone or methylprednisolone or hydrocortisone or predniso$ or cortisone or dexamethasone or betamethasone).tw. (146883)

20 or/17‐19 (477533)

21 exp Injections/ (286306)

22 inject$.tw. (674934)

23 21 or 22 (831792)

24 20 and 23 (41175)

25 randomized controlled trial.pt. (522672)

26 controlled clinical trial.pt. (94047)

27 randomized.ab. (435882)

28 placebo.ab. (193776)

29 drug therapy.fs. (2279437)

30 randomly.ab. (295601)

31 trial.ab. (458427)

32 groups.ab. (1824800)

33 or/25‐32 (4471460)

34 exp animals/ not humans.sh. (4788181)

35 33 not 34 (3828179)

36 16 and 24 and 35 (633)

Appendix 3. Embase search strategy

Database: Embase Classic+Embase <1947 to 2021 February 15>

Search Strategy:

‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐‐

1 shoulder/ (41649)

2 rotator cuff/ (6808)

3 1 or 2 (46178)

4 calcium/ (326795)

5 exp bursitis/ (5666)

6 4 or 5 (332414)

7 3 and 6 (951)

8 shoulder impingement syndrome/ (2943)

9 rotator cuff injury/ (2669)

10 rotator cuff rupture/ (7567)

11 (rotator cuff or supraspinatus or infraspinatus or subscapular$).mp. or teres.tw. [mp=title, abstract, heading word, drug trade name, original title, device manufacturer, drug manufacturer, device trade name, keyword, floating subheading word, candidate term word] (28009)

12 ((shoulder$ or subacromial or rotator cuff) adj5 (tendon$ or tendin$ or bursitis or calcium or calcif$ or impinge$ or tear$ or pain or periarthritis or stiff$)).tw. (27117)

13 exp shoulder pain/ (17510)

14 exp frozen shoulder/ (2289)

15 exp humeroscapular periarthritis/ (2365)

16 (glenohumeral adj5 (joint or capsule)).tw. (3612)

17 (adhesive capsuliti$ or frozen shoulder$).tw. (2538)

18 or/7‐17 (54626)

19 corticosteroid/ (267864)

20 corticosteroid therapy/ (51967)

21 (adrenal hormone$ or steroid$ or corticosteroid$ or corticoid$ or glucocorticoid$ or sub‐acromial or subacromial or hydroxycorticosteroid$).tw. (591540)

22 (triamcinolone or methylprednisolone or hydrocortisone or predniso$ or cortisone or dexamethasone or betamethasone).tw. (259780)

23 or/19‐22 (907684)

24 exp injection/ (251615)

25 inject$.tw. (1133111)

26 24 or 25 (1159032)

27 23 and 26 (64060)

28 18 and 27 (1855)

29 random$.tw. (1652459)

30 factorial$.tw. (41088)

31 crossover$.tw. (80422)

32 cross over.tw. (34315)

33 cross‐over.tw. (34315)

34 placebo$.tw. (328887)

35 (doubl$ adj blind$).tw. (224698)

36 (singl$ adj blind$).tw. (26723)

37 assign$.tw. (422103)

38 allocat$.tw. (165987)

39 volunteer$.tw. (273171)

40 crossover procedure/ (66753)

41 double blind procedure/ (185300)

42 randomized controlled trial/ (651777)

43 single blind procedure/ (42092)

44 or/29‐43 (2501102)

45 28 and 44 (592)

Appendix 4. Trial Registry search strategy

WHO International Clinical Trials Search Platform (ICTRP)

Shoulder in condition AND corticosteroid* injection in intervention. Date: 06/07/20 # Hits: 49 trials

ClinicalTrials.gov, using advanced search option:

“Shoulder” in condition AND corticosteroid in intervention. Date: 18/02/21 # Hits: 121 studies

Data and analyses

Comparison 1. Ultrasound‐guided injection versus anatomic landmark‐guided or systemic injection.

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
1.1 Overall pain (or daytime, activity‐related or unspecified)	18		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.1.1 up to 3 weeks	9	554	Mean Difference (IV, Random, 95% CI)	‐0.67 [‐1.31, ‐0.03]
1.1.2 > 3 weeks up to 6 weeks	15	1003	Mean Difference (IV, Random, 95% CI)	‐0.52 [‐0.84, ‐0.20]
1.1.3 > 6 weeks up to 3 months	6	424	Mean Difference (IV, Random, 95% CI)	‐0.07 [‐0.24, 0.09]
1.1.4 > 3 months to 6 months	2	205	Mean Difference (IV, Random, 95% CI)	‐0.61 [‐1.53, 0.31]
1.1.5 > 6 months	2	205	Mean Difference (IV, Random, 95% CI)	0.17 [‐0.73, 1.06]
1.2 Function	17		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.2.1 up to 3 weeks	8	487	Mean Difference (IV, Random, 95% CI)	5.45 [‐0.29, 11.20]
1.2.2 > 3 weeks up to 6 weeks	14	895	Mean Difference (IV, Random, 95% CI)	2.44 [‐0.17, 5.06]
1.2.3 > 6 weeks up to 3 months	4	257	Mean Difference (IV, Random, 95% CI)	9.15 [‐5.78, 24.09]
1.2.4 > 3 months to 6 months	2	205	Mean Difference (IV, Random, 95% CI)	‐1.05 [‐9.30, 7.19]
1.2.5 > 6 months	2	205	Mean Difference (IV, Random, 95% CI)	3.35 [‐4.69, 11.38]
1.3 Treatment success (50% pain improvement on VAS)	5	350	Risk Ratio (M‐H, Random, 95% CI)	1.56 [0.89, 2.75]
1.4 Quality of life	3		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.4.1 up to 3 weeks	2	122	Mean Difference (IV, Random, 95% CI)	4.17 [‐0.97, 9.31]
1.4.2 > 3 weeks up to 6 weeks	2	220	Mean Difference (IV, Random, 95% CI)	2.81 [‐0.73, 6.35]
1.4.3 > 6 weeks to 3 months	1	30	Mean Difference (IV, Random, 95% CI)	4.60 [‐1.27, 10.47]
1.4.4 > 3 months to 6 months	1	128	Mean Difference (IV, Random, 95% CI)	1.60 [‐2.49, 5.69]
1.4.5 > 6 months	1	128	Mean Difference (IV, Random, 95% CI)	‐0.70 [‐4.88, 3.48]
1.5 Number of adverse events	5	402	Risk Ratio (M‐H, Random, 95% CI)	0.72 [0.40, 1.28]
1.6 Withdrawals due to adverse events	1	106	Risk Ratio (M‐H, Random, 95% CI)	0.33 [0.01, 8.00]
1.7 Additional injections or surgery	3		Risk Ratio (M‐H, Random, 95% CI)	Subtotals only
1.7.1 Additional injections	3	229	Risk Ratio (M‐H, Random, 95% CI)	0.56 [0.14, 2.15]
1.7.2 Surgery	1	56	Risk Ratio (M‐H, Random, 95% CI)	0.67 [0.21, 2.11]
1.8 Range of flexion	10		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.8.1 up to 3 weeks	5	305	Mean Difference (IV, Random, 95% CI)	3.45 [‐1.60, 8.49]
1.8.2 > 3 weeks up to 6 weeks	8	461	Mean Difference (IV, Random, 95% CI)	1.09 [‐2.32, 4.49]
1.8.3 > 6 weeks up to 3 months	2	167	Mean Difference (IV, Random, 95% CI)	3.70 [‐1.42, 8.82]
1.8.4 > 3 months to 6 months	1	77	Mean Difference (IV, Random, 95% CI)	0.50 [‐5.49, 6.49]
1.8.5 > 6 months	1	77	Mean Difference (IV, Random, 95% CI)	4.00 [‐4.53, 12.53]
1.9 Range of abduction	10		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.9.1 up to 3 weeks	5	317	Mean Difference (IV, Random, 95% CI)	12.57 [2.62, 22.53]
1.9.2 > 3 weeks up to 6 weeks	9	528	Mean Difference (IV, Random, 95% CI)	6.85 [1.47, 12.22]
1.9.3 > 6 weeks to 3 months	1	90	Mean Difference (IV, Random, 95% CI)	1.10 [‐8.46, 10.66]
1.10 Range of external rotation	8		Mean Difference (IV, Random, 95% CI)	Subtotals only
1.10.1 up to 3 weeks	4	199	Mean Difference (IV, Random, 95% CI)	1.01 [‐3.88, 5.90]
1.10.2 > 3 weeks up to 6 weeks	6	334	Mean Difference (IV, Random, 95% CI)	‐0.73 [‐3.50, 2.04]
1.10.3 > 6 weeks up to 3 months	2	167	Mean Difference (IV, Random, 95% CI)	3.87 [‐0.42, 8.17]
1.10.4 > 3 months to 6 months	1	77	Mean Difference (IV, Random, 95% CI)	3.20 [‐8.10, 14.50]
1.10.5 > 6 months	1	77	Mean Difference (IV, Random, 95% CI)	‐2.70 [‐13.33, 7.93]
1.11 Serious adverse events	1	128	Risk Ratio (M‐H, Fixed, 95% CI)	3.00 [0.12, 72.29]

1.11. Analysis.

Comparison 1: Ultrasound‐guided injection versus anatomic landmark‐guided or systemic injection, Outcome 11: Serious adverse events

Comparison 2. Subgroup analyses (stratified by shoulder condition).

Outcome or subgroup title	No. of studies	No. of participants	Statistical method	Effect size
2.1 Overall pain	16		Mean Difference (IV, Random, 95% CI)	Subtotals only
2.1.1 Subacromial disease	12	777	Mean Difference (IV, Random, 95% CI)	‐0.58 [‐1.05, ‐0.11]
2.1.2 Adhesive capsulitis	3	189	Mean Difference (IV, Random, 95% CI)	‐0.21 [‐0.39, ‐0.03]
2.1.3 Mixed or undefined shoulder pain	1	65	Mean Difference (IV, Random, 95% CI)	‐1.20 [‐2.46, 0.06]
2.2 Function	15		Mean Difference (IV, Random, 95% CI)	Subtotals only
2.2.1 Subacromial disease	11	687	Mean Difference (IV, Random, 95% CI)	5.06 [‐3.23, 13.35]
2.2.2 Adhesive capsulitis	3	171	Mean Difference (IV, Random, 95% CI)	1.05 [‐1.18, 3.29]
2.2.3 Mixed or undefined shoulder pain	1	65	Mean Difference (IV, Random, 95% CI)	4.00 [0.05, 7.95]

Characteristics of studies

Characteristics of included studies [ordered by study ID]

Akbari 2020.

Study characteristics
Methods	Design: Single‐centre two‐arm parallel‐group randomised controlled trial Setting: Baskent University Faculty of Medicine, Physical Medicine and Rehabilitation (PMR) Department Outpatient clinic,Turkey Timing: 01 February 2017 and 31 May 2017 Interventions: US‐guided subacromial glucocorticoid and local anaesthetic injection vs anatomic landmark‐guided subacromial glucocorticoid and local anaesthetic injection Sample size calculation: The power analysis was performed based on the mean and standard deviation of the Constant score variables as previously used by Ucuncu 2009. The Pass 11 NCSS (LLC. Kaysville, Utah, USA) statistical program was used. According to the power analysis calculation, in order to obtain a power of 95% with a 5% type I error, each group would require at least 14 participants. Analysis: completers' analysis
Participants	Number of participants: Number of participants screened for eligibility: 42 Number enrolled: 29 (excluded; dialysis = 3; cervical vertebra pathology: 1; surgical intervention planned: 3; non‐consent to study participation: 7) Number randomised: 29 (14 to the US‐guided injection group and 15 to the blind injection group) Number included in analyses at each follow‐up: 28 (14 in the US‐guided injection group and 14 in the blind injection group) Inclusion criteria: (i) a history of posterolateral shoulder pain for more than three months which increased on shoulder abduction; (ii) painful restriction of active flexion and/or abduction of the shoulder with more restriction on passive ROM; (iii) a positive Hawkins‐Kennedy impingement sign; (iv) magnetic resonance imaging (MRI) consistent with SIS (rotator cuff impingement). Exclusion criteria: (i) a history of inflammatory arthritis; (ii) erythema/swelling of the shoulder joint; (iii) neurological deficit of the upper extremities; (iv) shoulder dislocation; (v) presence of partial/full thickness rotator cuff tear, bursitis, calcific tendinitis, or labral tears on MRI; (vi) significant chronic disease; (vii) a history of or current malignancy; (viii) shoulder trauma occurring within the past three months; (ix) SIS treatment within the past three months; (x) physical therapy of the ipsilateral shoulder within the past six months; (xi) non‐consent to subacromial injection; (xii) cases in which surgical intervention was deemed to be the appropriate treatment of choice; (xiii) cervical disc/suspicion of a cervical disc pathology. Baseline characteristics: US‐guided subacromial injection group (n = 14): Median (range) age in years: 39.5 (20‐64) No.(%) female: 8 (57.1) Median (range) VAS pain: 9 (4‐10) Median (range) Active shoulder flexion ROM in degrees: 150 (80‐170) Median (range) Active shoulder abduction ROM in degrees: 150 (80‐170) Median (range) DASH: 110 (45‐145) Median (range) Constant‐Murley score: 39 (19‐73) Anatomical landmark‐guided subacromial injection group (n = 14): Median (range) age in years: 42.5 (20‐64) No.(%) female: 9 (64.3) Median (range) VAS pain: 8.5 (4‐10) Median (range) Active shoulder flexion ROM in degrees: 140 (70‐165) Median (range) Active shoulder abduction ROM in degrees: 130 (70‐160) Median (range) DASH: 117 (62‐141) Median (range) Constant‐Murley score: 37 (13‐80) Pretreatment group differences: None apparent
Interventions	Ultrasound‐guided subacromial glucocorticoid injections: All US‐guided subacromial glucocorticoid injections were performed by a radiologist with 10 years of experience in musculoskeletal US working in the Radiology Department of Baskent University Faculty of Medicine. US imaging of the shoulder was obtained using the 2014 model Siemens Acuson S2000 (Siemens Healthcare, Erlangen, Germany) and a 9 MHz linear probe. The patient was seated and the shoulder was internally rotated with the ipsilateral hand positioned on the hip in the modified Crass position. Methylprednisolone acetate 40 mg in 1 mL and procaine 2% 4 mL were prepared in a 5 mL syringe. The anterolateral aspect of the shoulder was cleaned using 10% povidone iodine solution. The US probe was placed on the anterolateral aspect of the shoulder and the subacromial bursa was visualised. A 21‐gauge needle was used to enter the anteromedial aspect of the shoulder under continuous US guidance. Once the bevel of the needle was visualised in the subacromial bursa, the solution was injected. Anatomical landmark‐guided subacromial glucocorticoid injections: The subacromial injection was performed in the PMR Outpatient Department of Baskent University Faculty of Medicine by a single physiatrist with more than 10 years of experience in the field. The injection was performed using a standard posterolateral approach and an aseptic technique. The patient was seated upright with the arms resting comfortably at the side. The distal, lateral, and posterior edges of the acromion were palpated and the needle was inserted just inferiorly to the posterolateral edge of the acromion and directed towards the opposite nipple. Aspiration was performed to ensure that the needle was not in a blood vessel prior to administration of the drug. Methylprednisolone acetate 40 mg in 1 mL and procaine 2% 4 mL were injected slowly using a 21‐gauge needle 1 cm into the subacromial space.
Outcomes	Outcomes were measured at 4 weeks. Primary outcome: Visual analog scale (VAS) for shoulder pain from 0‐10 cm marked in 1 cm increments where 0 cm represents no pain Secondary outcomes: Active shoulder ROM in flexion and abduction using a goniometer Function using Disabilities of the Arm, Shoulder, and Hand (DASH) questionnaire score ranging from 0 (no disability) to 100 (most severe disability). In this study, the Turkish version of the DASH was used. Modified Constant‐Murley Score (CMS) score. A total score out of 100 is obtained; the higher the score the better the patient’s functionality. In this study, the Turkish version of the modified CMS was used. Outcomes used in this review: Pain on VAS Function on DASH ROM flexion and abduction scores
Sources of funding	This study was supported by the Baskent University Research Fund.
Notes	Trial registration: Not reported Withdrawals: 0/14 in US‐guided injection group; 1/15 in blind injection group lost to follow‐up Adverse events: No adverse events reported in either group at 4‐week follow‐up Data analysis: For pain, ROM and function, we used the median in place of the mean and used the SD from Ekeberg 2009.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Participants were randomised into two groups using the Random Allocation Software version 1.0.
Allocation concealment (selection bias)	Low risk	Participants were allocated by an independent junior doctor of the PMR department who had no other involvement in the study.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and study personnel were not blinded. Procedures were performed in different sites (radiology department for ultrasound‐guided injection or outpatient department for the non‐guided injection).
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	High risk	Self‐reported outcomes could have been subject to detection bias as participants were aware of their treatment allocation.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	"All patients were screened and assessed for study participation by a single PMR specialist prior to the procedure and at four weeks post‐injection. This PMR specialist was blinded to the method of injection applied and had no involvement in the injection procedure".
Incomplete outcome data (attrition bias) All outcomes	Low risk	0/14 in US‐guided injection group; 1/15 in blind injection group was lost to follow‐up.
Selective reporting (reporting bias)	Unclear risk	Trial was not registered and no study protocol was found.
Other bias	Low risk	None apparent

Azadvari 2020.

Study characteristics
Methods	Design: Single‐centre single‐blinded two‐arm parallel‐group randomised controlled trial Setting: Physical medicine clinic of Tehran University of Medical Sciences, Iran Timing: 2017 to 2019 Interventions: Ultrasound‐guided subacromial glucocorticoid injection vs anatomical landmark‐guided subacromial glucocorticoid injection Sample size calculation: not reported Analysis: not reported
Participants	Number of participants: Number of participants screened for eligibility: not reported Number enrolled: 30 Number randomised: 30 (15 to the US‐guided injection group and 15 to the blind injection group) Number included in analyses at each follow‐up: 30 (15 in the US‐guided injection group and 15 in the blind injection group) Inclusion criteria: Subacromial impingement in paraplegic patients with spinal cord injury (below T6) Aged between 18 and 50 years with complaint of shoulder pain and VAS score higher than 4 whose Neer test and Hawkins‐ Kennedy test results were positive Exclusion criteria: Rheumatoid arthritis Glenohumeral osteoarthritis Evident clinical rupture in rotator cuff Fractured or dislocated shoulder Diabetes mellitus Infection and tumour Receiving subacromial injection in last eight weeks Pregnancy or decision for that Risk factors for intra‐joint injection (i.e. blood coagulant disorders or application of anticoagulant drugs) and steroid/lidocaine injection contraindications (allergy to these drugs) Baseline characteristics: US‐guided injection group (n = 15): Mean (SD) age in years: 33.8 (9.34) Mean (SD) duration of shoulder pain in months: 23.46 (14.78) Mean (SD) duration of spinal cord injury: 8.86 (4.06) Blind injection group (n = 15): Mean (SD) age in years: 28.13 (11.02) Mean (SD) duration of shoulder pain in months: 22.53 (14.87) Mean (SD) duration of spinal cord injury: 9.46 (4.86) Pretreatment group differences: none apparent
Interventions	Ultrasound‐guided subacromial glucocorticoid injection: 1 cc depo‐medrol (methylprednisolone, Merck, Germany) was injected along with 2 cc lidocaine under ultrasound guide within the subacromial space. The injection was performed with lateral approach. The patients were at sitting position, they put their hand on iliac crest by internal rotation of shoulder joint. The probe was placed on supraspinatus muscle in longitudinal plane. Acromion was found by ultrasound guidance. Then the probe was derived to the inferior part so the subacromial‐subdeltoid bursa was determined. A hypodermic needle (gauge 22) was entered beneath the probe in in‐plane manner in upward medial direction until the needle tip entered the bursa and its dilatation was observed during injection. Injection in subacromial space was performed by a physiatrist. Anatomic landmark‐guided subacromial glucocorticoid injection: Participants received lateral approached blind glucocorticoid injection guided by anatomic landmarks under sterile conditions. The injection site was marked and then sterilised. Lateral edge of the acromion was touched by hand and the needle with gauge of 22–23 was guided 2–3 cm beneath the lateral edge of the acromion toward the medial side (slightly superior). Under resistance‐free condition, 1 cc depo‐medrol (methylprednisolone, Merck, Germany) was injected along with 2 cc lidocaine. Post‐intervention: Exercise therapy was used in both groups after seven days. The exercises included range of motion, posterior capsule stretching and isometric strengthening exercises presented as an instruction.
Outcomes	Outcomes were measured at 2 weeks and 2 months. Study outcomes: Pain on VAS Function on Constant score Quality of life on BREF questionnaire (assume: World Health Organization WHOQOL‐BREF quality of life assessment) Outcomes used in this review: Pain on VAS Function on Constant score Quality of life on BREF questionnaire (total scores)
Sources of funding	None reported
Notes	Trial registration: Not reported Withdrawals: None in either group Adverse events: Not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"The patients were first divided into two groups by 4‐block randomisation method".
Allocation concealment (selection bias)	Low risk	These two words were then written on paper in 6 different configurations (AABB, ABAB, BBAA, BAAB, ABBA, BABA) and the papers were sealed. One of the papers was then randomly selected and the patients were allocated to each group according to the order of A and B in the paper.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Study was reported to be single‐blind but it appeared that participants and study personnel were not blinded as procedures were described to have been performed either with or without ultrasound guidance without any attempt to blind participants to treatment allocation.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	High risk	Self‐reported outcomes could have been subject to detection bias as participants were aware of their treatment allocation.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	No assessor‐reported outcomes
Incomplete outcome data (attrition bias) All outcomes	Low risk	No withdrawals from either group
Selective reporting (reporting bias)	Unclear risk	Trial was not registered.
Other bias	Low risk	None apparent

Bhayana 2018.

Study characteristics
Methods	Design: Single‐centre, parallel‐group, two‐arm randomised controlled trial in participants with rotator cuff syndrome Setting: Orthopaedic outpatient department in Guru Teg Bahadur Hospital, Delhi Timing: October 2012–December 2013 Interventions: Ultrasound‐guided injection of 2 mL (40 mg/mL) methylprednisolone and 2 mL of 1% lignocaine into the subacromial bursa or anatomical landmark‐guided injection of 2 mL (40 mg/mL) methylprednisolone and 2 mL of 1% lignocaine into the subacromial space Sample size: Not reported Analysis: Not reported
Participants	Number of participants: Number of participants screened for eligibility: not reported Number enrolled: 60 (30 per group) Number randomised: 60 (30 per group) Number included in analyses at each follow‐up: 60 (30 in the US‐guided group and 30 in the blind group) Inclusion criteria: 1. At least 18 years old 2. History of shoulder pain for at least two months that has been not satisfactorily responsive to a trial of oral medication and physical therapy for at least last one month 3. History of night pain 4. Pain on overhead abduction 5. Less than 50% reduction in glenohumeral range of motion in not more than one direction of external rotation, internal rotation or abduction 6. Positive impingement test Exclusion criteria: 1. History of inflammatory joint disease 2. Significant trauma 3. Periarticular arthritis 4. Allergy to contrast agents 5. Previous history of steroid injection in the same shoulder 6. Fracture, glenohumeral osteoarthritis, osteonecrosis or other bone conditions 7. Symptomatic acromioclavicular arthritis 8. Full thickness rotator cuff tear 9. Pregnancy Baseline characteristics: Ultrasound‐guided injection group Mean (SD) age: 44.53 (9.2) years 17 female; 13 male Current smoker: 18 (60%) Heavy occupational demands: 22 (73.3%) Mean (SD) VAS for pain: 52.93 (1.9) Mean (SD) Constant score: 63.9 (7.8) Ultrasound diagnosis: supraspinatus tendinitis 22; supraspinatus bursitis 13; infraspinatus bursitis 4; infraspinatus partial tear 3; subscapularis tendinitis 3; subscapularis partial tear 2; biceps tendinitis 2; subacromial/subdeltoid bursitis 2; acromioclavicular involvement 0; impingement 30; biceps tendon subluxation; biceps tendon rupture 0; glenohumeral joint effusion 0; teres minor partial tear 0 Anatomical‐guided injection group Mean (SD) age: 42.03 (9.9) years 10 female; 20 male Current smoker: 20 (66.6%) Heavy occupational demands: 16 (53.3%) Mean (SD) VAS for pain: 48.13 (11.5) Mean (SD) Constant score: 58.90 (9.2) Ultrasound diagnosis: supraspinatus tendinitis 18; supraspinatus bursitis 25; infraspinatus bursitis 1; infraspinatus partial tear 1; subscapularis tendinitis 2; subscapularis partial tear 1; biceps tendinitis 3; subacromial/subdeltoid bursitis 5; acromioclavicular involvement 0; impingement 30; biceps tendon subluxation; biceps tendon rupture 0; glenohumeral joint effusion 0; teres minor partial tear 0
Interventions	All procedures were performed under strict sterile and aseptic conditions. All participants received an injection of 2 mL (40 mg/mL) methylprednisolone and 2 mL of 1% lignocaine combination. Ultrasound‐guided injection (N = 30): A single experienced radiologist performed the ultrasound imaging using the Phillips HD 7 US machine with 7–10 MHz multi frequency broad band transducer. Intradermal drug testing with lignocaine was performed 30 minutes prior to the procedure. The patient was seated with the affected arm in hyperextension and internal rotation with the elbow bent and back of the hand resting against the lower back. A 10 mL syringe connected to a 5 cm, 21 gauge needle was prepared with 2 mL of 40 mg/mL methylprednisolone acetate suspension mixed and 2 mL of 1% lignocaine and was inserted parallel to the transducer in a semi‐oblique plane from the anterior side of the shoulder. The bevelled side facing the transducer, the needle was advanced under the real time ultrasound assistance till the tip of the needle appeared in the subacromial bursa. Rotation of the bevelled side of the needle by 180 degrees was done for confirmation of the tip positioning and the bulging of subacromial bursa ascertained in the real time imaging. Anatomical landmark‐guided injection (N = 30): A single orthopaedic surgeon performed all the landmark‐guided injections. A 10 mL syringe connected to a 5 cm, 21 gauge needle was prepared with 2 mL of 40 mg/mL methylprednisolone acetate suspension mixed, 2 mL lignocaine and 2 mL of radio opaque non‐ionic contrast media iohexol. Intradermal skin testing with lignocaine and iohexol was performed 30 minutes prior to the procedure. Access to the subacromial space was achieved using a lateral approach. The needle was inserted just inferior to the midlateral aspect of the acromion, with the needle angled slightly cephalad, passing through the deltoid muscle, and directed medially and slightly anterior to the subacromial bursa. Care was taken to avoid injection directly into the tendons of the rotator cuff. Correct placement of the drug in the subacromial subdeltoid space was confirmed by the fluoroscopic assessment under the C arm within 10 min of the injection. Accuracy of correct instillation of methylprednisolone‐lignocaine‐iohexol combination was assessed by a senior radiologist. Co‐interventions: All the patients were prescribed a 5‐day course of antibiotics for prevention of injection‐induced subacromial bursitis. All participants had explained to them a home‐based shoulder rehabilitation programme, consisting of shoulder abduction and pendulum exercises (no further details reported). Patients were not prohibited from the use of analgesics or using the affected shoulder during the follow‐up period.
Outcomes	Outcomes were measured at baseline, and at 5 days, 3 weeks, 6 weeks and 3 months post‐intervention. Outcomes: 1. Mean visual analog scale (VAS) for pain (0 to 100; 0 no pain, 100 maximum pain) 2. Mean Constant score (0 to 100, higher score indicates better function. Domains: pain 15 points; activities of daily living 20 points, range of motion 40 points; strength 25 points) 3. Adverse events No primary outcome was specified. Outcomes included in this review 1. Mean visual analog scale (VAS) for pain 2. Mean Constant score 3. Adverse events
Sources of funding	No funding received
Notes	Conflicts of interest: Authors declared no conflicts of interest. Trial registration: Not reported Timepoints included in this review: 5 days, 3 weeks, 6 weeks and 3 months Data analysis: Range of motion data were requested from the authors but we did not receive it. Withdrawals: None Adverse events: None in either group
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	'Random number sequence' used to allocate participants, but method of sequence generation not reported.
Allocation concealment (selection bias)	Unclear risk	Allocation concealment was not reported.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Procedures were described to have been performed either with or without ultrasound guidance without any attempt to blind participants to treatment allocation. Procedures were performed by different specialists (radiologist for ultrasound‐guided injection and orthopaedic surgeon for the non‐guided injection). Only the outcome assessor was blinded to treatment allocation.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	High risk	Self‐reported outcomes could have been subject to detection bias as participants were aware of their treatment allocation.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	The assessor of range of motion was blinded to treatment allocation, thus there was a low risk of bias.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All randomised participants appeared to complete follow‐up assessments.
Selective reporting (reporting bias)	Unclear risk	There was no protocol paper and the trial was not registered. Range of motion was measured but not reported.
Other bias	Low risk	None apparent

Chen 2006.

Study characteristics
Methods	Design: Parallel‐group, two‐arm controlled trial in participants with subacromial bursitis (unclear if randomised) Setting: Taiwan Timing: Not reported Interventions: Ultrasound‐guided or anatomical landmark‐guided injection into the subacromial bursa Sample size: Not reported Analysis: Intention‐to‐treat analysis not specified
Participants	Number of participants: Number of participants screened for eligibility: not reported Number enrolled: 40 (20 per group) Number randomised: 40 participants were divided into 2 groups (20 per group) Number included in analyses at each follow‐up: not reported Inclusion criteria: 1. Shoulder pain for > 1 month 2. Shoulder pain that could be elicited during abduction manoeuvres 3. Shoulder range of motion limitation and the existence of painful arc syndrome 4. Sonographic confirmation of subacromial bursitis Exclusion criteria: 1. Capsular lesions of the shoulder Baseline characteristics: Baseline characteristics by group were not reported. Mean (range) age: 53 (30‐66) years Ratio of men:women 2:1. Exact numbers not reported Range of duration of symptoms: 2‐10 months Ultrasound‐guided group Mean (SD) abduction ROM: 69.05 (14.72) Anatomical landmark‐guided injection group Mean (SD) abduction ROM: 71.03 (12.38)
Interventions	All participants had an ultrasound performed prior to the procedure. They were seated in an upright position with their back well supported and arms behind their backs with their elbows bent. Both groups received a single injection of a suspension containing 1 mL of betamethasone and 1 mL of 1% lidocaine using a 21‐gauge needle. Ultrasound‐guided injection (N = 20): The needle was inserted into the subacromial bursa under ultrasound guidance using the LOGIQ 9 with 10L probe (4‐10 MHz) (General Electronic Company, Milwaukee, WI). Aspiration of the effusion was done first before injecting the steroid‐lidocaine suspension into the subacromial bursa. Advancement of the needle to the lesion site was observed as continuous and real‐time images. The ultrasound‐guided injections were performed by a physician experienced in using ultrasound probes and in reading musculoskeletal sonograms. Anatomical landmark‐guided injection (N = 20): The acromion was palpated by the thumb and the needle inserted in a horizontal approach. The needle was adjusted in different depths and angles to try to aspirate the effusion. Exaggerated needle adjusting motions were prevented to avoid pain and discomfort that may be experienced by the patients. If no effusion could be aspirated, the physician injected the suspension into the subacromial bursa. The blind injection was performed by a physician experienced with peripheral joint and soft‐tissue injections.  Co‐interventions: Not reported
Outcomes	Outcomes were measured before and 1 week after the injection. Outcomes 1. Mean shoulder abduction ROM using a goniometer (not specified if active or passive) No other outcomes were specified. Outcomes included in this review 1. Mean shoulder abduction ROM
Sources of funding	No funding reported
Notes	Conflicts of interest: No conflicts of interest statement Trial registration: Not reported Time points included in this review: 1 week Withdrawals: Not reported Adverse events: Not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Quote: "the patients were divided into blind and ultrasound‐guided injection groups". Method not reported, but unlikely this was randomised
Allocation concealment (selection bias)	High risk	Unlikely that the sequence allocation was concealed
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and study personnel were not blinded as procedures were described to have been performed either with or without ultrasound guidance without any attempt to blind participants to treatment allocation.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	Low risk	No participant‐reported outcomes were assessed.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Unclear risk	Not reported if assessor of range of motion was blinded
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	The number of participants who completed the one‐week assessment was not reported.
Selective reporting (reporting bias)	Unclear risk	The trial was not registered, no study protocol was found and it was unclear if any other range of motion outcomes were measured.
Other bias	Low risk	None apparent

Cho 2010.

Study characteristics
Methods	Design: Single‐centre, parallel‐group, two‐arm, randomised controlled trial in participants with subacromial or subdeltoid bursitis Setting: Korea. No further details specified Timing: Not specified Interventions: Ultrasound‐guided or anatomical landmark‐guided glucocorticoid injection into the subacromial subdeltoid bursa Sample size: Not reported Analysis: Intention‐to‐treat analysis not specified
Participants	Number of participants: Number of participants screened for eligibility: not reported Number enrolled: 28 (14 per group) Number randomised: 28 (14 per group) Number included in analyses at each follow‐up: not reported Inclusion criteria: 1. Shoulder pain for longer than 1 month 2. Positive findings on Hawkin's test (i.e. reproduction of pain) 3. Thickening of the bursal wall or effusion on ultrasound examination Exclusion criteria: 1. History of shoulder surgery 2. Previously received glucocorticoids Baseline characteristics: Ultrasound‐guided group Mean (SD) age: 60.3 (5.2) 8 female; 6 male Mean (SD) VAS for pain during Hawkins impingement test: 7.1 (1.36) Mean (SD) flexion ROM: 105.5 (11.31) Mean (SD) external rotation ROM: 36.6 (1.84) Mean (SD) UCLA Shoulder Rating Scale score: 3.3 (0.74) Anatomical landmark‐guided group Mean (SD) age: 60.7 (6.5) 7 female; 7 male Mean (SD) VAS for pain during Hawkins impingement test: 7.2 (1.05) Mean (SD) flexion ROM: 101.5 (14.12) Mean (SD) external rotation ROM: 37.12 (8.86) Mean (SD) UCLA Shoulder Rating Scale score: 4.5 (1.34)
Interventions	All participants received an injection of 5 mL of 0.5% lidocaine and triamcinolone acetate. All injections were performed by the same researcher. The study did not specify the training of this researcher. In both groups, successful injection was confirmed by the ultrasonographic examination after the injection. Ultrasound‐guided injection (N = 14): Ultrasound‐guided injections were performed using a Philips Newand 13 MHz linear transducer via a posterior approach. After the patient was seated in a chair, the deformed shoulder was extended, the elbow joint was flexed, and the modified crass position was adopted. The transducer was placed in front of the acromion, the needle was inserted in parallel with the probe to position the tip to the bursa and real‐time ultrasound was used to confirm that the tip of the needle was in the bursa and solution was then injected into the bursa. Anatomical landmark‐guided injection (N = 14): A posterior approach was used. Co‐interventions: Not reported
Outcomes	Outcomes were measured before the injection and 1 week after the injection. Outcomes: 1. Mean pain on a Visual Analog Scale (VAS) during Hawkin's impingement test (0 to 10; 0 no pain, 10 maximum pain) 2. Mean active shoulder ROM in flexion and external rotation 3. Mean UCLA shoulder rating scale score (2 to 35, higher score indicates better function) (Domains: Pain 9 points; function 9 points; active forward flexion 5 points; strength of forward flexion (manual muscle test) 5 points; satisfaction of patient 5 points) 4. Ultrasound examination of accuracy of injection Outcomes included in this review 1. Mean pain on a Visual Analog Scale (VAS) during Hawkin's impingement test 2. Mean UCLA shoulder rating scale score 3. Mean active shoulder ROM in flexion and external rotation
Sources of funding	Not reported
Notes	Conflicts of interest: No conflicts of interest statement Trial registration: Not reported Time points included in this review: 1 week Data analysis: For function, SD from the US group was used in the landmark‐guided group in the 1‐week data. This manuscript was translated from Korean to English. Withdrawals: None Adverse events: Ultrasound was performed 1 week post‐injection in both groups to look for side effects, however none were reported in either group.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Randomisation was performed by drawing lots.
Allocation concealment (selection bias)	Unclear risk	The study did not report if or how sequence allocation was concealed.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and study personnel were not blinded as procedures were described to have been performed either with or without ultrasound guidance without any attempt to blind participants to treatment allocation.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	High risk	It is likely participants were aware of their treatment allocation.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Unclear risk	The authors did not report whether the outcome assessor was blinded or not.
Incomplete outcome data (attrition bias) All outcomes	Low risk	No withdrawals in either group
Selective reporting (reporting bias)	Unclear risk	The trial was not registered and no study protocol was found.
Other bias	Low risk	None apparent

Cho 2021.

Study characteristics
Methods	Design: Single‐centre two‐arm parallel randomised controlled trial Setting: Keimyung University Dongsan Medical Center, Korea Timing: 1 November 2017 and 31 November 2018 Interventions: US‐guided glucocorticoid injection vs blind glucocorticoid injection Sample size: A power analysis indicated that sample sizes of 36 patients in each group would be required to show significant differences in ASES scores with a mean difference of eight points and a standard deviation of 12 points at an α level of 0.05 and a β value of 0.20. Considering a potential dropout rate of 20%, a minimum of 45 patients per group were enrolled. Analysis: Intention‐to‐treat
Participants	Number of participants: Number of participants screened for eligibility: 113 Number enrolled: 90 (20 not meeting inclusion criteria; 3 declined to participate) Number randomised: 90 (45 to US‐guided injection group and 45 to blind injection group) Number included in analyses at each follow‐up: 90 (45 in the US‐guided injection group and 45 in blind injection group) Inclusion criteria: Primary frozen shoulder patients with age ≥ 20 years Shoulder pain with limitation of passive motion of greater than 30 degrees in two or more planes of movement Exclusion criteria: Secondary frozen shoulder, such as that due to a rotator cuff tear Calcific tendinitis and rheumatic diseases Infection A history of high‐energy trauma Previous shoulder surgery Osteoarthritis A glucocorticoid injection on the affected side within the previous three months Poor cognitive function Bleeding diathesis Baseline characteristics: US‐guided injection group (n = 45) Mean (SD) age in years: 57.1 (8.2) No. female: 26 Mean (SD) duration of symptoms: 8.2 (7.4) Mean (SD) VAS: 7.2 (1.9) Mean (SD) ASES: 35.3 (16.7) Mean (SD) SSV: 39.1 (16.9) Mean (SD) ROM flexion: 118.3 (22.5) Mean (SD) ROM abduction: 100.7 (23.4) Mean (SD) ROM external rotation: 43.8 (17.8) Mean (SD) ROM internal rotation: 17.3 (2.8) Blind injection group (n = 45): Mean (SD) age in years: 58.2 (11.1) No. female: 30 Mean (SD) duration of symptoms: 7.1 (6.9) Mean (SD) VAS: 6.7 (1.9) Mean (SD) ASES: 36.7 (14.3) Mean (SD) SSV: 37.8 (17.8) Mean (SD) ROM flexion: 112.3 (23.1) Mean (SD) ROM abduction: 92.9 (20.4) Mean (SD) ROM external rotation: 37.6 (21.3) Mean (SD) ROM internal rotation: 17.6 (2.9) Pretreatment group differences: None apparent
Interventions	The injection mixture for both groups consisted of 40 mg of triamcinolone acetonide, 4 mL of 1% lidocaine, 4 mL of normal saline, and 3 mL of water soluble unionised contrast. A posterior approach was used for the injection in both groups, with the patient in the semi‐lateral decubitus position on the unaffected side and 45° anterior tilting of affected side. All procedures were performed by a single specialist with 15 years of experience in the field. US group (glenohumeral injection under US guidance): For the US group, the needle was advanced laterally to medially with visualisation of its shaft using a linear 5‐ to 12‐MHz probe (HD15 ultrasound system; Philips, Bothell, Washington, USA), and reaching the glenohumeral joint space between the posterior aspect of the humeral head and the glenoid labrum. Blind group (glenohumeral injection without guidance): For the blind group, the needle was inserted 2 cm inferior to the posterolateral margin of the acromion and directed anteriorly towards the coracoid process. After the needle contacted the humeral head, it was slightly withdrawn and the injection was administered. For the blind group, the US probe was placed just under the acromion without visualisation, to blind the patient to their allocation group. Post‐intervention: All patients were instructed to use a home‐based exercise programme to increase ROM and were allowed to perform an exercise three times a day (15 minutes each round). The programme included pendulum exercises, wall‐climbing exercises, and gentle ROM exercises with a bar. During this programme, patients were asked to refrain from provoking post‐mobilisation soreness with self‐feedback. Patients were forbidden from having acupuncture or receiving additional injections from other hospitals.
Outcomes	Outcomes were measured at 3, 6 and 12 weeks. Study outcomes: Pain measured on VAS American Shoulder and Elbow Surgeons (ASES) score Subjective shoulder value (SSV) Passive ROMs: The ROMs, including forward flexion, abduction, external rotation, and internal rotation were assessed using a goniometer with the patient in the sitting position. Outcomes used in this review: VAS pain ASES ROM scores
Sources of funding	Support was received from the National Research Foundation of Korea, funded by the Korean government.
Notes	Trial registration: Clinical Research Information Service (KCT0003105) Conflict of interest: C‐H. Cho and D. H. Kim reported grants from the National Research Foundation of Korea, which related to this article. Withdrawals: 4/45 were lost to follow‐up in the US‐guided injection group and 6/45 were lost to follow‐up in the blind injection group. Adverse events: None reported in either group
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	"Computer‐generated blocked‐randomisation number"
Allocation concealment (selection bias)	Low risk	Allocation was concealed in a sealed envelope.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants were blinded to group allocation. "For the blind group, the US probe was placed just under the acromion without visualization, to blind the patient to their allocation group." The outcome assessor was blinded to treatment allocation. The single interventionalist clearly would have known the treatment allocation but did not participate further and did not check the accuracy of injection.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	Low risk	As participants were blinded to treatment group, there was low risk of bias in the measurement of self‐reported outcomes.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	"The evaluator was blind to the study assignments." Low risk of bias in the measurement of range of motion
Incomplete outcome data (attrition bias) All outcomes	Low risk	4/45 in the US‐guided injection group and 6/45 in the blind injection group were lost to follow‐up.
Selective reporting (reporting bias)	Low risk	Clinical Research Information Service (KCT0003105): all prespecified outcomes were collected and reported in the study.
Other bias	Low risk	None apparent

Cole 2016.

Study characteristics
Methods	Design: Single‐centre, parallel‐group, two‐arm, double‐blind randomised controlled trial in participants with subacromial impingement syndrome Setting: Orthopaedic Research Institute, St George Hospital, Sydney, Australia Timing: Not reported Interventions: Ultrasound‐guided or anatomical landmark‐guided glucocorticoid injection into the subacromial bursa Sample size: 38 patients (19 per group) required to detect at least a 2 cm (SD 2.13) advantage of ultrasound‐guided injections over anatomic landmark‐guided injections in terms of pain (0‐10 cm VAS scale), based on a significance level of 0.05 and power 80% To account for a 10% rate of loss at follow‐up and still meet these requirements, 51 patients (56 shoulders) were included in the study. Analysis: Analyses performed as per intention‐to‐treat
Participants	Number of participants: Number of participants screened for eligibility: 55 participants (four lost to follow‐up before receiving the injection) Number enrolled: 51 participants; 56 shoulders (27 participants; 28 shoulders in ultrasound‐guided injection group, 26 participants; 28 shoulders in anatomical landmark‐guided injection group) Number randomised: 56 shoulders (28 in each group) Number included in analyses at each follow‐up: 56 shoulders (28 in each group) at 6 weeks. Four of the patients (14% [4/28 shoulders]) in the blind group were offered a further injection at 6 weeks. One went on to have an acromioclavicular joint injection, two patients had a second subacromial injection performed at follow‐up and one declined. None of the US‐guided group had any repeat injections. 6/28 shoulders in the blind group and 4/28 in the US‐guided group underwent surgery after 6 weeks. Inclusion criteria: 1. Symptoms and signs of subacromial impingement syndrome: history of shoulder pain with overhead activities and clinical signs of impingement (either in internal rotation or external rotation) 2. Over 18 years of age To evaluate their existing pathological abnormalities and exclude any cause of pain other than subacromial impingement syndrome, ultrasound and radiography were performed on all participants before finalising their inclusion in the study. Exclusion criteria: 1. Previous surgery of the affected shoulder 2. Previous rotator cuff tears, calcific tendinitis, adhesive capsulitis, inflammatory arthritis, acromioclavicular joint pain, os acromiale, osteoarthritis, fractures, bone tumours, osteonecrosis, or other bone conditions seen on radiographs Baseline characteristics: Ultrasound‐guided injection group Mean (range) age: 46 (19‐68) years 14 female; 14 male Mean (range) duration of shoulder pain: 26 (1‐108) months Presence of workers compensation claim: 10 (35.7%) Mean (SEM) VAS for pain: 59 (5) (0 to 10; 0 no pain, 10 maximum pain) Mean (SEM) American Shoulder and Elbow Surgeons (ASES) scores: 57 (2) 5‐point Likert Scales: Mean (SEM) difficulty with overhead activities: 2.7 (0.2) Mean (SEM) frequency of pain during activity: 3.4 (0.1) Mean (SEM) frequency of pain when sleeping: 2.8 (0.3) Mean (SEM) level of pain when sleeping: 2.1 (0.2) Mean (SEM) frequency of extreme pain: 2.4 (0.3) Anatomic landmark‐guided injection group Mean (range) age: 42 (23‐62) years 18 female; 10 male Mean (range) duration of shoulder pain: 16 (2‐108) months Presence of workers compensation claim: 9 (32.1%) Mean (SEM) VAS for pain: 63 (4) (0 to 10; 0 no pain, 10 maximum pain) Mean (SEM) American Shoulder and Elbow Surgeons (ASES) scores: 54 (3) 5‐point Likert Scales: Mean (SEM) difficulty with overhead activities: 2.5 (0.2) Mean (SEM) frequency of pain during activity: 3.5 (0.1) Mean (SEM) frequency of pain when sleeping: 2.9 (0.2) Mean (SEM) level of pain when sleeping: 2.3 (0.2) Mean (SEM) frequency of extreme pain: 2.6 (0.2) Pretreatment group differences: Participants in the ultrasound‐guided group were slightly older and had longer duration of symptoms.
Interventions	All participants received an injection of 1 mL of 40 mg/mL methylprednisolone acetate (depo‐medrol; Pfizer) and 5 mL of 1% lidocaine hydrochloride (xylocaine; AstraZeneca). All injections were performed by a surgeon with over 10 years of experience performing subacromial injections. All participants received an ultrasound at the time of injection. Ultrasound‐guided injection (N = 28 shoulders): Ultrasound‐guided injections were performed using a lateral approach and a General Electric Logiq E9 machine with a 6‐ to 15‐MHz linear transducer and 50 x 10–mm footprint. With the patient in an upright sitting position, a 22‐gauge needle was placed directly underneath the midpoint of the lateral acromion. The needle was directed anteriorly and cephaladly toward the subacromial bursa (as guided by the ultrasound image) until the tip of the needle was seen in the bursa. Anatomical landmark‐guided injection (N = 28 shoulders): Blind injections were performed with the patient in the same upright sitting position with the ultrasound probe on the acromion to keep the patients blinded to the treatment group. An image could be seen on the screen by the patient, but it was not possible for this to be used by the surgeon to guide the injection, as the way that he was positioned behind the patient meant that he did not have direct vision of the screen. A posterior approach was used, and the needle was inserted 1 cm medially and inferiorly to the posterolateral corner of the acromion and directed cephaladly, anteriorly, and medially toward the subacromial bursa. Co‐interventions: Participants were not restricted from the use of any analgesic or anti‐inflammatory medication or other treatments such as physical therapy.
Outcomes	Outcomes were measured before the injection and at 6 weeks after the injection. Primary outcome: 1. Mean VAS for pain during overhead activities A 10‐cm line with ‘‘no pain’’ at one end and ‘‘the worst imaginable pain’’ at the other end was marked by the patient, and the distance from the no pain end was converted to a score out of 100 (1 mm = 1 point). Secondary outcomes: 2. Mean American Shoulder and Elbow Surgeons (ASES) score (0 to 100, higher score indicates better function. Domains: VAS pain (1 item); 4‐point Likert scale for function/disability (10 items)) 3. Mean 5‐point Likert Scale for frequency of pain during activity, pain when sleeping, and extreme pain, as well as difficulty with overhead activities and level of pain during sleep 4. Mean passive shoulder ROM in forward flexion, abduction, external rotation, and internal rotation, via visual assessment 5. Mean muscle force using a dynamometer in internal rotation and external rotation at neutral with the elbow at 90 degrees, supraspinatus position of 90 degrees of shoulder abduction in the scapula plane, hand behind the back in the lift‐off position, and adduction 6. Proportion with 50% improvement in pain during internal rotation and external rotation impingement tests 7. Proportion that received additional injections 8. Proportion that went on to have surgery Outcomes included in this review 1. Mean VAS for pain during overhead activities 2. Mean American Shoulder and Elbow Surgeons (ASES) score 3. Internal rotation and external rotation impingement tests 4. Proportion that received additional injections 5. Proportion that underwent surgery 6. Mean passive shoulder ROM in forward flexion and abduction
Sources of funding	Not reported
Notes	Conflicts of interest: Authors declared no conflicts of interest. Trial registration: Australian New Zealand Clinical Trials Registry: ACTRN12615000562572 Time points included in this review: 6 weeks Data analysis: Dr Cole provided mean (no SD or SEM) values for shoulder ROM in flexion, abduction, internal rotation and external rotation, and shoulder strength in internal rotation and external rotation at neutral with the elbow at 90 degrees, supraspinatus position of 90 degrees of shoulder abduction in the scapula plane, hand behind the back in the lift‐off position, and adduction for each treatment group at baseline and 6 weeks (as these data were not reported in the published paper). Results were reported in mean (SEM). Standard errors were transformed to standard deviations in the data analysis section. We used the SDs from the best study for ROM in abduction and flexion (Ekeberg 2009). We transformed the pain scores to be out of 10. Withdrawals: None Adverse events: No adverse events were reported in either group.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Computer‐generated randomisation
Allocation concealment (selection bias)	Unclear risk	Not explicitly reported
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants were blinded; the surgeon and sonographer performing the injections were not blinded, however they were not involved in providing any other care for the participants in the study. Quote: "Blind injections were performed with the patient in the same upright sitting position with the ultrasound probe on the acromion to keep the patients blinded to the treatment group. An image could be seen on the screen by the patient, but it was not possible for this to be used by the surgeon to guide the injection, as the way that he was positioned behind the patient meant that he did not have direct vision of the screen."
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	Low risk	As participants were blinded, there was a low risk of detection bias in the measurement of self‐reported outcomes.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	The assessor of range of motion was blinded, hence there was low risk of bias in the measurement of range of motion.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Data reported at follow‐up from all randomised participants
Selective reporting (reporting bias)	Low risk	All proposed outcomes were reported in the results and missing data on range of motion was provided by authors. Trial was registered.
Other bias	Low risk	None apparent

Dogu 2012.

Study characteristics
Methods	Design: Parallel‐group, two‐arm randomised controlled trial in participants with subacromial impingement syndrome Setting: Istanbul, Turkey Timing: Not reported Interventions: Ultrasound‐guided or anatomical landmark‐guided glucocorticoid injections into the subacromial space Sample size: Not reported Analysis: Intention‐to‐treat analysis not specified
Participants	Number of participants: Number of participants screened for eligibility: 50 (two had full‐thickness tears in the initial MRIs, one had a total rupture, and one was excluded from the study for refusing the control MRI after the injection) Number enrolled: 46 (23 per group) Number randomised: 46 (23 per group) Number included in analyses: 46 (23 per group) Inclusion criteria: 1. Patients diagnosed with subacromial impingement syndrome that had shoulder pain for at least 3 months. Subacromial impingement syndrome was suspected with the history taken; open passive ROM; no erythema, swelling, or atrophy at the muscles around shoulder; no pain extending to the below part of the elbow; no neurological deficit; and at least two positive results in the provocative tests (Neer, Hawkins, and Jobe’s ‘‘Empty Can’’). 2. MRI criteria for subacromial impingement syndrome: microstructural changes of rotator cuff muscle and tendons, signal intensity changes in tendons and contour deformation, disappearance of subacromial fat in coronal plane, or acromial spur indentation without microstructural changes 3. No shoulder problems other than subacromial impingement syndrome, confirmed by MRI Exclusion criteria: 1. Evidence of another shoulder pathology (e.g. adhesive capsulitis or calcific tendinitis, dislocations, chronic inflammatory arthritis, full‐thickness tears or total rupture, and labral tears) 2. Previous treatment such as physiotherapy or glucocorticoid injections 3. Use of regular systemic nonsteroidal anti‐inflammatory drugs or glucocorticoids 4. Cervical pain or other painful conditions such as fibromyalgia and polymyalgia rheumatica Baseline characteristics: Ultrasound‐guided injection group Mean (SD) age: 55.17 (9.24) years Female 15; male 8 Mean (SD) duration of shoulder pain (SD): 7.43 (5.37) months Mean (SD) VAS for pain at rest: 3.43 (2.74) Mean (SD) VAS for pain at activity: 7.04 (2.33) Mean (SD) VAS for pain at sleep: 6.83 (2.87) Mean (SD) Shoulder Disability Questionnaire (SDQ) scores: 81.48 (17.86) Mean (SD) Constant ROM score: 30.78 (5.96) Mean (SD) Constant General Pain score: 3.7 (2.7) Mean (SD) Constant Activities of Daily Living score: 14.39 (2.64) Anatomic landmark‐guided injection group Mean (SD) age: 56.74 (8.02) years Female 16; male 7 Mean (SD) duration of shoulder pain (SD): 9.74 (7.67) months Mean (SD) VAS for pain at rest: 2.61 (1.92) Mean (SD) VAS for pain at activity: 8 (1.38) Mean (SD) VAS for pain at sleep: 6.87 (2.51) Mean (SD) Shoulder Disability Questionnaire (SDQ) scores: 81.22 (14.4) Mean (SD) Constant ROM score: 32.26 (5.89) Mean (SD) Constant General Pain score: 3.26 (2.86) Mean (SD) Constant Activities of Daily Living score: 14.57 (3.03)
Interventions	All participants received an injection of 1 mL of 5 mg/mL betamethasone dipropionate, 9 mL of 10 mg/mL prilocaine hydrochloride and 0.02 mL of 0.01 mmol gadolinium diethylenetriaminepentaacetic acid. Ultrasound‐guided injection (N = 23) Ultrasound‐guided injections were performed by a musculoskeletal radiologist using an Aplio XV (Toshiba, Tokyo, Japan) and a 7.5‐ to 14‐MHz linear probe using a lateral approach. Sterile gel was applied to the probe. The probe was held in one hand, and the syringe with the mixture was held in the other hand. With the patient in an upright sitting position, a 21‐gauge needle was placed under the probe, and a lateral injection was made directly underneath the midlateral aspect of the acromion, directed anteriorly and cephaladly. Anatomical landmark‐guided injection (N = 23): Injections were given by a physiatrist using a posterior approach. Under the guidance of a Toshiba Power Vision ultrasonograph, gel was applied to a 7.5‐ to 14‐MHz linear probe, and it was placed on the trapezius muscle. A 21‐gauge needle was located 1 cm posteriorly and inferiorly to the border of the acromion and directed cephaladly, anteriorly, and medially toward the subacromial space. Co‐interventions: Participants could use only 2000 mg paracetamol daily; no patient received physical therapy.
Outcomes	Outcomes were measured before treatment and 6 weeks after treatment. Outcomes 1. Mean Constant ROM score (2 to 40, higher score indicates better ROM) 2. Mean VAS for pain at rest, and during activity and sleep (0 to 10; 0 no pain, 10 maximum pain) 3. Mean Constant General Pain score (0 to 15, higher score indicates less pain) 4. Mean Constant Activities of Daily Living score (0 to 10, higher score indicates better daily activity levels) 5. Mean Shoulder Disability Questionnaire (SDQ) (0 to 100, 16‐items, higher indicates worse disability) 6. Injection‐related adverse events or increases in pain 7. MRI assessment of injection location accuracy Outcomes included in this review 1. Mean VAS for pain at rest 2. Mean Shoulder Disability Questionnaire (SDQ) 3. Injection‐related adverse events or increases in pain
Sources of funding	Not reported
Notes	Conflicts of interest: Authors declared no conflicts of interest. Trial registration: Not reported Time points included in this review: 6 weeks Data analysis: Because a higher score in function measures in other trials indicated better function while a higher score on the SDQ indicated worse function, we transformed the SDQ scores by subtracting the value from 100 for the purpose of pooling data in Analysis 1.2. The authors were contacted for data on total Constant score, however, we did not receive any response. Withdrawals: None Adverse events: None were recorded in either group.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "The patients were randomised by a random number sequence to receive either ultrasound‐guided subacromial injections (group 1) or blind injections (group 2)."
Allocation concealment (selection bias)	Unclear risk	The study did not report if or how sequence allocation was concealed.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants were blinded. Personnel performing the injections were not blinded but were not involved in any other care of the participants in the trial.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	Low risk	Since participants were blinded, there was a low risk of detection bias.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	The assessor that measured range of motion was blinded to treatment allocation. Quote: "The patients were evaluated two times, before treatment and 6 weeks after treatment onset, by another physiatrist blinded to the type of subacromial injection."
Incomplete outcome data (attrition bias) All outcomes	Low risk	There were no withdrawals in either group for the 6‐week follow‐up.
Selective reporting (reporting bias)	Low risk	All proposed outcomes were reported in the results.
Other bias	Low risk	None apparent

Ekeberg 2009.

Study characteristics
Methods	Design: Single‐centre, parallel‐group, two‐arm, double‐blind randomised controlled trial in participants with rotator cuff disease Setting: Outpatient clinic of the Physical Medicine and Rehabilitation Department at Ullevål University Hospital in Norway Timing: April 2005 and October 2006 Interventions: Ultrasound‐guided glucocorticoid and lidocaine injection in the subacromial bursa and lidocaine injection in the gluteal region ('local' group) or glucocorticoid and lidocaine injection in the gluteal region and ultrasound‐guided lidocaine injection in the subacromial bursa ('systemic' group) Sample size: 106 participants were estimated to be needed based upon detecting a clinically relevant difference of 10 points in the Shoulder Pain and Disability Index (SPADI) (SD 20) at 6 weeks with 95% probability and 80% power (including a 10% rate of loss at follow‐up). Analysis: Intention‐to‐treat
Participants	Number of participants: Number of participants screened for eligibility: 312 (not meeting inclusion criteria = 173, refused to participate = 33) Number enrolled: 106 (53 per group) Number randomised: 106 (53 per group) Number included in analyses at each follow‐up: 104 (52 per group) at 2 weeks and 104 (53 in 'local' group and 51 in 'systemic' group) at 6 weeks Inclusion Criteria: Patients with rotator cuff disease according to the following criteria: 1. Age > 18 years 2. Shoulder pain for > 3 months 3. Pain on abduction 4. Less than a 50% reduced glenohumeral range of motion in no more than one direction of external rotation, internal rotation or abduction 5. Pain on two of three isometric tests for abduction, external rotation and internal rotation 6. Positive Hawkins‐Kennedy impingement sign Exclusion Criteria: 1. Symptomatic acromioclavicular arthritis 2. Clinical and radiological findings indicating glenohumeral joint pathology 3. Referred pain from the neck or internal organs 4. Generalised muscular pain syndrome with bilateral muscular pain in the neck and shoulders 5. History of inflammatory arthritis 6. Diabetes mellitus type 1 7. Previous fractures or surgery to the shoulder 8. Contraindications to local steroid injections 9. Receipt of any glucocorticoids in the month before inclusion 10. Shoulder pain and disability index (SPADI) score < 30/100 Baseline characteristics: 'Local' group (image‐guided steroid injection into the subacromial bursa): Mean (SD) age = 51 (11) years 32 female; 21 male Duration of symptoms (%): < 6 months: 15 (28%); 6 months‐1 year: 17 (32%); 1‐2 years: 7 (13%); > 2 years: 14 (26%) On sick leave: 17 (32%) Median (IQR) duration of sick leave: 8 (1‐14) Presence of bilateral pain: 8 (15%) Presence of a workers compensation claim: 5 (9%) Presence of a full thickness rotator cuff rupture: 5 (9%) Mean (SD) Hopkins symptom checklist score: 1.53 (0.42) Treatment history: Physiotherapy: 29 (55%); glucocorticoid injection: 22 (42%) Concurrent treatment: Physiotherapy: 9 (17%); analgesics: 19 (36%) Mean (SD) Shoulder Pain and Disability Index (SPADI) score: 53 (18)  Mean (SD) Western Ontario rotator cuff index: 45 (17) Median VAS for pain at rest: 6.0 Median VAS for pain during activity: 6.0 Mean (IQR) ROM abduction: 131 (98‐144) Mean (IQR) ROM flexion: 151 (132‐160) 'Systemic' group (Steroid injection into the gluteal region): Mean (SD) age = 50 (12) years 33 female; 20 male Duration of symptoms (%): < 6 months: 15 (28%); 6 months ‐1 year: 15 (28%); 1‐2 years: 11 (21%); > 2 years: 12 (23%) On sick leave: 14 (26%) Median (IQR) duration of sick leave: 4 (3‐17) Presence of bilateral pain: 9 (17%) Presence of a workers compensation claim: 3 (6%) Presence of a full thickness rotator cuff rupture: 5 (9%) Mean (SD) Hopkins symptom checklist score: 1.53 (0.38) Treatment history: Physiotherapy: 23 (43%); glucocorticoid injection: 20 (38%) Concurrent treatment: Physiotherapy: 6 (11%); analgesics: 21 (40%) Mean Shoulder Pain and Disability Index (SPADI) score (SD): 51 (17) Mean (SD) Western Ontario rotator cuff index: 47(16) Median VAS for pain at rest: 7.0 Median VAS for pain during activity: 7.0 Mean (IQR) ROM abduction: 126 (88‐144) Mean (IQR) ROM flexion: 150 (129‐158)
Interventions	All injections were performed in a standardised way by the same independent consultant physician who was responsible for preparing the syringes immediately before use according to the randomisation schedule. Both groups received injections of local anaesthetic in the shoulder and the gluteal region to improve blinding of participants by inducing a temporary pain relief (impingement test) and mask possible post‐injection pain. In both groups, the sonographically‐guided injection into the subacromial bursa was performed first, using commercial ultrasound equipment (Medison 128 BWprime, Medison Co, Seoul, Korea) with a 5‐9 MHz linear transducer for guidance. Patients sat facing the ultrasound screen with the affected arm rotated internally behind the back, elbow bent, and the back of the hand resting against the lower back. The physician used the ultrasound probe to visualise the insertion of the supraspinatus tendon and the subacromial bursa on the longitudinal axis, taking care that the contents of the syringes were never shown to the participants. The physician used the anterior approach with a 0.8 × 50 mm intra‐muscular needle for the subacromial injections, perforating the skin and tracking the needle in real time until it reached the subacromial bursa. The physician emptied the content of the syringe into the subacromial bursa, taking care to avoid direct injection into the rotator cuff tendons. The physician then gave patients the intramuscular injection in the upper gluteal region. 'Local' group (N = 53): Participants were allowed to use analgesics, and to continue any physiotherapy programme that they were attending at baseline. Participants were not allowed other additional treatment during the study.
Outcomes	Outcomes were measured at baseline, 2 weeks and 6 weeks. Primary outcome: 1. Mean Shoulder Pain and Disability Index (SPADI) (primary endpoint at 6 weeks) (0 to 100, higher score indicates worse pain and disability. Included five questions on Pain (0 = no pain; 10 = the worst pain imaginable) and eight on disability (0 = no difficulty; 10 = so difficult it requires help)) Secondary outcomes: 2. Mean Western Ontario Rotator Cuff Index (0 to 100, higher score indicates worse function. Domains: Physical symptoms (6 questions); Sports/Recreation (4 questions); Work (4 questions); Lifestyle (4 questions); Emotions (3 questions). All questions were rated on a VAS) 3. Mean active shoulder abduction and flexion ROM, measured in degrees using an electronic digital inclinometer 4. Mean 'participant’s assessment of change in the main complaint compared with baseline', measured on an 18‐point ordinal scale 5. Mean pain at rest measured on a nine‐point ordinal scale (1 to 9, 1 no pain; 9 maximum pain) 6. Mean pain with activity measured on a nine‐point ordinal scale (1 to 9, 1 no pain; 9 maximum pain) 7. Proportion that received additional injections 8. Adverse events Outcomes included in this review 1. Mean pain at rest measured on a nine‐point ordinal scale 2. Mean Shoulder Pain and Disability Index (SPADI) 3. Serious and other adverse events 4. Proportion that received additional injections 5. Mean active shoulder abduction and flexion ROM
Sources of funding	University of Oslo funded the study.
Notes	Conflicts of interest: Authors declared no conflicts of interest. Trial registration: Clinical trials NCT00640575 https://clinicaltrials.gov/ct2/show/NCT00640575 Time points included in this review: 2 weeks and 6 weeks Data analysis: Dr Ekeberg provided mean (SD) values for pain at rest, pain with activities, active abduction and flexion for each treatment group at baseline, 2 and 6 weeks (as these data were not reported in the published paper). We transformed the SPADI scores by subtracting the value from 100 for the purpose of pooling data in Analysis 1.2. We were unable to use 'participant’s assessment of change in the main complaint compared with baseline' as a measure of treatment success as this was reported as a median value in both groups. Withdrawals: 1/53 in the local group withdrew (unavailable for follow‐up) and 2/53 withdrew from the steroid group (1 was unavailable for follow‐up and 1 withdrew and then received a local steroid injection) at 2 weeks; 0/53 in the local group and 2/53 in the steroid group (1 was unavailable for follow‐up and 1 had a medical problem) withdrew at 6 weeks. Adverse events: 'Local' group: Serious adverse events: 0/53 Other adverse events: 1/53 reported post‐injection pain in the shoulder. 'Systemic' group: Serious adverse events: 0/53 Other adverse events: 4/53 reported post‐injection pain in the shoulder. Nine participants from both groups reported mild adverse effects such as facial redness, dizziness, and feeling of warmth, however, per group data was not given.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: “We used the computer program Clinstat to generate a predetermined randomisation sequence with a randomised variable block size of three, four, and five”.
Allocation concealment (selection bias)	Low risk	Quote: “The consultant physician responsible for the injections held the only copy of the randomisation sequence. After the baseline registrations, we gave each participant a study number. To ensure concealed allocation, each patient was referred to the consultant physician responsible for injections, who then allocated the patient to one of the treatment groups according to the randomisation sequence”.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants and study personnel were blinded to treatment allocation. The same interventionist administered injections in both groups, was not blinded but not involved in outcome assessments. Quote: “Both groups received injections of local anaesthetic in the shoulder and the gluteal region. We provided local anaesthetic to improve blinding of participants by inducing a temporary pain relief (impingement test) and mask possible post‐injection pain.”
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	Low risk	Since participants were blinded, there was a low risk of detection bias.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	Quote: "Patients ... were assessed by the same physician (OME) blinded for treatment group".
Incomplete outcome data (attrition bias) All outcomes	Low risk	1/53 in the local group withdrew (unavailable for follow‐up) and 2/53 withdrew from the steroid group (1 was unavailable for follow‐up and 1 withdrew and then received a local steroid injection) at 2 weeks; 0/53 in the local group and 2/53 in the steroid group (1 was unavailable for follow‐up and 1 had a medical problem) withdrew at 6 weeks. These losses to follow‐up were reported and analysed on an intention‐to‐treat basis.
Selective reporting (reporting bias)	Low risk	All proposed outcomes were reported in the results. Trial registered
Other bias	Low risk	None apparent

Haghighat 2015.

Study characteristics
Methods	Design: Single‐centre parallel‐group, two‐arm randomised controlled trial in participants with subacromial impingement syndrome Setting: Physical medicine and rehabilitation outpatient clinics in Isfahan University of Medical Science (Al‐Zahra hospital), Iran Timing: February 2014 to February 2015 Interventions: Ultrasound‐guided or anatomical landmark‐guided glucocorticoid injections into the subacromial bursa Sample size: Not reported Analysis: Intention‐to‐treat analysis not specified
Participants	Number of participants: Number of participants screened for eligibility: 48 (three were excluded due to predisposing conditions including diabetes and five were unwilling to participate) Number enrolled: 40 (20 per group) Number randomised: 40 (20 per group) Number included in analyses at follow‐up: not reported Inclusion criteria: 1. Posterolateral shoulder pain 2. Pain in abduction or painful restriction of glenohumeral mobility 3. Provocation of symptoms by Neer and Hawkins tests Exclusion criteria: 1. Shoulder pain due to osseous pathology (e.g. osteoarthritis, osteonecrosis) 2. Duration of shoulder pain more than three months 3. Previous trauma in the shoulder region 4. Physiotherapy or local steroid injections in last three months 5. Any evidence of predisposing conditions such as diabetes mellitus, rheumatoid arthritis, and hypothyroidism 6. Unwillingness to participate in the present study Baseline characteristics: Ultrasound‐guided injection group Mean age (SD): 50.45 (6.78) years 12 female; 8 male Mean (SD) BMI: 27.05 (2.21) Mean (SD) duration of shoulder pain: 1.8 (0.54) months Mean (SD) VAS: 8.1 (1.11) Mean (SD) Shoulder Pain And Disability Index (SPADI) pain score: 37.65 (6.5) Mean (SD) Shoulder Pain And Disability Index (SPADI) disability score: 53.45 (9.94) Mean (SD) abduction ROM: 132.5 (17.5) Mean (SD) flexion ROM: 147.5 (15.43) Mean (SD) external rotation ROM: 68 (12.07) Mean (SD) internal rotation ROM: 62.25 (16.42) Anatomic landmark‐guided injection group Mean age (SD): 52.3 (7.48) years 13 female; 7 male Mean (SD) BMI: 26.3 (3.8) Mean (SD) duration of shoulder pain: 1.87 (0.48) months Mean (SD) VAS: 6.75 (1.16) Mean (SD) Shoulder Pain And Disability Index (SPADI) pain score: 32.75 (6.12) Mean (SD) Shoulder Pain And Disability Index (SPADI) disability score: 42.1 (10.41) Mean (SD) abduction ROM: 129 (21.67) Mean (SD) flexion ROM: 145.25 (9.93) Mean (SD) external rotation ROM: 71.5 (16.94) Mean (SD) internal rotation ROM: 62 (17.27) Patients were diagnosed as having subacromial impingement syndrome based on clinical criteria, which were also confirmed by ultrasounds and x‐rays for ruling out osseous pathology.
Interventions	All patients received 40 mg methylprednisolone with 1 cc lidocaine 2%. Ultrasound‐guided injection (N = 20) A lateral approach using the ultrasound equipment with high frequency linear transducer for guidance was adopted. The ultrasound probe was positioned parallel to the long axis of the supraspinatus. The skin was punctured at a distance of about 2–3 cm from the probe in order to avoid contact between the needle and the probe. As soon as the needle had penetrated the subcutaneous tissues, its progress was real‐time monitored on the US image. The needle was visualised as a hyperechoic structure with posterior comet‐tail artefact. Progress of the needle until the tip had entered the subacromial bursa was observed. When the tip of the needle appeared to be inside the subacromial bursa, a small amount of liquid was injected to confirm the correct position (fluid passing from the needle tip into the bursa). The injection can be visualised in real‐time as a spreading "cloud" of hyperechoic echoes inside the bursa. Anatomical landmark‐guided injection (N = 20): A posterior approach was used. The needle was placed approximately 2 cm below the posterior‐lateral aspect of the acromion. The needle was guided forward and slightly to the left under the acromion. Co‐interventions: Not reported
Outcomes	All outcomes were assessed at baseline and 6 weeks after the injection. Outcomes: 1. Mean VAS for pain (0 to 10, 0 no pain; 10 maximum pain) 2. Mean Shoulder Pain and Disability Index (SPADI) (0 to 100, higher score indicates worse pain and disability. Included five questions on Pain (0 = no pain; 10 = the worst pain imaginable) and eight on disability (0 = no difficulty; 10 = so difficult it requires help)) 3. Mean shoulder ROM in flexion, abduction, external rotation, and internal rotation using a goniometer (not specified if active or passive) Outcomes included in this review 1. Mean VAS for pain 2. Mean shoulder Pain and Disability Index (SPADI) 3. Mean shoulder ROM in flexion, abduction and external rotation
Sources of funding	Not reported
Notes	Conflicts of interest: Authors declared no conflicts of interest. Trial registration: Not reported Timepoints included in this review: 6 weeks Data analysis: Because a higher score in function measures in other trials indicated better function while a higher score on the SPADI indicated worse function, we transformed the SPADI scores by subtracting the value from 100 for the purpose of pooling data in Analysis 1.2. Withdrawals: Not reported Adverse events: Not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Unclear how randomisation was performed. Quote: "Participants were randomly assigned to control and case groups, two groups of 20 subjects, using randomised allocation."
Allocation concealment (selection bias)	Unclear risk	The study did not report if or how sequence allocation was concealed.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Study was reported to be single‐blind but it appeared that participants and study personnel other than the outcome assessor were not blinded as procedures were described to have been performed either with or without ultrasound guidance without any attempt to blind participants to treatment allocation.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	High risk	Self‐reported outcomes could have been subject to detection bias as participants were aware of their treatment allocation.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	The study did not report whether or not assessors were blinded to treatment allocation.
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	The number of participants who completed the six‐week assessment was not reported.
Selective reporting (reporting bias)	Low risk	Although the trial was not registered, all proposed outcomes were reported in the results.
Other bias	Unclear risk	Baseline function (assessed by SPADI) was higher (worse) in the ultrasound‐guided injection group vs. the anatomic landmark‐guided injection group which may have biased the result in favour of the ultrasound‐guided group.

Hsieh 2013.

Study characteristics
Methods	Design: Single‐centre, parallel‐group, two‐arm randomised controlled trial in participants with chronic subacromial bursitis Setting: Outpatient clinic of the Department of Physical Medicine and Rehabilitation at Shin Kong Wu Ho‐Su Memorial Hospital, Taiwan Timing: March 2010 to November 2011 Interventions: Ultrasound‐guided or palpation‐guided glucocorticoid injections into the subacromial (or sub‐deltoid) bursa Sample size: Not reported Analysis: Data were analysed in those who completed the study.
Participants	Number of participants: Number of participants screened for eligibility: 145 (31 did not satisfy the inclusion criteria and 18 were unwilling to sign the consent form) Number enrolled: 96 (48 per group) Number randomised: 96 (48 per group) Number included in analyses at each follow‐up: 95 (47 in ultrasound‐guided injection group and 48 in palpation‐guided injection group) at week 1 and 92 (46 per group) at 1 month Baseline characteristics: Ultrasound‐guided group Mean (SD) age: 57.59 (10.30) years 27 female; 19 male Mean (SD) duration of shoulder pain: 6.28 (3.59) months 28 exercisers; 18 non exercisers Mean (SD) weight: 64.33 (12.84) kg Mean (SD) height: 159.93 (8.27) cm Mean (SD) VAS for pain: 6.57 (2.04) Mean (SD) Shoulder Disability Questionnaire (SDQ): 40.47 (7.79) Mean (SD) Shoulder Pain And Disability Index (SPADI) total score: 50.85 (18.28) Mean (SD) Shoulder Pain And Disability Index (SPADI) pain score: 58.17 (20.49) Mean (SD) Shoulder Pain And Disability Index (SPADI) disability score: 43.53 (19.01) Mean (SD) sub‐item scores on the SF‐36: Social functioning: 73.91 (17.51) Physical role: 35.33 (40.34) Bodily pain: 49.27 (17.33) General health: 59.04 (23.11) Vitality: 60.87 (19.73) Physical functioning: 78.26 (19.08) Emotional role: 63.77 (44.35) Mental health: 68.43 (19.34) Anatomic landmark‐guided group Mean (SD) age: 55.87 (11.42) years 29 female; 17 male Mean (SD) duration of shoulder pain: 7.14 (4.72) months 24 exercisers; 22 non exercisers Mean (SD) weight: 61.20 (10.10) kg Mean (SD) height: 160.67 (7.78) cm Mean (SD) VAS for pain: 7.39 (1.86) Mean (SD) Shoulder Disability Questionnaire (SDQ): 40.31 (9.01) Mean (SD) Shoulder Pain And Disability Index (SPADI) total score: 59.78 (20.88) Mean (SD) Shoulder Pain And Disability Index (SPADI) pain score: 66.04 (22.10) Mean (SD) Shoulder Pain And Disability Index (SPADI) disability score: 53.51 (22.58) Mean (SD) sub‐item scores on the SF‐36: Social functioning: 66.38 (20.82) Physical role: 34.57 (39.20) Bodily pain: 44.21 (15.95) General health: 56.47 (19.09) Vitality: 52.55 (21.36) Physical functioning: 81.12 (17.46) Emotional role: 61.70 (42.27) Mental health: 64.17 (18.31) Inclusion criteria: 1. Shoulder pain for more than 1 month 2. Painful abduction with a visual analog scale (VAS) score for pain of 4 or more 3. Painful arc of motion or pain at the mid to end range of shoulder abduction or internal rotation and a soft end feel 4. Tenderness over the subacromial bursa 5. A reduction in pain of > 40% on active shoulder abduction at end range after the injection of 3 mL of 1% lidocaine into the subacromial bursa Exclusion criteria: 1. History of uncontrolled chronic diseases (e.g. malignant neoplasms, hypocoagulability, and infection) 2. Previous surgery of the affected shoulder 3. Any evidence of a rotator cuff tear or tendinopathy, demonstrated by positive resistive tests or sonographic findings 4. Calcification of the rotator cuff, demonstrated by x‐ray or sonographic findings 5. Presence of inflammatory arthritis (e.g. rheumatoid arthritis, seronegative spondyloarthropathy, and crystal‐related arthropathy), osteoarthritis, frozen shoulder, subacromial spurs, or deformity of the acromion 6. Instability of the affected shoulder 7. Previous fracture near the shoulder region 8. Presence of cervical radiculopathy or myelopathy 9. Having received a glucocorticoid or hyaluronate subacromial or shoulder joint injection in the past 3 months Pretreatment group differences: Baseline SF‐36 sub‐item values were higher in the USG group compared with the palpation‐guided group (study reported better SF‐36 outcomes in USG group)
Interventions	All of the injections were administered by a senior physician who was a board‐certified rheumatologist, physiatrist, and ultrasonographer in musculoskeletal medicine. All participants received an injection of 0.5 mL dexamethasone suspension (5mg/mL) and 3 mL lidocaine (10 mg/mL) into the subacromial bursa. Ultrasound‐guided injection (N = 48) The physician used the LOGIQ P5 (General Electronic Company, Milwaukee, WI) machine for ultrasonographic examinations and injection guidance. The probe, a 5‐ to 12‐MHz linear array transducer, was used to assess the shoulders and to guide the injection needles. After the sterilisation of the skin on the lateral side of the affected arm, a 21‐gauge needle was inserted into the subacromial bursa under ultrasound guidance. Any effusion, if present, was aspirated prior to the injection. Palpation‐guided injection (N = 48): The physician injected the subacromial bursa according to Cyriax’s method, where they first localised the lateral edge of the acromion. They then asked the patient to relax the affected arm, and at 1–2 cm beneath the middle point of the lateral edge of the acromion, inserted a 21‐gauge needle medially and in a slightly cranial direction into the bursa. If the needle encountered resistance, either the coracoacromial ligament or the capsulotendinous structures have been contacted, and the needle position was slightly adjusted. Co‐interventions: Not reported
Outcomes	The primary outcome measures were evaluated before, immediately, 1 week, and 1 month after the injection; the secondary outcome measures were evaluated before, 1 week, and 1 month after the injection. Primary outcomes: 1. Mean VAS for pain (0 to 10, 0 no pain; 10 maximum pain) 2. Mean active and passive shoulder ROM in flexion, abduction, external rotation, and internal rotation, using a goniometer Secondary outcomes: 1. Mean Shoulder Pain and Disability Index (SPADI) (0 to 100, higher score indicates worse pain and disability. Included five questions on Pain (0 = no pain; 10 = the worst pain imaginable) and eight on disability (0 = no difficulty; 10 = so difficult it requires help)) 2. Mean Shoulder Disability Questionnaire (SDQ) (0 to 100, 16 items, higher score indicates worse function) 3. Mean Short‐Form Health Survey (SF‐36) (0 to 100, higher score indicates better quality of life. Sub‐items: Social functioning; Physical role; Bodily pain; General Health; Vitality; Physical functioning; Emotional role; Mental health) Outcomes included in this review 1. Mean VAS for pain 2. Mean Shoulder Pain and Disability Index (SPADI) 3. Mean Short‐Form Health Survey (SF‐36) mental health scores
Sources of funding	Financial support was provided by the Shin Kong Wu Ho‐Su Memorial Hospital.
Notes	Conflicts of interest: Authors declared no conflicts of interest. Trial registration: Not reported Time points included in this review: baseline, 1 weeks and 1 month Data analysis: Because a higher score in function measures in other trials indicated better function while a higher score on the SPADI indicated worse function, we transformed the SPADI scores by subtracting the value from 100 for the purpose of pooling data in Analysis 1.2. Mean active and passive shoulder ROM in flexion and abduction could not be included because the figures were unclear and the authors were unable to provide these data. Withdrawals: 1/48 in the palpation‐guided group and 0/48 in the ultrasound‐guided group did not complete the 1‐week follow‐up. 2/48 in the palpation‐guided group and 2/48 in the ultrasound‐guided group did not complete the 1‐month follow‐up. Reasons for non‐completion were provided and were similar across groups. Adverse events: Not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "The assignment scheme was generated using a table of computer‐generated random numbers."
Allocation concealment (selection bias)	Unclear risk	The study did not report if or how sequence allocation was concealed.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Quote: "The patients and the physician administering the injections were not blinded."
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	High risk	Since participants were not blinded, there was a high risk of detection bias.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	Quote: "The outcome assessor was blinded to the treatment assignment."
Incomplete outcome data (attrition bias) All outcomes	Low risk	1/48 in the palpation‐guided group and 0/48 in the ultrasound‐guided group did not complete the 1‐week follow‐up. 2/48 in the palpation‐guided group and 2/48 in the ultrasound‐guided group did not complete the 1‐month follow‐up. Reasons for non‐completion were provided and were similar across groups.
Selective reporting (reporting bias)	Unclear risk	Range of motion data were not reported clearly and authors failed to provide these data when requested. Trial was not registered and study protocol was not reported.
Other bias	Low risk	None apparent

Lee 2009.

Study characteristics
Methods	Design: Single‐centre, parallel‐group, two‐arm controlled trial in participants with adhesive capsulitis Setting: Department of Rehabilitation Medicine, Ilsan‐Paik Hospital, Gyonggido, South Korea Timing: October 2006 to May 2007 Interventions: Ultrasound‐guided or anatomical landmark‐guided intra‐articular injection Sample size: Not specified Analysis: Intention‐to‐treat analysis not specified
Participants	Number of participants: Number of participants screened for eligibility: not reported Number enrolled: 43 (21 in ultrasound‐guided injection group and 22 in the anatomic landmark‐guided injection group) Number randomised: 43 participants were alternately allocated to 2 groups (21 in the ultrasound‐guided injection group and 22 in the anatomical landmark‐guided injection group) Number included in analyses at each follow‐up: 40 (20 per group) at 6 weeks Baseline characteristics: Ultrasound‐guided injection group Mean (SD) age: 53.1 (7.3) years 11 female; 9 male Mean (SD) height: 163.0 (7.7) Mean (SD) weight: 57.3 (11.4) Mean (SD) duration of shoulder pain: 8.5 (4.5) months Presence of diabetes mellitus: 5 Mean (SD) VAS for pain during the day: 5.4 (0.5) Mean (SD) VAS for pain before sleep: 5.6 (0.5) Mean (SD) Functional Activities score: 27.3 (2.7) Mean (SD) flexion ROM: 112.9 (5.6) Mean (SD) abduction ROM: 106.7 (5.9) Mean (SD) external rotation ROM: 48.9 (7.7) Mean (SD) internal rotation ROM: 31.2 (3.5) Anatomical landmark‐guided injection group Mean (SD) age: 54.1 (7.74) years 10 female; 10 male Mean (SD) height: 165.2 (7.2) Mean (SD) weight: 51.5 (7.4) Mean (SD) duration of shoulder pain: 10.6 (7.7) months Presence of diabetes mellitus: 3 Mean (SD) VAS for pain during the day: 6.0 (0.4) Mean (SD) VAS for pain before sleep: 6.1 (0.4) Mean (SD) Functional Activities score: 26.7 (1.8) Mean (SD) flexion ROM: 110.7 (6.8) Mean (SD) abduction ROM: 101.1 (6.3) Mean (SD) external rotation ROM: 46.4 (5.3) Mean (SD) internal rotation ROM: 28.7 (3.4) Inclusion criteria: Diagnosis of clinical disease stage II idiopathic adhesive capsulitis on the basis of: 1. Main complaint of limited range of movement of the shoulder joint and pain; and 2. normal X rays of the shoulder joint; and 3. ultrasound of the shoulder showing no particular problems such as rotator cuff injury, calcific tendinopathy, and subacromial bursitis.  Exclusion Criteria: 1. Other disorders that could cause similar clinical symptoms 2. Cervical examination showing significant dysfunction suggestive of cervical disk disorder or radiculopathy 3. Ultrasound showing rotator cuff injury, calcific tendinopathy or subacromial bursitis
Interventions	All participants received a 0.5 mL intra‐articular injection of triamcinolone (20 mg) mixed with 1.5 mL of 2% lidocaine and 3 mL of normal saline in the first week. This was followed by a 2.5 mL injection of low‐molecular‐weight sodium hyaluronate (25 mg) in the second week and then once weekly for the next 4 weeks for a total of 5 hyaluronate injections. The posterior approach was used for both groups. Patients were seated on a chair, with the affected shoulder bent and adducted. Ultrasound‐guided injection (N = 21): The team used the LOGIQ 5a and a 10‐MHz linear probe. The probe touched the lower part of the acromion sideways, and a long needle (23G, 7 cm) was inserted to the posterior articular surface of the humeral head and positioned inside the articular capsule. The expansion of the articular capsule was checked while the fluid was being injected. All ultrasound‐guided injections were performed by one doctor with two years experience in ultrasound‐guided shoulder injections. Anatomical landmark‐guided injection (N = 22): The acromion of the scapula was palpated and the needle inserted 1 cm inferior to the tip of the acromion. The needle was directed toward the coracoid process and advanced into the articular capsule, after which the drug was injected. All anatomic landmark‐guided injections were performed by one doctor with seven years experience in anatomic landmark‐guided shoulder injections. Co‐interventions: All participants were taught exercises for increasing joint ROM including stretching forward and bending down to a desk, Codman exercise, and a wall‐climbing exercise with the fingers. They were instructed to practice these at home. Participants were checked and encouraged to keep up with the exercises every time they visited the hospital. All had 5 weekly intra‐articular injections of 25 mg sodium hyaluronate.
Outcomes	All outcomes were measured at baseline and 1, 2, 3, 4, 5, 6 weeks. Outcomes: 1. Mean VAS for pain during the day time (0 to 10, 0 no pain; 10 maximum pain) 2. Mean VAS for pain just before sleep (0 to 10, 0 no pain; 10 maximum pain) 3. Mean passive shoulder ROM in abduction, flexion, extension, internal rotation and external rotation, using a goniometer 4. Mean functional activities of the shoulder score (evaluated pain intensity while the patient performs ten common functional activities in daily life: putting on clothes, combing hair with the affected arm, using a spoon and chopsticks during a meal, writing or typing on keyboards, carrying a handbag, cleaning the anus, catching something overhead, putting a hand on the back, doing routine sports, and sleeping on the painful side. Each function was evaluated on a 0 to 4 scale where 4 = normal condition, 3 = slight difficulty, 2 = considerable difficulty, 1 = requires help and 0 = impossible to perform the function. Total score from 0 to 40, higher scores indicate better function) No primary outcome was specified. Outcomes included in this review 1. Mean VAS for pain before sleep 2. Mean functional activities of the shoulder score 3. Mean passive shoulder ROM in abduction, flexion and external rotation
Sources of funding	No funding specified
Notes	Conflicts of interest: Authors declared no conflicts of interest. Trial registration: Not reported Time points included in this review: 1, 2, 3, 4, 5, 6 weeks Data analysis: As co‐interventions were applied equally across groups, data from all time points were used in the meta‐analysis. Mean functional activities of the shoulder score data were transformed from 0 to 40 to 0 to 100. Withdrawals: 2/22 in the anatomical landmark‐guided injection group and 1/21 in the ultrasound‐guided injection group at 6 weeks Adverse events: Not reported
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	High risk	Quote: “They were alternately assigned to either US‐guided injection or blind injection groups in the sequence in which they joined the study.”
Allocation concealment (selection bias)	High risk	As participants were alternately allocated into groups (see above), allocation was not concealed.
Blinding of participants and personnel (performance bias) All outcomes	Low risk	Participants were blind to treatment allocation. Quote: “To mimic injection under the guidance of US, we first contacted the patients in the blind injection group by an ultrasound probe at the skin under the acromion before the intra‐articular injection”. Personnel performing the injections were not blinded but were not involved in any other care of the participants in the trial.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	Low risk	Since participants were blinded, there was a low risk of detection bias.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	Quote: “One tester, who did not know which patient belonged to which group, performed the evaluations before and after treatment”.
Incomplete outcome data (attrition bias) All outcomes	Low risk	2/22 in the anatomical landmark‐guided injection group and 1/21 in the ultrasound‐guided injection group did not complete the 6‐week follow‐up. Reasons for non‐completion were not given.
Selective reporting (reporting bias)	Unclear risk	Results for all prespecified outcomes in the methods section were reported by authors. The standard deviation for continuous outcomes looked small, and it is possible that the trial authors had inadvertently reported standard errors.
Other bias	Low risk	None apparent

Lee 2015.

Study characteristics
Methods	Design: Single‐centre, parallel‐group, two‐arm randomised controlled trial in participants with shoulder stiffness Setting: Seoul St. Mary’s Hospital, School of Medicine, The Catholic University of Korea, Seoul, South Korea Timing: April 2008 to March 2012. Interventions: Ultrasound‐guided or anatomical landmark‐guided intra‐articular injection Sample size: Sample sizes were calculated, with a significant difference of 2 points in pain VAS between 2 groups. Minimal 2‐point difference in 11‐point pain numerical system was reported to have clinical significance in a broad population of patients with various musculoskeletal conditions. A sample size of 27 patients in each group was required for power of 95% at a type I error of 0.05 with a standard deviation of 2 points Analysis: Intention‐to‐treat analysis not specified
Participants	Number of participants: Number of participants screened for eligibility: 115 Number enrolled: 90 (excluded = 21; refused to participate = 4) Number randomised: 90 (45 in the ultrasound‐guided group and 45 in the anatomical landmark‐guided injection group) Number included in analyses at each follow‐up: 77 (38 in the ultrasound‐guided injection group and 39 in the anatomical landmark‐guided injection group) for all time points Baseline characteristics: Ultrasound‐guided injection group Median (range) age: 57.4 (47‐78) years 27 female; 11 male No. with diabetes mellitus: 2 No. with thyroid diseases: 1 Mean follow‐up period: 15.36 months Mean flexion ROM: 127.7 Mean external rotation at 90 degrees abduction ROM: 73.9 Mean external rotation at side ROM: 72.7 Mean internal rotation ROM: 3.88 Mean VAS for pain: 6.8 Mean American Shoulder and Elbow Surgeons (ASES) score: 48.1 Mean Simple Shoulder Test (SST): 47.9 Anatomical landmark‐guided injection group Median (range) age: 53.9 (45‐76) years 32 female; 7 male Presence of diabetes mellitus: 3 Presence of thyroid diseases: 1 Mean follow‐up period: 16.38 months Mean flexion ROM: 117.5 Mean external rotation at 90 degrees abduction ROM: 70.4 Mean external rotation at side ROM: 64.6 Mean internal rotation ROM: 3.96 Mean VAS for pain: 6.9 Mean American Shoulder and Elbow Surgeons (ASES) score: 42.5 Mean Simple Shoulder Test (SST): 42.9 Inclusion criteria: Diagnosis of shoulder stiffness with respect to ROM on the basis of: 1. Forward flexion (FF) less than 100 degrees of glenohumeral motion or external rotation less than 30 degrees, or internal rotation (IR) of the back at a level lower than the first lumbar spine junction Exclusion Criteria: 1. Secondary cause of shoulder stiffness ‐ plain radiography for shoulder (true anteroposterior, supraspinatus outlet and axial views) and ultrasonography were performed to determine the secondary cause of shoulder stiffness including rotator cuff tear, calcific tendinitis, osteoarthritis, and fracture. 2. Previous history of operation 3. Trauma or fracture
Interventions	All participants received 40 mg triamcinolone acetonide and 2 mL of 2% lidocaine. The anterior approach was used for both groups. In both groups, the patients were positioned on a standard gurney with the arm at the side. All ultrasound‐guided injections were performed by one experienced senior surgeon with more than 5 years of experience in performance of ultrasound‐guided injections at the shoulder joint. Ultrasound‐guided injection (N = 38): The US‐guided injection was performed according to the Valls and Melloni method with a slight modification: other than performing injections in the supine position, the patient position and preparation were the same as those in the blind injection group. A high‐resolution transducer with 12 MHz linear array was used to visualise the needle. After skin and transducer preparation with alcohol and povidone, the 21‐gauge 1.5‐inch needle on a 3 mL syringe with triamcinolone acetonide 40 mg and 2 mL of 2% lidocaine was inserted at the level of the coracoids, from lateral to medial, aimed at the medial border of the humeral head. When the needle made contact with the articular cartilage of the humeral head, the needle was tilted to position the point of the needle in the articular cavity. The intra‐articular position of the needle and fluid infiltration in the shoulder joint was confirmed by ultrasonography. Anatomical landmark‐guided injection(N = 39): With the patient sitting, the sulcus between the lateral tip of the coracoid and the humeral head was palpated. Then, a 21‐gauge 1.5‐inch needle on a 3 mL syringe with triamcinolone acetonide 40 mg and 2 mL of 2% lidocaine was inserted in a slightly cephalad direction at the anterior‐lateral tip of the coracoid. The needle was slowly advanced while infiltrating the local anaesthetic until resistance was lost, indicating intra‐articular position. The injection was performed slowly. Co‐interventions: All patients were prescribed the same home‐based exercise programme for rehabilitation. Starting from the first day after injection, pendulum circumduction and passive shoulder exercises for self‐stretching in forward flexion, external rotation and internal rotation in the supine position were recommended until 4 weeks after injection. Pulley exercise was prescribed for achieving advanced further flexion after 4 weeks of injection, and isometric exercise in all planes was recommended. From 8 weeks after injection, thera‐band exercise, strengthening exercise for scapular stabilisation and advanced muscle strengthening exercise with dumbbells were taught. All listed procedures were recommended until the last visit after 12 months. No limit was imposed on use of the shoulder within a tolerable extent.
Outcomes	All outcomes were measured at baseline and 3, 6 and 12 months. Outcomes: 1. Mean VAS (0 to 10, 0 no pain; 10 maximum pain) 2. Mean American Shoulder and Elbow Surgeons (ASES) score (0 to 100, higher score indicates better function. Domains: VAS pain (1 item); 4‐point Likert scale for function/disability (10 items)) 3. Mean Simple Shoulder Test (SST) (0 to 12, higher scores mean better function. Consisted of 12 yes/no questions. For each question, the patients indicated that they were able or are not able to do the activity specified). 4. Mean shoulder ROM in flexion, external rotation at 90 degrees abduction, external rotation by side, and internal rotation (assessment method not specified; not specified if active or passive) No primary outcome was specified. Outcomes included in this review 1. Mean VAS for pain 2. Mean American Shoulder and Elbow Surgeons (ASES) score 3. Mean shoulder ROM in flexion and external rotation by side
Sources of funding	None reported
Notes	Conflicts of interest: Authors declared no conflicts of interest. Trial registration: Not reported Timepoints included in this review: 3, 6 and 12 months Data analysis: We used shoulder ROM data for external rotation by side (instead of external rotation at 90 degrees abduction) to be consistent with other studies. We calculated SDs from P values for pain, function and ROM. Withdrawals: 6/45 (13%) in the anatomical landmark‐guided injection group (lost to follow‐up = 3, discontinued intervention due to continued pain after index injection = 3) and 7/45 (15%) in the ultrasound‐guided injection group (lost to follow‐up = 5, discontinued intervention due to continued pain after index injection = 2) Adverse events: None reported in either group
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "patients were randomised into 2 groups through computer‐generated blocked randomization."
Allocation concealment (selection bias)	Unclear risk	It was unclear whether allocation was concealed.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and study personnel other than the outcome assessor were not blinded as procedures were described to have been performed either with or without ultrasound guidance without any attempt to blind participants to treatment allocation.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	High risk	Since participants were not blinded, there was a high risk of detection bias for self‐reported outcomes.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	As the outcome assessor was blinded. there was low risk of bias in the measurement of assessor‐reported outcomes.
Incomplete outcome data (attrition bias) All outcomes	Low risk	6/45 (13%) in the anatomical landmark‐guided injection group (lost to follow‐up = 3, discontinued intervention due to continued pain after index injection = 3) and 7/45 (15%) in the ultrasound‐guided injection group (lost to follow‐up = 5, discontinued intervention due to continued pain after index injection = 2)
Selective reporting (reporting bias)	Unclear risk	There was no protocol paper and the trial was not registered. Authors did not report measures of variance in the paper.
Other bias	Low risk	None apparent

Naredo 2004.

Study characteristics
Methods	Design: Single‐centre, parallel‐group, two‐arm randomised controlled trial in participants with periarticular disorders of the shoulder (impingement syndrome, rotator cuff lesions, subacromial‐subdeltoid bursitis, and/or biceps tendon abnormalities) Setting: Madrid, Spain. No further details reported Timing: Not reported Interventions: Ultrasound‐guided injection into either the subacromial‐subdeltoid bursa, biceps tendon sheath or rotator cuff calcifications depending upon the ultrasound findings, or anatomical landmark‐guided injection into the subacromial space Sample size: Not specified Analysis: Intention‐to‐treat analysis not specified
Participants	Participants: Number of participants screened for eligibility: not reported Number enrolled: 41 (21 in ultrasound‐guided group and 20 in anatomic landmark‐guided group) Number randomised: 41 (21 in ultrasound‐guided group and 20 in anatomic landmark‐guided group) Number included in analyses at each follow‐up: not reported Baseline characteristics: Ultrasound‐guided group Mean (SD) age: 52.9 (11) years 15 female; 6 male Mean (SD) symptom duration: 11.9 months (14.6) months Shoulder demand: High 9 (43%); moderate 11 (52%); low 1 (5%) Presence of nocturnal pain: 20 (95%) Presence of active ROM impairment: 8 (38%) Mean (SD) VAS for pain during the previous week (SD): 61.2 (21.2) Mean Shoulder Function Assessment (SFA) score (SD): 42.6 (14.5) Ultrasound findings: increased fluid (biceps tendon sheath), 5 (24%); biceps tendon partial tear, 1 (5%); rotator cuff tendinosis, 3 (14%); rotator cuff partial‐thickness tear, 3 (14%); rotator cuff full‐thickness tear, 10 (48%); rotator cuff calcification, 4 (19%); increased fluid (subacromial/subdeltoid bursa), 9 (43%); rotator cuff impingement, 17 (81%); degenerative changes in the acromioclavicular joint, 16 (76%) Anatomic landmark‐guided group Mean (SD) age: 51.9 (13.8) years 12 female; 8 male Mean (SD) duration of symptoms: 10.2 (14.7) months Shoulder demand: High 6 (30%); moderate 9 (45%); low 5 (25%) Presence of nocturnal pain: 20 (100%) Presence of active ROM impairment: 9 (45%) Mean (SD) VAS for pain during the previous week: 63.7 (19.8) Mean (SD) Shoulder Function Assessment (SFA) score: 39.3 (13.4)  Ultrasound findings: increased fluid (biceps tendon sheath), 5 (25%); biceps tendon partial tear, 1 (5%); rotator cuff tendinosis, 6 (30%); rotator cuff partial‐thickness tear, 5 (25%); rotator cuff partial‐thickness tear, 6 (30%); rotator cuff calcification, 3 (15%); increased fluid (subacromial/subdeltoid bursa), 8 (40%); rotator cuff impingement, 16 (80%); degenerative changes in the acromioclavicular joint, 13 (65%) Inclusion criteria: 1. First flare of shoulder pain of periarticular aetiology (impingement syndrome, rotator cuff lesions, subacromial‐subdeltoid bursitis, and/or biceps tendon abnormalities) diagnosed by clinical history and examination 2. At least one month duration 3. Without response to nonsteroidal anti‐inflammatory drugs   Exclusion criteria: 1. Chronic inflammatory arthritis, previous trauma, fracture, glenohumeral osteoarthritis, bone tumours, osteonecrosis or other bone conditions 2. Previous physiotherapy or local glucocorticoid injections in the same shoulder 3. Local glucocorticoid injections in any musculoskeletal location in the three months prior to the study
Interventions	Prior to injection, all patients underwent a sonographic examination of their shoulders. All participants received an injection containing 20 mg of triamcinolone. A single rheumatologist experienced in ultrasound performed all injections in both groups, without knowledge of the clinical evaluation. Ultrasound‐guided injection group (N = 20): Commercial ultrasound equipment was used (Sonoline, Prima, Siemens, Seattle, WA, USA) with a 7.5 MHz linear phased array transducer. The transducer and the patient’s skin were sterilised with alcohol. Sterile gel was applied to the probe. The transducer was held in one hand and the syringe with glucocorticoid in the other hand. The needle was placed under the probe and its route was visualised in real‐time by US, as a hyper‐reflective line, often with reverberation, from the skin to the target. Injection was directed into SA‐SD bursa or biceps tendon sheath when increased fluid was detected. When there was effusion in both the SA‐SD bursa and the biceps tendon sheath, injection was directed into the SA‐SD bursa due to its easier approach. Peri and intralesional injection was performed when rotator cuff calcifications were found. Perilesional injection directed into the SA‐SD bursa was performed when tendon lesions without increased fluid were detected, avoiding direct injection into the rotator cuff tendons. In Group 2, glucocorticoid injection was sonographically guided to the SA‐SD bursa in 14 patients, to biceps tendon sheath in 3 patients, and to rotator cuff calcification in 4 patients. Post‐injection steroid location was confirmed by US in all patients. Anatomic landmark‐guided group (N = 21): A 21‐gauge needle was used. For blind injection, a standard technique was performed. The patient’s skin was sterilised with alcohol. Access to the subacromial space was achieved with a lateral approach, inserting the needle under the anterolateral aspect of the acromion process, passing it through the deltoid muscle, and directing it medially and slightly anterior to the subacromial‐subdeltoid (SA‐SD) bursa, with care taken to avoid injection directly into the tendons of the rotator cuff. Immediately after blind injection, an US examination was performed to search for steroid deposit as hyperechoic foci or lines, with or without acoustic shadowing. Co‐interventions: No patient received physical therapy during the follow‐up period. However, patients with loss of shoulder ROM were instructed to start a home physical therapy programme consisting of pendulum exercises and slow shoulder abduction. No restriction was placed on the patient’s ability to work or to use their shoulder as tolerated, or to take nonsteroidal anti‐inflammatory drugs.
Outcomes	Outcomes were measured at baseline (within 5 days of intervention) and at 6 weeks after the injection. Outcomes: 1. Mean VAS for pain during the previous week (0 to 100, 0 no pain; 100 maximum pain) 2. Mean Shoulder Function Assessment (SFA) scale (includes 2 items on pain during movement and at rest; 4 items for shoulder function in activities of daily living; and 3 objective ROM measures: the active abduction measure (1 point per 10° of abduction, score: 0–18) and 2 semi‐quantitative measures of combined movements. Total score from 0 to 70; higher score indicates better function) 3. Presence of nocturnal pain 4. Intake of nonsteroidal anti‐inflammatory drugs 5. The percentage of patients with a 50% improvement in VAS 6. The percentage of patients with a 50% improvement in SFA 7. Mean active and passive shoulder ROM in flexion, internal rotation and external rotation, measured subjectively 8. Mean active and passive shoulder ROM in abduction, using a goniometer  9. Post‐injection adverse effects (immediate or delayed) 10. Accuracy of injection checked via US imaging No primary outcome was specified. Outcomes included in this review 1. Mean VAS for pain during the previous week 2. Mean Shoulder Function Assessment (SFA) scale 3. The percentage of patients with a 50% improvement in VAS 4. The percentage of patients with a 50% improvement in SFA 5. Post‐injection adverse effects (serious or other; immediate or delayed)
Sources of funding	None reported
Notes	Conflicts of interest: No conflicts of interest statement Trial registration: Not reported Time points included in this review: 5 days and 6 weeks Data analysis: Change scores (mean, SD) were used for pain. VAS scores were transformed from 0 to 100 to 0 to 10 and function scores were transformed from 0 to 70 to 0 to 100. Mean active and passive shoulder ROM were not reported and we could not obtain these data from the authors. Withdrawals: Not reported Adverse events: Not reported Ultrasound‐guided injection group: Serious adverse events: 0/21 Other adverse events: 0/21 Anatomical landmark‐guided injection group: Serious adverse events: 0/20 Other adverse events: 1/20 had a mild post‐injection adverse effect (not specified)
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: "Patients were randomised by a random‐number sequence."
Allocation concealment (selection bias)	Unclear risk	Not reported
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and personnel other than the outcome assessor were not blinded to treatment allocation.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	High risk	Since participants were not blinded, there was potential for detection bias in assessment of self‐reported outcomes.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	Quote: "Each patient was reviewed 6 weeks post injection by the same rheumatologist who performed the initial clinical evaluation (FC), blinded to the injection technique and the sonographic findings."
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	The study provided no information on whether or not any participants were lost to follow‐up.
Selective reporting (reporting bias)	High risk	Mean active and passive shoulder ROM were not reported and we could not obtain these data from the authors.
Other bias	Low risk	None apparent

Raeissadat 2017.

Study characteristics
Methods	Design: Single‐centre, parallel‐group, two‐arm randomised controlled trial in participants with adhesive capsulitis Setting: Shahid Modarres Hospital, Saadat Abad, Tehran, Iran Timing: 2015 Interventions: Ultrasound‐guided or anatomic landmark‐guided glucocorticoid injection into the glenohumeral joint Sample size: sample size was calculated to be 29, however the method was not described, but in order to compensate for dropouts, 41 were included. Analysis: Intention‐to‐treat analysis not specified
Participants	Number of participants: Number of participants screened for eligibility: not reported Number enrolled: 41 (20 in the ultrasound‐guided injection group, 21 in the anatomical landmark‐guided injection group) Number randomised: 41 (20 in the ultrasound‐guided injection group, 21 in the anatomical landmark‐guided injection group) Number included in analyses at each follow‐up: 41 (20 in the ultrasound‐guided injection group, 21 in the anatomical landmark‐guided injection group) Baseline characteristics: Ultrasound‐guided injection group Median age (range): 57.8 (6.0725) years 7 female; 13 male Presence of diabetes: 3 (15.0%) Presence of hypothyroidism: 2 (9.0%) Presence of coronary artery disease: 4 (20.0%) Mean (SD) VAS for pain: 7.6 (1.57) Mean (SD) duration of symptoms: 11 (6.98) months Mean (SD) abduction ROM: 97.8 (24.19) Mean (SD) flexion ROM: 90.35 (19.21) Mean (SD) extension ROM: 46.0 (16.79) Mean (SD) internal rotation ROM: 76.8 (21.33) Mean (SD) external rotation ROM: 54.25 (23.52) Mean (SD) Oxford score: 21.45 (6.84) Anatomical landmark‐guided injection group Median age (range): 59.9 (11.01) years 8 female; 13 male Presence of diabetes: 1 (4.8%) Presence of hypothyroidism: 1 (5.0%) Presence of coronary artery disease: 3 (15.0%) Mean (SD) VAS for pain: 6.66 (2.35) Mean (SD) duration of symptoms: 10.64 (11.38) months Mean (SD) abduction ROM: 93.28 (20.58) Mean (SD) flexion ROM: 102 (23.04) Mean (SD) extension ROM: 55.33 (7.786) Mean (SD) internal rotation ROM: 79.57 (16.32) Mean (SD) external rotation ROM: 52.81 (27.73) Mean (SD) Oxford score: 24.5 (7.31) Inclusion criteria: 1. Diagnosis of adhesive capsulitis on the basis of findings from the history and physical examination 2. Duration of symptoms at least 3 months Exclusion criteria: 1. Previous fracture in the shoulder 2. History of joint inflammatory diseases or any bone disorders 3. History of surgical procedures in the affected shoulder 4. History of shoulder physical therapy or injections in the shoulder in the past 3 months 5. Hypersensitivity reactions to steroids 6. Recurrence of the disease
Interventions	All the injections were performed by one physical medicine and rehabilitation specialist with 15 years’ experience in his field. All participants received an injection of 1 cm3 lidocaine 1%, then, 3 cm3 water soluble un‐ionised contrast with 1 cm3 distilled water, and finally, 1 cm3 triamcinolone 40 mg/cm3 with 1 cm3 lidocaine 1%. Ultrasound‐guided injection (N = 20): The specialist used a Alpinion E‐cube 7 ultrasound device with linear 3–12‐MHz probe while the patient was sitting and the affected hand was resting on his/her thigh. A glenohumeral joint injection was given using the posterior short axis approach, and the needle was inserted in the plane relative to the ultrasound probe and medial to the posterior aspect of the head of the humerus. Palpation‐guided injection (N = 21): A glenohumeral joint injection was given using a posterior approach. A 25‐gauge needle was inserted 2.5 cm lower than the posterolateral aspect of the acromion. If the physician felt that the needle was not accurately inserted, he was not allowed to withdraw and re‐enter again. Co‐interventions: After the injection, all participants received a naproxen tablet 500 mg twice daily for 5 days and were told to perform Codman’s exercises.
Outcomes	All outcome measures were collected before the injection and 1 and 4 weeks after the injection. Outcomes: 1. Mean VAS for pain (0 to 10, 0 no pain; 10 maximum pain) 2. Mean Oxford questionnaire score (0 to 48, higher score indicates better function, 12 items scored from 0‐4) 3. Mean shoulder ROM in flexion, extension, abduction, internal rotation and external rotation, using a goniometer (not specified if active or passive) 4. Adverse events 5. Radiologic assessment of accuracy of injections No primary outcome was specified. Outcomes included in this review 1. Mean VAS for pain 2. Mean Oxford questionnaire score 3. Adverse events 4. Mean shoulder ROM in flexion, extension, abduction internal rotation and external rotation
Sources of funding	Not reported
Notes	No adverse events occurred in either group after the injection. Conflicts of interest: Authors declared no conflicts of interest. Trial registration: Not reported Time points included in this review: 1 week and 4 weeks Data analysis: Authors reported change scores for pain, function and ROM which were used in the meta‐analysis. Withdrawals: None reported Adverse events: None in either group
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "Using block randomisation method, the patients were divided into two equal‐sized groups."
Allocation concealment (selection bias)	Unclear risk	The study did not report if or how sequence allocation was concealed.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and personnel performing injections were not blinded.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	High risk	Since participants were not blinded, there was a high risk of detection bias for self‐reported outcomes.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Unclear risk	Quote: "All the injections were performed by one physical medicine and rehabilitation specialist with 15 years’ experience in his field, different from the physician who measured range of motion of the patients." However, authors have not specified whether the physician who measured range of motion was blinded to treatment allocation or not. The radiologist assessing the post‐injection x‐rays was blinded to treatment allocation.
Incomplete outcome data (attrition bias) All outcomes	Low risk	Quote: "The 41 patients, 20 subjects in the ultrasound‐guided group and 21 in the blind group, completed all the follow‐ups of the survey and cooperated until the end of the study."
Selective reporting (reporting bias)	Low risk	All proposed outcomes are reported in the results.
Other bias	Low risk	None apparent

Roddy 2020.

Study characteristics
Methods	Design: 2 x 2 factorial designed randomised controlled trial in participants with subacromial impingement syndrome Setting: Primary Care Sciences Research Centre (lead centre) and two musculoskeletal Clinical Assessment and Treatment Services in Staffordshire, United Kingdom Timing: 31 May 2011 to 29 November 2012 Interventions: US‐guided injection plus physiotherapist‐led exercise vs US‐guided injection plus advice and exercise leaflet vs unguided injection plus physiotherapist‐led exercise vs unguided injection plus advice and exercise leaflet Sample size calculation: To obtain an effect size (standardised mean difference of 0.4), equating to an approximately 8‐point difference (SD 20) in the SPADI total score for the two main effects (ultrasound‐guided vs unguided injection; physiotherapist‐led exercise vs leaflet), a sample size of 250 participants was needed (80% power; 5% two‐tailed significance; 20% lost to follow‐up). Analysis: Analysis was by intention‐to‐treat.
Participants	Number of participants: Number of participants screened for eligibility: 1275 (not meeting inclusion criteria = 801; declined to participate = 218) Number randomised: 256 (64 to the US‐guided injection plus physiotherapist‐led exercise group; 64 to the US‐guided injection plus advice and exercise leaflet; 64 to the unguided injection plus physiotherapist‐led exercise group; 64 to the unguided injection plus advice and exercise leaflet group) Number included in analysis at each follow‐up: at 6 weeks, 6 months and 12 months (64 in the US‐guided injection plus physiotherapist‐led exercise group; 64 in the US‐guided injection plus advice and exercise leaflet; 64 in the unguided injection plus physiotherapist‐led exercise group; 64 in the unguided injection plus advice and exercise leaflet group) Inclusion criteria: 1. 18 years and over, referred to interface service shoulder clinics with shoulder problems 2. No history of significant shoulder trauma, for example, fracture or full thickness cuff tear 3. A clinical diagnosis of subacromial impingement syndrome i.e. pain in deltoid insertion area, positive Neer and Hawkins Kennedy tests, pain on shoulder abduction Patients were not required to undergo diagnostic imaging (e.g. MRI) to reflect current practice where treatment choices are informed by clinical findings. Exclusion criteria: 1. Below 18 years old 2. Those whose main complaint is due to neck problems, acromioclavicular pathology, or other primary shoulder disorders including adhesive capsulitis or full thickness cuff tear 3. Potentially serious pathology (inflammatory arthritis, polymyalgia rheumatica, malignancy etc.) or ipsilateral shoulder surgery/replacement 4. Those already on a surgical waiting list for shoulder surgery 5. Contraindications to local glucocorticoid injection (known blood coagulation disorders, warfarin therapy) 6. Participation in a shoulder‐focused exercise programme or shoulder injection in the last month 7. Inability to provide written informed consent, complete written questionnaires, or read instruction leaflets written in English Baseline characteristics Ultrasound‐guided injection and physiotherapist‐led exercise (n = 64): Mean (SD) age: 55.6 (10.5) No. (%) female: 37 (57.8) Mean (SD) Shoulder pain interference with work (NRS scale; 0–10): 4.6 (3.0) Mean (SD) SPADI total score: 60.0 (19.2) Mean (SD) SPADI pain subscale: 69.1 (17.3) Mean (SD) SPADI disability subscale: 54.3 (22.4) Mean (SD) Pain severity today (NRS scale; 0–10): 5.6 (2.1) No. (%) both shoulders affected: 7 (10.9) No. (%) Duration of shoulder pain: < 3 months, 6 (9.4); 3‐6 months, 16 (25.0); 6‐12 months, 17 (26.6); > 12 months, 25 (39.1) No. (%) Previous episode of shoulder pain: 16 (25) No. (%) Troubled by shoulder pain in bed most or every night: 48 (75.0) Mean (SD) SF‐12 PCS: 37.3 (8.9) Mean (SD) SF‐12 MCS: 47.9 (12.7) Ultrasound‐guided injection and exercise leaflet (n = 64) Mean (SD) age: 54.8 (10.0) No. (%) female: 29 (45.3) Mean (SD) Shoulder pain interference with work (NRS scale; 0–10): 5.2 (2.8) Mean (SD) SPADI total score: 63.4 (17.6) Mean (SD) SPADI pain subscale: 72.9 (14.8) Mean (SD) SPADI disability subscale: 57.3 (21.1) Mean (SD) Pain severity today (NRS scale; 0–10): 5.9 (2.0) No. (%) both shoulders affected: 3 (4.7) No. (%) Duration of shoulder pain: < 3 months, 5 (7.8); 3‐6 months, 13 (20.3); 6‐12 months, 18 (28.1); > 12 months, 28 (43.8) No. (%) Previous episode of shoulder pain: 24 (37.5) No. (%) Troubled by shoulder pain in bed most or every night: 51 (79.7) Mean (SD) SF‐12 PCS: 38.5 (9.5) Mean (SD) SF‐12 MCS: 46.8 (13.3) Unguided injection and physiotherapist‐led exercise (n = 64): Mean (SD) age: 51.9 (10.7) No. (%) female: 45 (70.3) Mean (SD) Shoulder pain interference with work (NRS scale; 0–10): 4.2 (3.0) Mean (SD) SPADI total score: 62.4 (16.5) Mean (SD) SPADI pain subscale: 70.9 (15.6) Mean (SD) SPADI disability subscale: 57.1 (19.3) Mean (SD) Pain severity today (NRS scale; 0–10): 5.4 (1.9) No. (%) both shoulders affected: 5 (7.8) No. (%) Duration of shoulder pain: < 3 months, 8 (12.5); 3‐6 months, 9 (14.0); 6‐12 months, 26 (40.6); > 12 months, 21 (32.8) No. (%) Previous episode of shoulder pain: 17 (26.6) No. (%) Troubled by shoulder pain in bed most or every night: 51 (79.7) Mean (SD) SF‐12 PCS: 38.1 (10.9) Mean (SD) SF‐12 MCS: 45.8 (12.4) Unguided injection and exercise leaflet (n = 64): Mean (SD) age: 53.0 (9.5) No. (%) female: 22 (34.4) Mean (SD) Shoulder pain interference with work (NRS scale; 0–10): 4.9 (2.9) Mean (SD) SPADI total score: 58.4 (18.9) Mean (SD) SPADI pain subscale: 69.4 (15.5) Mean (SD) SPADI disability subscale: 51.6 (22.9) Mean (SD) Pain severity today (NRS scale; 0–10): 5.8 (2.0) No. (%) both shoulders affected: 5 (7.8) No. (%) Duration of shoulder pain: < 3 months, 9 (14.0); 3‐6 months, 12 (18.8); 6‐12 months, 13 (20.3); > 12 months, 30 (46.9) No. (%) Previous episode of shoulder pain: 30 (46.9) No. (%) Troubled by shoulder pain in bed most or every night: 55 (85.9) Mean (SD) SF‐12 PCS: 38.5 (9.8) Mean (SD) SF‐12 MCS: 48.1 (12.2) Pretreatment group differences: None apparent
Interventions	Intervention groups: (1) Ultrasound‐guided subacromial glucocorticoidinjection and physiotherapist‐led exercise: Injection procedure: Ultrasound‐guided injections were performed by one of nine clinicians using the LOGIQ e system with a 12 MHz transducer. Clinicians either had extensive clinical experience performing ultrasound‐guided injections or completed an accredited course on ultrasound‐guided subacromial injections; all attended a half‐day injection protocol workshop and passed a clinical competency test by a consultant musculoskeletal sonographer. The skin and transducer were cleaned with chlorhexidine 0.5% solution and sterile gel applied to the transducer. The participant sat with the shoulder internally rotated and the ipsilateral hand on the buttock to maximise visibility of and access to the subacromial bursa. The transducer was placed anterolaterally, the hypoechoic subacromial bursa visualised, and a 21 G needle inserted under real‐time ultrasound guidance until the needle‐tip entered the bursa. A commercially available premixed solution of methylprednisolone 40 mg and 1 mL 1% lidocaine was injected into the bursa. Exercise procedure: The protocol for the physiotherapist‐led exercise programme can be found at: https://www.keele.ac.uk/media/keeleuniversity/ri/primarycare/docs/SUPPORT_Physiotherapy_Intervention_Manual_v3.0_04_01_11_Internet_Version.pdf. The exercise programme was delivered by 20 community‐based, musculoskeletal physiotherapists who completed a 2‐day training workshop about the exercise treatment protocol. It commenced within 3 weeks of recruitment and injection. Exercise type and dose were individualised, supervised and progressed in 6–8 sessions over a period of 12–16 weeks. Exercises were progressed through three stages guided by an individualised written exercise sheet and a computerised package, aiming to support the patient back to their specific everyday physical, sporting and occupational activities: 1. Scapular stability exercise and active movement without resistance 2. Range of movement exercises, isometrics and stretches, with scapular control in pain‐free range 3. Through‐range resistance exercises, progressed to encourage rotator cuff muscle strengthening through all ranges of movement (2) Ultrasound‐guided subacromial glucocorticoid injection and exercise leaflet: The ultrasound‐guided injection was performed as described above. Advice and exercise leaflet: The leaflet was provided following glucocorticoid injection and included information about shoulder anatomy and SAPS; simple self‐help messages about analgesia, cold packs and activities; and six standardised specific strengthening and range of motion exercises to be performed 2–3 times daily, with no instructions for individualisation or progression. (3) Unguided subacromial glucocorticoid injection and physiotherapist‐led exercise: Unguided subacromial injections were performed by one of eight clinicians, different to those performing ultrasound‐guided injections. Clinicians had extensive clinical experience performing subacromial injections, and attended a half‐day injection protocol workshop. The participant sat with their arm hanging with the elbow bent and forearm resting on their lap. The skin was cleaned with chlorhexidine solution 0.5%. A 21 G needle was inserted through the deltoid under the acromion process laterally. The same premixed solution of methylprednisolone 40 mg and 1 mL 1% lidocaine was injected. Exercise was delivered as described above. (4) Unguided subacromial glucocorticoid injection and exercise leaflet: Procedure for the injection as described above. The content of the advice and exercise leaflet is as described above. Post‐intervention: Following ultrasound‐guided or unguided injection, participants were advised not to drive immediately after injection and to avoid pushing/pulling movements with the affected arm and heavy/repetitive tasks for 2 weeks. A second injection as per treatment allocation was permitted at the treating clinician’s discretion.
Outcomes	Outcomes were assessed at baseline, 6 weeks, 6 months and 12 months. Follow‐up rates were 94% at 6 weeks, 88% at 6 months and 80% at 12 months. Primary outcome: 1. Mean (SD) Shoulder Pain and Disability Index (SPADI) total scores (0 to 100, higher score indicates worse pain and disability. Included five questions on Pain (0 = no pain; 10 = the worst pain imaginable) and eight on disability (0 = no difficulty; 10 = so difficult it requires help)) Secondary outcomes: Current shoulder pain intensity (0‐10 NRS) 0, no pain SPADI pain scores: SPADI subscales, ranged from 0 to 100; 0 = no pain, 100 = worst pain SPADI disability scores: SPADI subscales, ranged from 0 to 100; 0 = no difficulty, 100 = so difficult it required help Patient’s self‐reported global impression of change Emotional representation: How much does your shoulder problem affect you emotionally? 0–10 NRS, 0 = not at all affected emotionally, 10 = extremely affected emotionally SF‐12 (PCS and MCS). Scales are based on a ‘normalised’ general population average of 50 with SD of 10 Shoulder pain at night (troubled most or every night) Pain self‐efficacy: 10‐item scale, score range = 0–60 (0 = not at all confident, 60 = completely confident) Fear of movement: Tampa scale for kinesiophobia‐11–score range from 11 to 44, with higher scores reflecting greater fear of movement or (re)injury Work performance Work presenteeism Time off work Healthcare utilisation (medications, visits to healthcare practitioners, further injections or physiotherapy) Treatment preferences and expectations Illness perceptions Exercise adherence Satisfaction with treatment (treatment success) Adverse events Serious adverse events Outcomes included in this review: Pain intensity of current shoulder pain (0‐10 NRS, 0 no pain) Function on SPADI scores QOL on SF‐12 MCS scores Treatment success Number of participants experiencing any adverse events Number of participants experiencing serious adverse events
Sources of funding	The trial was organised and sponsored by the Arthritis Research UK Primary Care Centre, at Keele University, and funded by the National Institute for Health Research: Research for Patient Benefit Programme
Notes	Trial registration: ISRCTN42399123 (prospectively registered) (http://www.isrctn.com/ISRCTN42399123) Withdrawals: At 6 weeks, 6/64 (lost to follow‐up = 4, withdrawn = 2) in US‐guided injection + physiotherapist‐led exercise group; 6/64: (lost to follow‐up = 6) in US‐guided injection + leaflet group; 4/64: (lost to follow‐up = 3, withdrawn = 1) in unguided injection + physiotherapist‐led exercise group; 0/64: in unguided injection + leaflet group. At 6 months, 7/64 (lost to follow‐up = 4, withdrawn = 3) in US‐guided injection + physiotherapist‐led exercise group; 8/64 (lost to follow‐up = 6, withdrawn = 2) in US‐guided injection + leaflet group; 8/64 (lost to follow‐up = 5, withdrawn = 3) in unguided injection + physiotherapist‐led exercise group; 8/64 (lost to follow‐up = 5, withdrawn = 3) in unguided injection + leaflet group. At 12 months, 15/64 (lost to follow‐up = 9, withdrawn = 6) in US‐guided injection + physiotherapist‐led exercise group; 12/64 (lost to follow‐up = 7, withdrawn = 5) in US‐guided injection + leaflet group; 10/64 (lost to follow‐up = 6, withdrawn = 4) in unguided injection + physiotherapist‐led exercise group; 13/64 (lost to follow‐up = 8, withdrawn = 5) in unguided injection + leaflet group Adverse events: Ultrasound‐guided injection and physiotherapist‐led exercise (n = 64): Serious adverse event: No. and nature of event: 1/64 ‐ pyelonephritis Other adverse event: No. and nature of event: 17/64 ‐ shoulder pain > 3 days following injection No. and nature of event: 16/64 ‐ discomfort during the injection or local skin changes, presyncope, nausea or flushing following the injection Ultrasound‐guided injection and exercise leaflet (n = 64): Other adverse event: No. and nature of event: 17/64 ‐ shoulder pain > 3 days following injection No. and nature of event: 16/64 ‐ discomfort during the injection or local skin changes, presyncope, nausea or flushing following the injection Unguided injection and physiotherapist‐led exercise (n = 64): Other adverse event: No. and nature of event: 20/64 ‐ shoulder pain > 3 days following injection No. and nature of event: 17/64 ‐ discomfort during the injection or local skin changes, presyncope, nausea or flushing following the injection Unguided injection and exercise leaflet (n = 64): Other adverse event: No. and nature of event: 20/64 ‐ shoulder pain > 3 days following injection No. and nature of event: 17/64 ‐ discomfort during the injection or local skin changes, presyncope, nausea or flushing following the injection
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	"Trial administrator in the clinic arranged randomisation through Keele University Clinical Trial Unit’s telephone randomisation service". it was unclear how the random sequence was generated.
Allocation concealment (selection bias)	Unclear risk	Unclear how the allocation to groups was concealed from the study participants
Blinding of participants and personnel (performance bias) All outcomes	High risk	"Participants and clinicians were aware of treatment allocation." Hence, there was high risk of performance bias.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	High risk	As participants were not blinded; there was high risk of bias in the assessment of self‐reported outcomes.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	"The research nurse remained blind to allocation." No assessor‐reported outcomes used in this review
Incomplete outcome data (attrition bias) All outcomes	Low risk	Although there were withdrawals (numbers and reasons across groups similar), ITT analysis (missing at random (MAR) assumption ‐ the missing data independent of the unobserved measurement) data from all 256 participants (64 per group) were analysed.
Selective reporting (reporting bias)	Unclear risk	Trial was registered and study protocol published. Authors reported "assessed only self‐reported outcomes but not the effect of exercise on shoulder strength, scapular stability or movement, nor can we determine which aspects of the physiotherapist‐led exercise programme contributed most to the benefit observed."
Other bias	Low risk	None apparent

Saeed 2014.

Study characteristics
Methods	Design: Single‐centre, parallel‐group, two‐arm randomised controlled trial in participants with 'shoulder impingement pain' Setting: Rheumatology outpatient clinic in Ireland Timing: Not reported. Interventions: Ultrasound‐guided ('musculoskeletal ultrasonography‐guided') steroid injection or palpation‐guided glucocorticoid injection directed to the main ultrasound pathology or subacromial‐subdeltoid bursa Sample size: Not reported Analysis: Intention‐to‐treat analysis not performed. Quote: "Twenty patients, 11 (20 shoulders) from the anatomical landmark‐guided injection group and 9 (15 shoulders) from the ultrasound‐guided group, were excluded from the study after 6 weeks. These patients showed little or no clinical improvement after the first shoulder injection and required either a repeat shoulder injection or further imaging and surgical opinion based on their symptoms."
Participants	Number of participants: Number of participants screened for eligibility: not reported Number enrolled: 100 participants; 125 shoulders (50 per group, 59 shoulders in ultrasound‐guided injection group and 66 shoulders in the anatomical landmark‐guided injection group) Number randomised: 100 participants; 125 shoulders (50 per group, 59 shoulders in ultrasound‐guided injection group and 66 shoulders in anatomical landmark‐guided injection group) Number included in analyses at each follow‐up: 80 participants; 90 shoulders at 12 weeks (41 participants and 44 shoulders in ultrasound‐guided injection group; 39 participants and 46 shoulders in the anatomical landmark‐guided injection group). All participants that did not complete the 12‐week follow‐up required a repeat injection or surgical referral. Baseline characteristics: Ultrasound‐guided group: Mean (SD) duration of symptoms: 19.64 (1.84) weeks Mean (SD) VAS for pain: 6.34 (1.74) Mean number of positive Shoulder Function Test (SFT): 1.81 Mean (SD) Physician Global Assessment (PGA) score: 5.5 (2.07) Ultrasound findings: biceps pathology (tendinopathy/tear), 10 (17%); rotator cuff impingement, 15 (25%); rotator cuff partial‐thickness tear, 14 (24%); rotator cuff full‐thickness tear, 13 (22%); subacromial/subdeltoid bursitis, 9 (15%); acromioclavicular joint pathology 7 (12%) Anatomical landmark‐guided injection group: Mean (SD) duration of symptoms: 20.02 (1.52) weeks Mean (SD) VAS for pain: 6.39 (1.83) Mean number of positive Shoulder Function Test (SFT): 1.5 Mean (SD) Physician Global Assessment (PGA) score: 5.48 (1.94) Ultrasound findings: biceps pathology (tendinopathy/tear), 6 (9%); rotator cuff impingement, 14 (21%); rotator cuff partial‐thickness tear, 17 (26%); rotator cuff full‐thickness tear, 13 (20%); subacromial/subdeltoid bursitis, 8 (12%); acromioclavicular joint pathology 8 (12%) Inclusion criteria: 1. Shoulder pain for at least 3 months duration with minimal or no response to nonsteroidal anti‐inflammatory drugs (NSAIDs) Exclusion Criteria: 1. Fracture, glenohumeral osteoarthritis, chronic inflammatory arthritis, bone tumours, osteonecrosis and other bone conditions 2. Patients who had clinical and radiological findings indicating moderate osteoarthritis, referred pain from the neck or internal organs and generalised muscular pain syndrome with bilateral muscular pain in the neck and shoulders, a history of inflammatory arthritis, previous fractures or surgery to the shoulder, or contraindications to local steroid injections 3. Patients who had been treated with local glucocorticoid injections and/or physiotherapy within 1 month of study initiation
Interventions	All participants received an injection of 40 mg of methylprednisolone acetate with 4 mL of lidocaine hydrochloride. All ultrasound‐guided and palpation‐guided injections were performed by the same experienced physician. Ultrasound‐guided injection (N = 50 participants; 59 shoulders): Acuson Sequoia 512 ultrasound systems (Siemens, CA, USA) using 8L‐RS MHz linear phased‐array transducer were used. A two‐handed technique was used with the transducer held in one hand and the syringe with 21G needle in the other hand. The needle was directed in real time by ultrasound from the skin to the target (e.g. subdeltoid bursa, acromio‐clavicular joint or biceps pathology). In those shoulders where ultrasound demonstrated more than one pathology, the pathology most consistent with clinical examination was injected. When there was effusion in both the subacromial‐subdeltoid bursa and the biceps tendon sheath, injection was directed into the subacromial‐subdeltoid bursa. For confirmed rotator cuff pathology and inpatients with clinical impingement but normal shoulder, ultrasound subacromial‐subdeltoid bursa injection was performed. When ultrasound confirmed the principal pathology as acromioclavicular joint inflammation or biceps tendon inflammation, ultrasound‐guided injection of the principal pathological structure was performed. Palpation‐guided injection (N = 50 participants; 66 shoulders): A standard technique was performed using a 21G needle via a lateral approach to inject the subacromial‐subdeltoid bursa. Co‐interventions: No participant received physical therapy during the follow‐up period. However, all participants with the loss of shoulder range of movement were given post‐injection instructions in pendulum exercises and slow shoulder abduction. No restriction was placed on participants' ability to work or to use their shoulder as tolerated, or intake of NSAIDs.
Outcomes	Outcomes were assessed at baseline and at 6 and 12 weeks post‐injection. Outcomes 1. Mean VAS for pain (0 to 10, 0 no pain; 10, maximum pain) 2. Mean Shoulder Function Tests (SFTs) score (tenderness on Hawkins–Kennedy test and supraspinatus tendon tenderness; 1 positive test = 1 score; total score from 0 to 2, higher score indicates worse shoulder function) 3. Mean active and passive shoulder ROM (method and movements not specified) 4. Mean Physician Global Assessment (PGA) score (0 to 10 cm, higher scores indicate worse function) 5. Any adverse outcomes of injections Outcomes included in this review 1. Mean VAS for pain 2. Any adverse outcomes of injections
Sources of funding	None reported
Notes	Conflicts of interest: No conflicts of interest statement Trial registration: Not reported Time points included in this review: 6 weeks and 12 weeks Data analysis: Mean Shoulder Function Tests (SFTs) scores were not included in the data analysis as we were unable to obtain SDs from the author or use the SDs from another trial (as none used this measure of function). Mean active and passive shoulder ROM were not included in the data analysis as they were not reported in the paper and we were unable to obtain the results from the author. The shoulder was the unit of analysis. Withdrawals: 20 patients including 35 shoulders (9 in ultrasound‐guided injection group including 15 shoulders, 11 in anatomical landmark‐guided injection group including 20 shoulders) at 12 weeks Adverse events: Ultrasound‐guided injection group: Serious adverse events: 0/44 Anatomical landmark‐guided injection group: Serious adverse events: 0/46 Other adverse events: Apart from mild shoulder pain in a small number of cases (group not specified) within a few hours of shoulder injection, no serious adverse event was observed.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "A computer‐generated randomization scheme with permitted block sizes of 2 and 4 was implemented."
Allocation concealment (selection bias)	Unclear risk	The study did not report if or how sequence allocation was concealed.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and personnel other than the outcome assessor were not blinded.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	High risk	Since participants were not blinded, there was a high risk of detection bias for self‐reported outcomes.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	Quote: "A single consultant rheumatologist ‐ blinded to the results of ultrasound and to the treatment received by the patient ‐ performed the shoulder evaluations at baseline and at 6 and 12 weeks post‐injection."
Incomplete outcome data (attrition bias) All outcomes	Low risk	11/50 in the anatomical landmark‐guided injection group and 9/50 in the ultrasound‐guided injection group did not complete the 12‐week assessment and were excluded from the study after the 6‐week assessments. All dropouts were due to the need for a repeat injection or referral for surgery.
Selective reporting (reporting bias)	High risk	Shoulder ROM was measured but not reported in the results. Function could not be included in the data analysis as we were unable to obtain SDs from authors.
Other bias	Low risk	None apparent

Ucuncu 2009.

Study characteristics
Methods	Design: Single‐centre, parallel‐group, two‐arm randomised controlled trial in participants with 'soft tissue' shoulder lesions (acromioclavicular degeneration, rotator cuff lesions (rupture, partial rupture, tendinosis, impingement, calcification), fluid accumulation in the biceps tendon, partial rupture in the biceps tendon and bursitis (subdeltoid, subacromial) Setting: Outpatient clinic, Department of Physical Medicine and Rehabilitation, Farabi Hospital, Karadeniz Technical University, Trabzon,Turkey Timing: Not reported Interventions: Ultrasound‐guided steroid injection or anatomic landmark‐guided injection into the subacromial region Sample size: Not reported Analysis: Intention‐to‐treat analysis not specified
Participants	Number of participants: Number of participants screened for eligibility: not reported Number enrolled: 60 (30 per group) Number randomised: 60 (30 per group) Number included in analyses at each follow‐up: not reported Baseline characteristics: Ultrasound‐guided injection group: Mean (SD) age: 52.1 (11.6) years 22 female; 8 male Mean (SD) duration of symptoms: 10.7 (12.5) months Active ROM decrease: 24 (80%) Mean (SD) VAS score for pain: 6.3(1.8) Mean (SD) Constant score: 56.7 (21.6) Landmark‐guided injection group: Mean (SD) age: 52.9 (9.7) years 22 female; 8 male Mean (SD) duration of symptoms: 9.6 (8.7) months Active ROM decrease: 18 (60%) Mean (SD) VAS score for pain: 6.0 (1.4) Mean (SD) Constant score: 70 (16.4) Inclusion criteria: 1. Shoulder pain for at least 1 month and not satisfactorily responding to at least 1 month of nonsteroidal anti‐inflammatory drugs Exclusion criteria: 1. Fracture, glenohumeral osteoarthritis, characteristic findings of chronic inflammatory arthritis, bone tumours, osteonecrosis, and other bone conditions 2. Diabetes mellitus 3. Previous major trauma in shoulder area 4. Pain in both shoulders 5. Previous physical therapy or local steroid injections
Interventions	All participants were examined with ultrasound, performed by a physician experienced in ultrasound without knowledge of the clinical evaluation. All participants received a single injection of 1 mL 40 mg triamcinolone and 1 mL 1% lidocaine.  Ultrasound‐guided injection (N = 30): The injector was directed to the site of observed pathology beneath the probe by imaging a hyper‐reflective line. Injectates were administered perilesionally and intralesionally using commercial, real‐time, equipment (Esaote, Mylab 60, Italy) using a 6 to 18 MHz linear phased array transducer. Landmark‐guided injection (N = 30): Blind injection was administered with a lateral entry approach to the sub‐acromial region.  Co‐interventions: Participants did not receive physical therapy during the follow‐up period. Participants with a loss of shoulder ROM performed a home exercise programme consisting of shoulder abduction and pendulum exercises. No limit was imposed on nonsteroidal anti‐inflammatory consumption or use of the shoulder.
Outcomes	Outcomes were measured at baseline and 6 weeks following injection.  Outcomes: 1. Mean VAS for pain (0 to 10, 0 no pain; 10 maximum pain) 2. Mean Constant score (0 to 100, higher score indicates better function. Domains: pain, 15 points; activities of daily living, 20 points: range of motion, 40 points; strength, 25 points) 3. Mean active and passive shoulder ROM in flexion and abduction, using a goniometer 4. Adverse events 5. Treatment success measured as decrease of > 50% in VAS pain No primary outcome was specified. Outcomes included in this review 1. Mean VAS for pain 2. Mean Constant score 3. Adverse events 4. Mean active shoulder ROM in flexion and abduction 5. Treatment success measured as decrease of > 50% in VAS pain
Sources of funding	None reported
Notes	Conflicts of interest: No conflicts of interest statement Trial registration: Not reported Time points included in this review: 6 weeks Withdrawals: None Adverse events: Ultrasound‐guided injection group: Serious adverse events: 0/30 Other adverse events: 1/30 had mild post‐injection pain. Anatomical landmark‐guidedinjection group: Serious adverse events: 0/30 Other adverse events: 6/30 5/30 had a slight increase in pain and 1/30 had skin peeling after the injection.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Unclear risk	Quote: “This prospective randomised study allocated consecutive patients, stratified by sex to either LMG group or USG group.” Method of random sequence generation not specified
Allocation concealment (selection bias)	Unclear risk	The study did not report if or how sequence allocation was concealed.
Blinding of participants and personnel (performance bias) All outcomes	High risk	Participants and personnel were not blinded.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	High risk	Since participants were not blinded, there was a high risk of detection bias for self‐reported outcomes.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Unclear risk	The study did not provide details about whether the person who performed the outcome assessment was blinded to treatment allocation or not.
Incomplete outcome data (attrition bias) All outcomes	Low risk	All randomised participants completed the 6‐week follow‐up.
Selective reporting (reporting bias)	Low risk	The results for all prespecified outcomes were reported.
Other bias	Low risk	None apparent

Zufferey 2012.

Study characteristics
Methods	Design: Multicentre, parallel‐group, two‐arm randomised controlled trial in participants with acute shoulder pain Setting: Three Swiss rheumatology centres, Switzerland Timing: December 2008 to May 2010 Interventions: Ultrasound‐guided subacromial glucocorticoid injection (according to ultrasound diagnosis) or anatomical landmark‐guided subacromial glucocorticoid injection Sample size: The power calculation of the number of patients needed to show statistical significance between the two groups was based on previous published studies and results of a pilot study in one of the study centres. However, the exact number of participants in each group was not reported. Analysis: Intention‐to‐treat analysis not specified
Participants	Number of participants: Number of participants screened for eligibility: 143 Number enrolled: 70 (34 in the ultrasound‐guided injection group; 36 in the anatomical landmark‐guided injection group) Number randomised: 70 (34 in the ultrasound‐guided injection group; 36 in the anatomical landmark‐guided injection group) Number included in analyses at each follow‐up: 67 (32 in the ultrasound‐guided group; 35 in the anatomical landmark‐guided injection group) at 2 and 6 weeks, and 56 (27 in the ultrasound‐guided injection group; 29 in the anatomical landmark‐guided injection group) at 12 weeks Baseline data were only available for 65 participants (32 in the ultrasound‐guided injection group; 33 in the anatomical landmark‐guided injection group). Baseline characteristics: Ultrasound‐guided injection group: Mean (SD) age: 53 (11) years 13 female; 19 male Mean duration of shoulder pain: 20 weeks Mean duration of night pain: 11 weeks No. with mobility impairment: 28 Mean (SD) NRS for pain at rest: 6.7 (1.9) Mean (SD) NRS for pain during activity: 5.1 (1.4) Mean (SD) Constant score (maximum 75 points due to the exclusion of the 'strength' item): 39 (12) Ultrasound diagnosis: isolated tendonitis (n = 5); partial rupture (n = 11); complete rupture (n = 9); bursitis (n = 16); fluid or synovitis (n = 23); calcifications (n = 6); capsulitis (n = 6) Anatomical landmark‐guided injection group: Mean (SD) age: 54 (10) years 15 female; 18 male Mean duration of shoulder pain: 27 weeks Mean duration of night pain: 17 weeks No. with mobility impairment: 25 Mean (SD) NRS for pain at rest: 6.6 (2.3) Mean (SD) NRS for pain during activity: 5.3 (2.0) Mean (SD) Constant score (maximum 75 points due to the exclusion of the 'strength' domain): 40 (14) Ultrasound diagnosis: isolated tendonitis (n = 5); partial rupture (n = 11); complete rupture (n = 5); bursitis (n = 13); fluid or synovitis (n = 19); calcifications (n = 8); capsulitis (n = 6) Inclusion criteria: 1. Patients over age of 18 with shoulder pain that did not respond to nonsteroidal anti‐inflammatory drugs or physiotherapy Exclusion criteria: 1. History of inflammatory arthritis or radiological glenohumeral osteoarthritis 2. Previous local steroid injection within 12 weeks
Interventions	Shoulder ultrasound was performed with a 5‐9 MHz probe using a Philips HD11 machine in two centres and Easote mylab25 gold in one, by two rheumatologists experienced in musculoskeletal ultrasonography. All patients received 2 mL of Diprophos® (7 mg of a mixture made up of one third soluble and two‐third long‐acting bethamethasone). Ultrasound‐guided injection group (N = 34): Steroid injection under ultrasound guidance was performed according to the ultrasound diagnosis within two days of the clinical evaluation. The injection was directed into the location assessed by ultrasound to be the cause of shoulder pain. Anatomical landmark‐guided injection group (N = 36): Blind injections were directed at the subacromial bursa. Co‐interventions: Rescue analgesia was permitted and participants recorded their intake of nonsteroidal anti‐inflammatory drugs and paracetamol until the end of the study. New infiltrations after two weeks were tolerated but considered a poor result. In those cases, the succeeding NRS at 6 and 12 weeks were assimilated to the initial pre‐study NRS for calculation.
Outcomes	Outcomes were assessed at baseline, 2 weeks, 6 weeks and 12 weeks. The primary outcome measures were pain at 2 and 6 weeks and function at 6 weeks. Primary outcomes: 1. Mean Numeric Rating Scale (NRS) for pain at rest and during activity (0 to 10, 0 no pain; 10 maximum pain) 2. Mean Constant score (0 to 75 since the 'strength' domain was excluded for technical reasons, higher scores indicate better function. Domains: pain, 15 points; activities of daily living, 20 points, range of motion, 40 points) 3. Percentage of participants with greater or equal to 50% of improvement for pain at rest and during activity Secondary outcomes: 4. Mean active and passive shoulder ROM in abduction, external rotation and internal rotation (method not specified) 5. Use of analgesics 6. Proportion that received additional injections 7. Adverse events (e.g. progressing rotator cuff tear) Outcomes included in this review 1. Mean Numeric Rating Scale (NRS) for pain at rest 2. Mean Constant score 3. Percentage of participants with greater or equal to 50% of improvement for pain at rest 4. Proportion that received additional injections 5. Mean active shoulder ROM in abduction and external rotation 6. Adverse events (serious and other)
Sources of funding	Not reported
Notes	Conflicts of interest: Authors declared no conflicts of interest. Trial registration: Not reported Time points included in this review: 5 days, 3 weeks, 6 weeks and 3 months Data analysis: The authors provided adverse event data but did not respond when we asked for the SDs for pain at 12 weeks, function and ROM. We used baseline SDs at 12 weeks for pain, we used SDs from the following studies for function (Bhayana 2018), ROM in abduction (Ekeberg 2009) and ROM in external and internal rotation (Raeissadat 2017). Withdrawals: 3 (2 in ultrasound‐guided injection group out of which one was a dropout and one had a re‐injection, 1 in the anatomical landmark‐guided injection group due to re‐injection) at 2 and 6 weeks, and 14 (7 in the ultrasound‐guided injection group, 7 in the anatomical landmark‐guided injection group) at 12 weeks, reasons for withdrawal not provided Adverse events: Ultrasound‐guided injection group: Serious adverse events: 0/32 Other adverse events: 3/32 required a second local injection after 2 or 6 weeks Anatomical landmark‐guided injection group: Serious adverse events: 0/35 Other adverse events: 3/35 required a second local injection after 2 or 6 weeks Four patients reported flushing and one diabetic patient a transient hyperglycaemia, however, there were no details of which groups these participants belonged to.
Risk of bias
Bias	Authors' judgement	Support for judgement
Random sequence generation (selection bias)	Low risk	Quote: "All patients underwent shoulder US, and were randomised to either US‐guided local steroid injection or blind injection." "Patients signed an informed consent and were assigned treatment groups by random‐number sequence'.
Allocation concealment (selection bias)	Unclear risk	The study did not report if or how sequence allocation was concealed.
Blinding of participants and personnel (performance bias) All outcomes	High risk	The authors did not report whether or not participants were blinded to treatment allocation. Personnel performing the injections were not blinded. Two rheumatologists performed the ultrasound‐guided and anatomical landmark‐guided injections. If the expertise of these rheumatologists differed and the distribution of ultrasound‐guided and anatomical landmark‐guided injections was uneven, this could have biased the treatment effect estimates.
Blinding of outcome assessment (detection bias) Self‐reported outcomes (pain, function, success, quality of life, adverse events)	High risk	As it is not known if participants were blinded, there was a high risk of detection bias for self‐reported outcomes.
Blinding of outcome assessment (detection bias) Assessor‐reported outcomes (range of motion)	Low risk	Quote: "Follow‐up evaluation was performed by a rheumatologist who was blinded to the results of the initial US and clinical assessments as well as the route of steroid administration."
Incomplete outcome data (attrition bias) All outcomes	Unclear risk	1/36 in the anatomical landmark‐guided injection group and 2/34 in the ultrasound‐guided injection group did not complete the 2 and 6‐week follow‐up. 7/36 in the anatomical landmark‐guided injection group and 7/34 in the ultrasound‐guided injection group did not complete the 12‐week follow‐up. Reasons for withdrawals at 12 weeks were not provided.
Selective reporting (reporting bias)	Unclear risk	Authors did not provide SDs for all outcomes; function data were not reported for 12 weeks. Data for non‐responders were not presented clearly. The trial was not registered and no study protocol was found.
Other bias	Low risk	None apparent

US‐guided = ultrasound‐guided; SIS = subacromial impingement syndrome; ROM = range of motion; NRS = numerical rating scale; VAS = visual analog scale; SD = standard deviation; T6 = sixth thoracic vertebrae; BREF/WHOQOL‐BREF = World Health Organization Quality of Life Brief Version; UCLA = University of California Los Angeles; SEM = standard error of the mean; IQR = interquartile range; MHz = megahertz; BMI = body mass index; SF‐12 PCS = Short Form Health Survey Physical Component Score; SF‐12 MCS = Short Form Health Survey Mental Component Score; QOL = quality of life; ITT = intention to treat; NSAIDs = non‐steroidal anti‐inflammatory drugs.

Characteristics of excluded studies [ordered by study ID]

Study	Reason for exclusion
Alfredson 2006	Not a randomised controlled trial
Bamji 2004	Not a randomised controlled trial
Buchbinder 2004	Not correct intervention
Chavez‐Lopez 2009	Not correct comparator
Cohen 2009	Not correct participants
Esenyel 2003	Not a randomised controlled trial
Graber 1997	Not correct intervention
Gutierrez 2004	Not a randomised controlled trial
Hashiuchi 2010	Not a randomised controlled trial
Henkus 2006	Not correct intervention
Kang 2008	Not correct intervention
Kim 2016	Not correct comparator
Kim 2017	Not correct comparator
Koes 2009	Not a randomised controlled trial
McCormack 2009	Not a randomised controlled trial
Micu 2013	Not a randomised controlled trial
Rutten 2007	Measured only accuracy of needle placement ‐ which is an exclusion criterion for this review.
Sibbitt 2009	Not correct participants
Tveita 2008	Not correct comparator
Valtonen 1978	Data were not presented separately for participants who were randomly allocated to either subacromial glucocorticoid injection, intramuscular gluteal steroid injection or intramuscular gluteal placebo injection.
Widiastuti‐Samekto 2004	Not correct intervention
Yi 2006	Not a randomised controlled trial
Zhang 2011	Not correct participants

Characteristics of studies awaiting classification [ordered by study ID]

Cinar 2018.

Methods	Design: Parallel‐design, two‐arm, randomised controlled trial in participants with subacromial impingement syndrome Setting: Not reported Timing: Not reported Interventions: Steroid injection into the subacromial space using ultrasound guidance or anatomic landmarks Sample size: 40 Analysis: Not specified
Participants	Patients with subacromial impingement syndrome
Interventions	Ultrasound‐guided injection (N = 21): Method not reported Anatomical landmark‐guided injection (N = 19): Method not reported
Outcomes	Outcomes were measures before and four weeks after the injection. Outcomes 1. Mean (SD) Shoulder Pain and Disability Index (SPADI) (0 to 100, higher score indicates worse pain and disability. Included five questions on Pain (0 = no pain; 10 = the worst pain imaginable) and eight on disability (0 = no difficulty; 10 = so difficult it requires help)) 2. Mean (SD) Quick Disabilities of the Arm, Shoulder, and Hand (Q‐DASH) (0 to 100, higher score indicates worse disability. Included 11 items, scored 1‐5, that are transformed to a 0 to 100 scale) 3. Mean (SD) visual analog scale (VAS) for pain at rest 4. Mean (SD) VAS for pain during movement 5. Mean (SD) shoulder range of flexion and external rotation
Notes	Conference abstract; attempts to contact the authors were unsuccessful.

IRCT2017021524621N6.

Methods	Design: Single‐centre, parallel‐group, two‐arm randomised controlled trial in participants with shoulder tendinopathy or impingement syndrome Setting: Firoozgar Educational Hospital, Tehran, Iran Timing: October 2012–December 2013 Interventions: Ultrasound‐guided injection of 2 mL (40 mg/mL) triamcinolone and 2 cc of lidocaine into the subacromial space or anatomical landmark‐guided injection of 2 mL (40 mg/mL) triamcinolone and 2 cc of lidocaine into the subacromial space Sample size: 30 Analysis: Not specified
Participants	Inclusion criteria: 1. Age between 20 and 80 years 2. Shoulder tendinopathy as positive result of specific test like Neer & Hawkins test 3. Willingness to participate in research projects 4. No history of rotator cuff tears, diabetes, coagulation disorders, fractures, rheumatologic diseases, tumours and infections 5. No shoulder injections in last 8 weeks 6. The absence of intra‐articular injection contraindications like use of anticoagulant medications 7. The absence of glucocorticoid and lidocaine injection contraindications such as allergic reactions to these medications and skin infection at the injection site 8. The absence of bilateral shoulder problems 9. Visual analog scale > 4 10. Not pregnant or not deciding to get pregnant Exclusion criteria: 1. Willingness to exit the study at any stage of the project 2. Not attending follow‐up evaluations during the study 3. Daily use of any type of pain medication during the study
Interventions	All procedures were performed under sterile conditions. All participants received an injection of 2 mL (40 mg/mL) triamcinolone and 2 cc of lidocaine. Ultrasound‐guided injection (N = 15): Ultrasound‐guided injection of glucocorticoid in the subacromial space with lateral approach Anatomical landmark‐guided injection (N = 15): Blind injection of glucocorticoid in the subacromial space with lateral approach. Co‐interventions: All the patients received physical therapy, including shoulder range of motion exercises; posterior capsule stretching and isometric exercises will be taught to patients.
Outcomes	Outcomes were measured at baseline, and 2 weeks and 2 months post‐intervention. Primary outcomes: 1. Mean (SD) Constant score (0 to 100, higher score indicates better function. Domains: pain, 15 points; activities of daily living, 20 points, range of motion, 40 points; strength, 25 points) 2. Mean (SD) visual analog scale (VAS) for pain Secondary outcomes: 3. Adverse events
Notes	Trial registration: IRCT2017021524621N6 (not prospectively registered) http://apps.who.int/trialsearch/Trial2.aspx?TrialID=IRCT2017021524621N6 Primary sponsor: Vice chancellor for Research, Iran University of Medical Sciences Recruitment status: Complete Authors confirmed that they were in the process of publishing the data.

Moore 2018.

Methods	Design: Parallel‐group, two‐arm, randomised controlled trial in participants with a painful shoulder associated with osteoarthritis. Setting: Not reported. United States Timing: Not reported Interventions: Corticosteroid and lidocaine injection into the glenohumeral joint using ultrasound guidance or anatomic landmarks Sample size: A power calculation for the primary outcome of therapeutic duration (time after the injection when VAS rises to ≥ 5 cm) was made using preliminary data where α = 5%, power = 0.8, and allocation ratio = 1.0 indicated that n = 10 in each group would provide statistical power at the P < 0.05 level. Analysis: Not specified
Participants	Number of participants: Number of participants screened for eligibility: Not reported Number enrolled: 30 (15 per group) Number randomised: 30 (15 per group) Number included in analyses at each follow‐up: 30 (15 per group) at 2 weeks and 6 months Inclusion Criteria: Inclusion criteria for participants with painful shoulder associated with osteoarthritis included: 1) pain with passive abduction/elevation of the shoulder 2) pain to deep palpation lateral to the coracoid process 3) nocturnal shoulder pain 4) significant pain in the affected shoulder by 0‐10 cm Visual Analog Pain Sale (VAS) where VAS ≥ 5 cm 5) failure of rest, range of motion exercises, and nonsteroidal anti‐inflammatory drugs 6) radiographs that demonstrated only mild to moderate (Grade I‐II) osteoarthritis of the glenohumeral joint 7) the desire of the participant to have a glucocorticoid injection Exclusion Criteria: 1) shoulder deformity 2) complete rotator cuff tear 3) the diagnosis of cervical spine nerve root impingement 4) recent trauma 5) haemorrhagic diathesis 6) use of warfarin or other anticoagulants 7) the presence of infection 8) previous glucocorticoid injection into the shoulder in the preceding 6 months Baseline characteristics: 'Lankmark' guidance group: Mean (SD) age = 53.6 (15.8) years 14 female; 1 male Mean (SD) pain on VAS = 7.9 (1.3) Mean (SD) pain on needle introduction = 3.4 (3.4) Significant needle introduction pain (VAS > 5 cm) = 3 (20%) Mean (SD) injection pain on VAS = 1.4 (2.4) Significant injection pain (VAS > 5cm) = 1 (6.7%) Ultrasound‐guided group: Mean (SD) age = 57.3 (12.9) years 14 female; 1 male Mean (SD) pain on VAS = 8.0 (1.8) Mean (SD) pain on needle introduction = 2.5 (1.8) Significant needle introduction pain (VAS > 5 cm) = 0 (0%) Mean (SD) injection pain on VAS = 0.3 (0.6) Significant injection pain (VAS > 5 cm) = 0 (0%)
Interventions	The injection procedure of the shoulder was performed in a standardised manner using the anterior approach in both the ultrasound and anatomic landmark groups. The one needle two‐syringe technique was used where 1) one needle is used for anaesthesia and glucocorticoid injection; 2) the first syringe is used to anaesthetise the shoulder and perform arthrocentesis, and 3) the second syringe is used to inject the glucocorticoid therapy. Anatomical landmark‐guided injection (N = 15): In the palpation guided anatomic landmark technique with the forearm held relaxed in the patient’s lap, first the acromioclavicular joint and the coracoid process were palpated and marked with ink. The medial border of the humeral head was then palpated and marked. Chlorhexidine 2% was used for antisepsis. A 22‐gauge 2‐inch needle on a 3 mL syringe and filled with 3 mL of 1% lidocaine was introduced horizontally on the lateral side of the coracoid through the ligaments and joint capsule into the synovial joint with care not to forcefully inject into the glenoid labrum. Arthrocentesis was performed and joint effusion was aspirated, if present. The syringe was then detached from the needle while still in the glenohumeral joint and a 3 mL syringe prefilled with 60 mg triamcinolone acetonide suspension (Kenalog® 40) was attached to the indwelling needle, and the treatment was slowly injected. The needle was then extracted, firm pressure applied to the puncture site, and a sterile bandage was applied. Ultrasound‐guided injection (N = 15): For the ultrasound‐guided procedures, a portable ultrasound unit with a 10‐5 MHz 38 mm broadband liner array transducer (Sonosite M‐Turbo, SonoSite, Inc. 21919 30th Drive SE, Bothell, WA 98021) was used. In the ultrasound‐guided technique the linear probe was held horizontally and the coracoid process, humeral head, and glenoid labrum were identified. The needle was then introduced out‐of‐plane inferiorly and laterally to the coracoid and then the transducer moved 90 degrees so that the needle was in plane with the transducer using an ultrasound gel wedge to assist with in‐plane positioning. The needle was then introduced through the joint capsule into the glenohumeral synovial space and lidocaine injected to confirm needle positioning. After injection of the lidocaine, arthrocentesis was performed and joint effusion was aspirated if present. The syringe was then detached from the needle while still in the glenohumeral joint and a 3 mL syringe prefilled with 60 mg triamcinolone acetonide suspension was attached to the indwelling needle, and the treatment was slowly injected. If there was unexpected resistance to injection, needle tip positioning was examined with ultrasound and the needle was rotated to change bevel positioning until resistance subsided and flow into the joint space could be confirmed, and the remainder of treatment was injected. The needle was then extracted, firm pressure applied to the puncture site, and a sterile bandage was applied.
Outcomes	Most outcomes were measured at baseline, 2 weeks (pain at 2 weeks was the primary outcome) and 6 months (or directly before the next injection if less than 6 months; secondary outcome). Exceptions included pain assessed upon insertion of the needle and during injection, and time‐to‐next‐intervention which was assessed to 12 months. Primary outcomes: 1. Mean pain on VAS (0 to 10) assessed at 2 weeks Secondary outcomes: 2. Mean pain on VAS (0 to 10) assessed at 6 months 3. Mean pain on VAS (0 to 10) upon insertion of the needle 4. Mean pain on VAS (0 to 10) at time of injection 5. Percentage of participants reporting treatment success (VAS < 2 cm) 5. Therapeutic duration (duration of therapeutic response time or time to flare), assessed as the time interval in months when the shoulder became symptomatic (VAS ≥ 5 cm) after the injection (up to 6 months) 6. Time‐to‐next‐intervention assessed as the time in months to the next glucocorticoid injection (up to 12 months) 7. Total number of corticosteroid injections in 12 months 8. Cost of injection procedure in USD, expressed per patient and per patient who reported treatment success
Notes	Published pre‐print https://www.biorxiv.org/content/10.1101/395293v1.full#:~:text=Conclusions,%2Dto%2Dnext%2Dinjection. Trial registration: Unclear if prospectively registered as trial registration number reported in pre‐print (NCT00651625) is an unrelated trial Recruitment status: Complete Funding: There was no industry support for this study.

Pierce 2018.

Methods	Design: Parallel‐design, two‐arm, randomised controlled trial in participants with subacromial bursitis Setting: Not reported Timing: Not reported Interventions: Injections into the subacromial space by ultrasound guidance or anatomic landmarks Sample size: 40 Analysis: Not specified
Participants	Participants who presented to the clinic with subacromial bursitis (including clinical and radiographic signs) and were deemed good candidates for injections
Interventions	(1) Ultrasound‐guided injection (N = 20) (2) Anatomical landmark‐guided injection (N = 20)
Outcomes	Outcomes were assessed at baseline and 12 weeks. Additionally, all participants were asked to rate their pain 15 minutes after their injection. Outcomes 1. Mean (SD) QuickDASH score (0 to 100, higher score indicates worse disability) 2. Pain intensity (0 to 10, higher score indicates worse pain) 3. Mean (SD) University of California‐Los Angeles (UCLA) Activity score (0 to 10, higher score indicates higher levels of activity)
Notes	Conference abstract. Authors confirmed that they were in the process of preparing the manuscript for publication.

SD = standard deviation; VAS = visual analog scale; DASH = Disabilities of the Arm, Shoulder, and Hand questionnaire; USD = US dollars

Differences between protocol and review

Differences between protocol and first published version of the review

We have clarified that included trials could include comparators of anatomic landmark‐guided injection or intramuscular injection.

We planned a subgroup analysis based on three time points (≤ six weeks; > six weeks to six months; > six months), but included studies that only measured outcomes to six weeks. For the purpose of pooling data, we combined one to two weeks in one analysis and six weeks in another analysis.

We also planned to perform sensitivity analysis to investigate the effect of imputation of missing data (e.g. imputation of standard deviations) if data were available but this was not performed.

Differences between first version of the review and current review update

In the original review, we excluded trials including participants with a primary diagnosis of glenohumeral osteoarthritis. One excluded trial included participants with shoulder pain among a mixed population of participants with both osteoarthritis (N = 48) and rheumatoid arthritis (N = 100), of which 22 injections were for the shoulder but the results were not presented separately by disease or joint (Sibbitt 2009), so the trial would have been excluded in any case.

In this review, we elected to include trials including participants with a primary diagnosis of glenohumeral osteoarthritis that compared image‐guided to landmark‐guided intra‐articular injection, as they would provide valid information for our question and could be included as a separate subgroup for the subgroup analysis. Only one trial was identified on this basis but it is awaiting assessment as only a preprint prior to peer review has been published to date (Moore 2018).

We included seven major outcomes in the summary of findings table.

We included the following time points: up to three weeks; up to six weeks (primary time point); up to three months; up to six months; > six months.

We updated the search strategy to search only MEDLINE, Embase and Cochrane CENTRAL Register of Controlled Trials (CENTRAL), plus clinicaltrials.gov and the WHO trial registry portal ‐ to conform with MECIR guidance.

Where trials used different measures for the same outcome, we did not pool data using the standardised mean difference (SMD) statistic. Instead, we used the most common outcome measure as an index outcome measure (Guyatt 2013). We transformed the mean and standard deviations (SDs) of other outcome measures to the scale of the index instrument and pooled the data using MD as the summary estimate.

We changed the search strategy to include updated subject heading terms and randomised controlled trial filters.

Instead of MCIDs for pain and function, we considered the magnitude of the mean between‐group differences as defined by the American College of Physicians 2017 guidelines for low back pain (Chou 2017). For pain measured on a 0 to 10‐point scale, a difference of 0.5 to 1.0 points was considered slight/small, a difference of > 1 to 2 points was considered moderate and a difference of > 2 points was considered large. For function and health‐related quality of life measured on a 0 to 100‐point scale, a difference of 5 to 10 points was considered slight/small, a difference of > 10 to 20 points was considered moderate and a difference of > 20 points was considered large.

Contributions of authors

All authors contributed to the writing of the review.

Sources of support

Internal sources

• Department of Epidemiology and Preventive Medicine, School of Public Health and Preventive Medicine, Monash University, Australia

  In kind support

• Cabrini Institute, Cabrini Hospital, Malvern, Victoria, Australia

  In kind support

• The University of Sydney, Australia

  In kind support

External sources

• National Health and Medical Research Council (NHMRC), Australia

  R Buchbinder is supported by an Australian National Health and Medical Research Council Senior Principal Research Fellowship

Declarations of interest

JZ: none known.

AR: none known.

RB is the Co‐ordinating Editor of Cochrane Musculoskeletal, but was not involved in editorial decisions regarding this review. She is a recipient of a National Health and Medical Research Council (NHMRC) Cochrane Collaboration Round 7 Funding Program Grant, which supports the activities of the Cochrane Musculoskeletal Editorial base, but the funders do not participate in the conduct of reviews.

RJ is the Managing Editor of Cochrane Musculoskeletal, but was not involved in editorial decisions regarding this review. She is a recipient of an NHMRC (Australia) Cochrane Collaboration Round 7 Funding Program Grant, which supports the Cochrane Musculoskeletal Editorial base, but the funders do not participate in the conduct of this review.

Edited (no change to conclusions)