To address our hypothesis, we conducted a systematic review of the available literature. The four authors independently performed the search on April 12, 2010. The search was performed with use of the following databases: PubMed, EMBASE, CINAHL, SportDiscus, MEDLINE, and the Cochrane Central Register of Controlled Trials, Database of Systematic Reviews, and modified Coleman Methodology Register (see AppendixAppendix). Search terms included frozen shoulder, adhesive capsulitis, stiff shoulder, Duplay disease, injection, corticosteroid, and steroid. Level-I, II, III, IV, and V evidence (according to the Oxford Centre for Evidence-Based Medicine used by The Journal of Bone and Joint Surgery [American volume])13 was reviewed for inclusion in this study. Potential inclusive papers were manually reviewed and were discussed among the authors, and a decision was made regarding inclusion or exclusion. In the event of disagreement among authors for study inclusion, the final decision was made by the senior author (G.L.J.). The full-text article was reviewed, and the reference list was checked for potential studies not identified by our original search.
Studies that were considered for inclusion in the systematic review included studies with Level-I and II evidence (randomized controlled trials) that were written in the English language; that involved human subjects; that had a minimum of six months of follow-up; that had a publication date from January 1, 1950, to April 12, 2010; that investigated the treatment of primary adhesive capsulitis (including in subjects with diabetes mellitus); that investigated the treatment of primary adhesive capsulitis during the freezing14 or frozen stage14; and that investigated intra-articular glenohumeral joint corticosteroid injection as compared with another recognized treatment.
Studies that were excluded included those that were not written in the English language; that had less than six months of follow-up; that investigated the treatment of causes of shoulder pain other than adhesive capsulitis; that investigated the treatment of secondary causes of adhesive capsulitis; that investigated the treatment of primary adhesive capsulitis during the thawed stage14; that investigated intra-articular glenohumeral joint corticosteroid injections by comparing different doses, different injection techniques, or different types of corticosteroid; that investigated subacromial corticosteroid injections in comparison with non-intra-articular corticosteroid treatment; or that investigated intra-articular glenohumeral joint hyaluronate injection in comparison with non-intra-articular corticosteroid injection treatment.
Initial application of the inclusion criteria yielded 932 citations. Eight studies met the inclusion criteria and were retained for further analysis3,5,15-20. Five studies specifically excluded diabetic patients (n = 215 patients), whereas three studies specifically allowed diabetic patients (n = 191 patients). Nevertheless, only three diabetic patients were identified among the latter three studies. The control group for comparison within the present review included patients who received no treatment (n = 8 patients), ice pack treatment (n = 12 patients), or small-volume (∼2-mL) intra-articular saline solution injection (n = 49 patients).
For assessment of the quality of the randomized controlled trials, we utilized the modified Coleman Methodology Score (MCMS)21-23 and the Quality Appraisal Tool24 (see AppendixAppendix). Originally designed to grade clinical studies investigating Achilles25 and patellar21 tendinopathy, the original Coleman Methodology Score involved the use of ten criteria and had a potential overall score of between 0 and 100, with 100 signifying a study that optimally limits chance, bias, and confounders. The modified Coleman Methodology Score22 is the sum of fifteen unique components, with a scaled potential score of between 0 and 100, assessing the reporting of study quality. Scores are classified as excellent (85 to 100), good (70 to 84), fair (55 to 69), or poor (<55). The Quality Appraisal Tool contains twelve items. Each item on the Quality Appraisal Tool may receive a maximum of 2 points and a minimum of 0 points. Each item was reviewed by the authors, who attempted to achieve consensus. When consensus was not achieved, the lower score was used by default according to the recommendation in the original study utilizing the Quality Appraisal Tool26. The sum of the individual items allowed for the calculation of the percentage quality (the sum of the individual scores divided by twenty-four multiplied by 100).
An attempt was made to identify common outcome measures across all studies analyzed. However, several different clinical outcome measures were variably utilized (Constant-Murley scores, SF-36 [Short Form-36] Physical Component Summary [PCS] and Mental Component Summary [MCS] scores, SPADI [Shoulder Pain and Disability Index] scores, VAS [visual analog scale] scores, and SST [Simple Shoulder Test] scores). Demographic data and treatment data were analyzed with calculations of the mean, median, mode, and standard deviation, when appropriate. A p value of <0.05 was a priori deemed significant. A paired t test was utilized for outcome data comparison when possible on the basis of study reporting. The differences in the reporting of data prevented the assimilation of data between studies and, therefore, significance is only reported in the current review when the individual study demonstrated significance as such. The lack of subject-level-specific data and heterogeneity in outcome reporting precluded meta-analysis.
Eight studies comprising 406 subjects (409 shoulders) met the criteria for inclusion. The mean modified Coleman Methodology Score (and standard deviation) was 44 ± 14 (poor) (Table I). One study16 had a fair score, whereas the remaining seven studies had a poor score. The mean quality appraisal score was 17 ± 5.6, which correlated with a 70.8% ± 23.3% quality rating (Table I). Only three of the eight reports described the method of randomization; two studies involved the use of random-number tables in closed envelopes, and one involved a computerized random-number generator. Of the 406 subjects, 175 were male and 231 were female. The mean age for all subjects was 55 ± 1.7 years. Dominance was known for 170 (42%) of the affected shoulders, with the dominant shoulder being affected 53% of the time (ninety of 170 shoulders). When staging of disease was reported11,14, seventy-six subjects had Stage-2 disease (frozen stage). The mean pretreatment duration of symptoms was 182 ± 93 days. Five studies (n = 215 patients) excluded diabetic patients, whereas three studies (n = 191 patients) included three diabetic patients (1.6%; three of 191).
All treatments resulted in improved subjective and objective clinical outcome measures. Tables II and III show the treatment allocation distribution and injection data. On the basis of Constant-Murley scores (Table IV), there were no significant differences between the treatment groups in terms of the amount of improvement at the time of the latest follow-up or at the interim time points (p > 0.05). The mean Constant-Murley score improved 34.6 points at fifty-two weeks after intra-articular steroid injection, compared with 37.1 points at twenty-six weeks after intra-articular saline solution hydrodilation, 45 points at twelve to thirteen weeks after manipulation with the patient under anesthesia, and 28.3 points at fifty-two weeks after oral steroid taper. Low numbers of subjects within the individual studies both at the time of the latest follow-up and at the interim short-term follow-up periods precluded significant conclusions regarding the difference in improvements in the Constant-Murley score. On the basis of SPADI scores (Table V), there was no significant difference in the amount of improvement between treatment with intra-articular steroid and treatment with intra-articular saline solution (2 mL) (control) at the time of the latest follow-up (50-point reduction compared with 46-point reduction; p > 0.05). However, at six weeks of follow-up, there was a significant difference in SPADI scores, with the score following steroid injection being 19 points better than the control value. On the basis of SF-36 PCS and MCS scores (Table VI), intra-articular steroid treatment resulted in significantly greater improvements in comparison with manipulation under anesthesia at the time of the two-year follow-up (p < 0.05) but was not significantly different compared with intra-articular saline solution (2 mL) (control) at the time of the one-year follow-up. Most treatments resulted in improved passive range of motion at the time of early follow-up (four to ten weeks) and at the time of the latest follow-up (Table VII).
Pain, as measured with heterogeneous visual analog scales, was significantly reduced (p < 0.05) in seven of the eight studies (358 [88%] of 406 subjects). This pain relief was generally experienced at the time of early follow-up (four to six weeks) and was sustained until the time of the latest follow-up. Nevertheless, when intra-articular steroid treatment was compared with oral corticosteroid treatment, manipulation under anesthesia, hydraulic distention, and physical therapy, there were no significant differences in terms of the degree of pain relief as measured with visual analog scales. However, in terms of the SPADI pain subscore, there was a significant difference (p < 0.0302) demonstrating better pain relief in association with steroid injection plus physical therapy or isolated steroid injection as compared with saline solution plus physical therapy or isolated saline solution injection.
This systematic review of existing Level-I and II evidence reveals that all treatments of adhesive capsulitis of the shoulder resulted in improved clinical outcome measures according to Constant-Murley, SPADI, and VAS pain scores, with a trend toward greater improvement in SF-36 scores with intra-articular steroid injection as compared with manipulation under anesthesia but not as compared with intra-articular saline solution. Most treatments resulted in improved passive range of motion at the time of early follow-up, with both intra-articular steroid and oral steroid showing significantly greater improvements in terms of abduction and forward elevation as compared with intra-articular lidocaine and intra-articular saline solution (p < 0.05). These significant improvements do appear to be transient, as all treatments resulted in improved passive range of motion at the time of the latest follow-up, with no treatments reaching a significant difference (p < 0.05) compared with another. Pain, as measured with the SPADI and VAS, was significantly reduced (p < 0.05) at the time of both short-term and latest follow-up with intra-articular corticosteroid, oral corticosteroid, manipulation under anesthesia, hydraulic joint distention, and physical therapy. The only significant difference between intra-articular steroid and the remaining groups was found in one study that demonstrated better SPADI pain scores in association with intra-articular steroid plus physical therapy and intra-articular steroid alone as compared with a saline-based placebo control16.
The purpose of the present study was to review existing Level-I and II evidence regarding the effectiveness of intra-articular corticosteroid injections for the treatment of idiopathic adhesive capsulitis as compared with other common treatment modalities. We hypothesized that the results for intra-articular steroids would result in significant improvements in the short term but would show no significant difference as compared with other treatments in the long term. Our hypothesis was confirmed on the basis of this review.
In the comparison between intra-articular corticosteroid and oral corticosteroid, this systematic review showed that both treatments result in significant improvements in terms of Constant-Murley3,5, Simple Shoulder Test (SST)3, Visual Analog Scale (VAS)3,5, and range-of-motion scores3,5,15-17,19,20, with the differences in range of motion, the Constant-Murley score, the SST, and patient satisfaction reaching significance in favor of intra-articular steroid injections3.
In the comparison between intra-articular corticosteroid and manipulation under anesthesia, this systematic review showed mixed results. In one study, there was no significant difference between the two treatments in terms of the Constant-Murley, VAS, and SF-36 scores, and the final conclusion was in favor of intra-articular steroid injections as it carries less inherent risk than manipulation under anesthesia5. In another study, manipulation under anesthesia combined with intra-articular corticosteroid and lidocaine was compared with hydrodilation with normal saline solution until capsular rupture18. That study demonstrated significant improvement in terms of Constant-Murley and VAS scores in favor of hydrodilation with capsular rupture.
In the comparison between intra-articular corticosteroid injection and intrabursal injection of both corticosteroid and local anesthetic, no significant difference in pain relief or shoulder motion was reported19. The data in that study compared four groups: intra-articular corticosteroid and lidocaine, intrabursal corticosteroid and lidocaine, intra-articular lidocaine alone, and intrabursal lidocaine alone. This was the only study that evaluated subacromial bursa injection. Further analysis of a comparison between intra-articular and intrabursal injection could not be performed.
In the comparison between intra-articular corticosteroid and control (predefined as no treatment, ice pack, or low-volume intra-articular saline solution [∼2 mL] injection), intra-articular corticosteroid provided earlier return of shoulder motion3,5,15-20 and improvement in SPADI scores16, with no difference in longer-term outcome (fifty-two weeks). Nevertheless, only sixty-nine patients in this review were control subjects (including forty-nine who were managed with saline solution injection, twelve who were managed with an ice pack, and eight who received no treatment).
Despite the Level-I and II evidence nature of this systematic review, several limitations were present. On account of these deficiencies, the poor methodological characteristics of the studies analyzed in the present review mandate an overall need for improvement in study quality in the future. With use of standardized, validated study-quality-assessment tools (the modified Coleman Methodology Score [MCMS] and Quality Appraisal Tool), study quality was found to be poor (in seven of eight studies), with a mean Quality Appraisal Tool score of 70.8%. Scrutiny of the individual components within the MCMS illustrates severe shortcomings in the studies with regard to the calculations of the number-needed-to-treat (NNT), the lack of independent post-treatment investigators, patient and investigator blinding, and the performance of power analysis. Analysis of the individual components within the Quality Appraisal Tool illustrates shortcomings in of the studies in terms of the explicit description of hypotheses, appropriate psychometric properties, sample-size calculations and power analyses, and appropriate statistical error estimates, all of which were lacking.
Properly conducted randomization attempts to eliminate selection bias and support the internal validity of a study27. Only three of the eight reports in this review described the randomization methods (closed envelope in two studies and computer-based random-number generator in one). None of the studies indicated whether the authors were involved in the randomization of subjects. Another source of selection bias includes unequal numbers of subjects in each compared group (Table II). Only three subjects in our systematic review had diabetes mellitus. This disease is commonly associated with adhesive capsulitis that is more difficult to treat11.
The assessment of range of motion with use of a goniometer (as was done in only three studies in this review15-17) has high intratester reliability but only fair to moderate intertester reliability28. Nevertheless, range-of-motion measurement either was not reported or was visually performed in five of the eight included studies, despite recommendations for the use of a calibrated goniometer during the performance of high-level evidence studies28. Dissimilar techniques of intra-articular injection (e.g., with use of fluoroscopy), dissimilar types of intra-articular corticosteroid (triamcinolone versus methylprednisolone), and dissimilar doses of intra-articular corticosteroid also introduce performance bias. Confounding the treatment effect of isolated intra-articular injection was the concurrent intrabursal injection of corticosteroid in one study15. Furthermore, image-guided intra-articular injections have been shown to be more accurate, with a significantly improved early clinical benefit of this accuracy, compared with "blind" technique29,30. In this review, 56% (153) of all 271glenohumeral joint injections involved the use of fluoroscopy for intra-articular placement and injection localization. Dissimilar physical therapy regimens across the studies is a confounding variable in assessing the results of treatment. Furthermore, incomplete understanding of the true natural history of adhesive capsulitis precludes definitive conclusions on how treatment affects the course of disease. The stage of disease was known for only 19% (seventy-six) of the 406 subjects analyzed, further introducing bias.
The assessment of outcomes via an independent observer is necessary to minimize detection bias. However, independent evaluators of clinical outcome or shoulder motion following treatment were utilized in only three of the eight studies analyzed15,17,19. The primary outcome measures used in the studies analyzed were shoulder motion and the Constant-Murley, SPADI, and SF-36 scores. The SF-36 is the most commonly used general health-assessment tool in the orthopaedic literature31 and has high psychometric standards in the evaluation of quality of life32. Although a minimum clinically important difference for the SF-36 has not been determined in the analysis of adhesive capsulitis, the minimum clinically important difference for the physical function component of the SF-36 is 3 to 5 points in the assessment of knee and hip osteoarthritis33. Nevertheless, the SF-36 was utilized in only two of the eight studies in this review5,16. Although the minimum clinically important difference for the Constant-Murley score has yet to be determined26, this tool has been validated for use in the assessment of adhesive capsulitis34. However, the Constant-Murley score was only utilized in three studies in this review3,5,18. The minimum clinically important difference for the SPADI is 1826. This minimum clinically important difference was easily met by both subgroups analyzed in the one study in which the SPADI was used (with a difference of 50 points for the intra-articular steroid group and 46 points for the low-volume [2-mL] intra-articular saline solution group)16. Thus, based on the SPADI outcome measure, a significant finding is likely to correlate with a clinically important finding.
In this review, the duration of follow-up after treatment was short (minimum, six months). Although the natural history of adhesive capsulitis is incompletely understood, it is well recognized that, regardless of treatment, most patients achieve the maximal outcome between two and four years after treatment6,35. However, some studies have demonstrated severe loss of shoulder motion, inability to perform activities of daily living, and mild residual pain and weakness at longer-term follow-up (as long as seven years)7-10. Adhesive capsulitis is often described as a self-limited condition36, and if there were a way to determine the treatment modalities that could hasten its course or improve subjective and objective measures during the disease course, this would be relevant information to know. Therefore, the inclusion of studies with fewer than two years of follow-up was appropriate. As a result, our study is limited in that the longest mean duration of follow-up in any of the studies was two years.
A possible reason for the poor study quality in these Level-I and II reviews may not actually be the inherent limitations in the enrollment of subjects, the performance of treatment, and the post-treatment assessment, but rather may be the poor quality of reporting of what the authors did. Journal space requirements limit the amount of information conveyed. Nevertheless, authors should attempt to succinctly report and describe the necessary and salient methodological qualities of their studies.
This systematic review of existing Level-I and II evidence reveals that all treatments resulted in improved clinical outcome measures based on Constant-Murley and SPADI scores, with a trend toward greater improvement in SF-36 scores in association with intra-articular corticosteroid injection as compared with manipulation under anesthesia but not compared with intra-articular injections of saline solution. Most treatments resulted in improved passive shoulder motion at the time of early follow-up, with both intra-articular and oral corticosteroid treatment showing significantly greater improvements in shoulder abduction and forward elevation in comparison with intra-articular lidocaine and intra-articular saline solution (p < 0.05). These significant improvements appear to be transient as all treatments resulted in improved passive shoulder motion at the time of the latest follow-up, with no treatment reaching a significant difference (p < 0.05) as compared with another. Pain, as measured with the SPADI and VAS, was significantly reduced both at the time of short-term follow-up and at the time of the latest follow-up after treatment with intra-articular corticosteroid, oral corticosteroid, manipulation under anesthesia, hydraulic joint distention, and physical therapy. The only significant difference between intra-articular corticosteroid treatment and the remaining treatments was found in one study demonstrating better SPADI pain scores with intra-articular corticosteroid plus physical therapy and intra-articular corticosteroid alone compared with intra-articular saline solution plus physical therapy or isolated intra-articular saline solution alone. The poor methodological characteristics of the studies analyzed in this review emphasizes an overall need for improvement in study quality in the future.