Evidence-based approach to the shoulder examination for subacromial bursitis and rotator cuff tears: a systematic review and meta-analysis

Abstract

Introduction

Shoulder pain represents a common patient complaint evaluated in a primary care setting. Approximately two thirds of these patients have rotator cuff injuries, with rotator cuff tears (RCTs) and subacromial impingement syndrome (SIS) accounting for a majority of causes. An accurate and efficient diagnostic strategy focused on physical examination findings may lead to improved outcomes and less functional disability.

Objective

To identify the most accurate physical examination tests for the diagnosis of RCT and SIS, we performed a systematic review and meta-analysis.

Methods

The database for our systematic review was compiled by using keywords and common indexing strategies to search PubMed, Ovid MEDLINE, Embase and the Cochrane Library from January 1, 1980, to March 15, 2024. Included studies evaluated a physical examination being performed prior to a reference standard diagnostic test for either RCT or SIS. Data was extracted in dual fashion and meta-analyses were performed regarding physical examination tests identified in our included studies.

Results

A total of twenty studies, which include 3,438 patients, met our inclusion criteria and had data extracted for statistical analysis. Data was adequate to perform meta-analyses on ten physical examination tests for RCT and five physical examination tests for SIS. There were six physical examination tests which had significant diagnostic odds ratios (DORs) when used in the diagnostic evaluation of suspected RCT, with the External Rotation Lag Sign at 90 Degrees having the highest magnitude of significance (DOR, 12.70; 95% CI, 3.68 – 43.86; P < .0001). Four physical examination tests had significant DORs for physical examination tests when used in the diagnostic evaluation of SIS, with the Yergason’s Test having the highest magnitude of significance (DOR, 4.71; 95% CI; 2.16 – 10.32; P = .0001).

Conclusion

We present a large body of low-quality evidence for the diagnostic accuracies of physical examination tests for the identification of both RCT and SIS. We have identified novel data for the accuracy of the External Rotation Lag Sign at 90 Degrees and the Internal Rotation Lag Sign that have high to moderate diagnostic accuracy for ruling in tears of the rotator cuff.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12891-024-08144-z.

Background

The objective of our systematic review and meta-analysis was to attempt to improve upon the evidence base for the bedside diagnosis of rotator cuff tears and the subacromial impingement syndrome. To our knowledge we are presenting the largest database to date on physical examination tests used in the diagnostic evaluation of patients suspected of having rotator cuff tears or the subacromial impingement syndrome. Our study has significant potential to improve early detection and benefit public health worldwide, as these data may be useful to better define clinical practice guidelines. We present novel data regarding the accuracy of the External Rotation Lag Sign at 90 Degrees and the Internal Rotation Lag Sign in the diagnosis of rotator cuff tears. Our findings show these two tests are better positive predictors than previous studies.

Introduction

A 1988 decree by the United Nations called for raising awareness of the impact of musculoskeletal disorders on society by improving both treatment and prevention [1]. Shoulder pain comprises a significant portion of this societal impact as the third most common musculoskeletal complaint evaluated by primary care physicians (PCP), which has been shown to be associated with increases in both disability and health care costs [2, 3]. In the United States, musculoskeletal pain accounts for the third highest condition in healthcare spending, only behind diabetes mellitus and ischemic heart disease, respectively [3]. Unfortunately, only one-half of all new episodes of shoulder pain end in complete recovery within six months [4]. Shoulder pain is often multifactorial; however, rotator cuff injuries with rotator cuff tears (RCTs) and subacromial impingement syndrome (SIS) account for the majority of causes [5].

The physical examination has a long history as an integral element in the initial evaluation of musculoskeletal conditions. The abundance of physical examination tests used to evaluate patients with suspected SIS or RCT may be the most obvious diagnostic challenge in the primary care setting. Confusion over the taxonomy of naming these physical examination tests, the lack of clarity regarding the criteria for a positive test interpretation, and the paucity of evidence to support the diagnostic accuracy of a single examination maneuvers have also been cited in previous systematic reviews and meta-analyses as hindrances to diagnostic accuracy for many clinicians [6, 7]. Limitations in the clinical efficacy for shoulder examination tests in isolation compared with using serial examination maneuvers has been highlighted in several past reviews as a roadblock to performing an accurate shoulder examination [6–9]. Data from previous systematic reviews with meta-analyses also have limitations due to small number of included studies as well as low diagnostic accuracy of tests studied in patients having suspected SIS or a RCT. Hegedus et al., found the painful arc test to be the most accurate in diagnosing SIS with a likelihood ratio (LR) of only 2.25 (95% CI; 1.24 – 4.0) derived from four studies [6]. Gismervik et al., reported a diagnostic odds ratio (DOR) of 2.86 (95% CI; 1.14 – 7.17) from only two studies for the Hawkins-Kennedy test in the diagnosis of SIS [7]. This same study found a DOR for any type of RCT of 2.63 (95% CI; 1.62 -4.270 for Jobe’s test [7]. Alqunaee et al., reported a high LR of 16.47 (95% CI; 1.46 – 185.61) for the lift-off test in the diagnosis of SIS; however, these data should be viewed with caution as they were derived from a small sample size and the confidence intervals were very wide [10]. None of these studies reported any tests having negative LR or DOR of significance to help rule out SIS or RCT.

Based on the evidence summarized above, one could conclude that previous studies have not led to a pathognomonic diagnostic approach for shoulder conditions which are often evaluated in the primary care setting. Data supporting an early and accurate diagnosis for RCT and SIS have led to increased range of shoulder motion postoperatively as well as to improved pain and disability, respectively [11, 12]. These findings further emphasize the need for an evidence-based approach to the shoulder examination which would improve both diagnostic accuracy and early detection for all clinicians. Additionally, several years have passed since the last large-scale systematic review and meta-analysis was performed on the physical examination for rotator cuff pathology, which prompted our decision to conduct an updated study. To our knowledge, we have performed a unique study with regard to the choice of our topic and to the breath of our systematic review and meta-analysis. Our goal was to present the most comprehensive and current database which could enhance the diagnostic certainty for the physical examination to evaluate patients suspected of having SIS or a RCT.

Methods

We began our systematic review by developing a Patient, Intervention, Comparison, Outcomes, and Study design (PICOS) question and proceeded with adherence to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guidelines [13]. Our systematic review and meta-analysis has been registered with the International Prospective Register of Systematic Reviews (PROSPERO; CRD 42024541704).

Data sources and searches

We conducted our database searches by using keywords and medical subject headings (MeSH), along with common indexing practices that were rigorously tested to capture all articles potentially relevant to our systematic review. PubMed, Ovid MEDLINE, Embase and the Cochrane Library were searched from January 1, 1980, to March 15, 2024, to establish our database. We used MeSh terms when searching PubMed, Ovid MEDLINE and the Cochrane library. For our Embase search, we employed Emtree to classify articles. For an example of search strategy and keywords, please refer to the study protocol (Supplement 1). Author and refence tracking were used to augment our initial search.

Study selection and eligibility criteria

Title and abstracts were compiled and stored electronically by using software designed for systematic reviews. Following a dual extraction protocol, two authors (CD and RS) independently screened all titles and abstracts followed by full-text review. Disagreements were resolved by reaching a consensus between these same two authors with a third author (QZ) available if a consensus was not able to be reached. Cohort studies, case series, and case control studies were eligible for inclusion that collected data on patients undergoing a physical examination prior to having a reference standard imaging or surgical confirmation of either SIS or a RCT. Studies were excluded when the examiners were not blinded to the results of the reference standard confirmatory imaging test or surgical findings. Abstract-only publications were excluded due to the uncertainty of a rigorous peer review process. There were no language restrictions as a part of our exclusion criteria.

Data extraction and quality assessment

Two authors (CD and RS) independently extracted data into files for statistical analysis from studies which satisfied our inclusion criteria. Agreement on data extraction was derived though discussion between these same two authors to achieve a consensus. Study characteristics are listed in Table 1 and include: first author and the year of publication; study type; number of participants studied; shoulder pathology studied; physical examination tests studied; and reference standard used to confirm a diagnosis. Two authors (QZ and RS) independently assessed the risk of bias of all included studies using a modified version of the Newcastle–Ottawa Scale for assessing the quality of non-randomized studies in meta-analyses available from the National Institutes of Health [14, 15].

Table 1.

Study characteristics

Author, year	Study design	Number of study participants	Mean age of participants	Percent male	Shoulder pathology studied	Physical examination tests studies	Reference standard
Calis, 2000 [16]	Prospective cohort	120	52	40%	SIS	Neer Test Hawkins Test Painful Arc Test Drop Arm Sign Yergason’s Test	Magnetic Resonance Imaging
Litaker, 2000 [17]	Retrospective chart review	448	57	63%	RCT	Hawkins test	Arthrography
Park, 2005 [18]	Prospective cohort	552	Not reported	Not reported	SIS	Neer Test Hawkins Test Painful Arc Test	Arthroscopy
Kim, 2007 [19]	Prospective cohort	120	59	23.3%	RCT	Neer Test Hawkins Test Jobe’s Test Patte’s Test, Lift-off Test Yergason’s Test	Ultrasound
Miller, 2008 [20]	Same-subject, Correlation, double-blinded study	37	56	43.24%	RCT	External Rotation Lag Sign Drop Arm Sign Internal Rotation Lag Sign	Ultrasound
Silva, 2008 [21]	Prospective cohort	30	55	46.66%	SIS	Neer Test Hawkins Test Jobe’s Test	Magnetic Resonance Imaging
Fodor, 2009 [22]	Prospective cohort	100	57	35%	SIS	Neer Test Hawkins Test Painful Arc Test	Ultrasound
Michener, 2009 [23]	Prospective cohort	55	41	85.45%	SIS	Hawkins Test Neer Test Painful Arc Test Jobe’s Test	Surgery
Bak, 2010 [24]	Prospective cohort	52	57	65.5% in the study group	RCT	Neer Test Hawkins Test Jobe’s Test Painful Arc Test Internal Rotation Lag Sign External Rotation Lag Sign	Ultrasound and Arthroscopy
Fowler. 2010 [25]	Prospective cohort	101	41	825%		Jobe’s Test Hawkin’s Test Lift-off Test	Arthroscopy
Kelly, 2010 [26]	Cross-sectional	34	57	58.82%	SIS	Neer Test Hawkins Test Painful Arc Test Jobe’s Test Yergason’s Test	Ultrasound
Toprak, 2013 [27]	Prospective cohort	69	48	30.43%	SIS	Neer Test Hawkin’s Test	Ultrasound
Van Kampen, 2014 [28]	Prospective cohort	100	44	65%	RCT	Jobe’s Test Neer Test Hawkins Test Drop Arm Sign Lift-off Test Painful Arc Test External Rotation Lag sign Drop Arm Sign	Magnetic resonance arthrography
Villafane, 2015 [29]	Prospective cohort	100	52	48%	RCT	Jobe’s Test Patte test Neer Test Hawkins Test	Magnetic Resonance Imaging
Penning 2016 [30]	Prospective cohort	49	57	41%	SIS	Jobe’s Test Drop Arm Test	Ultrasound
Jain, 2017 [31]	Prospective cohort	187	62	54%	RCT	Lift-off Test Bear Hug Test External Rotation Lag Sign Drop Arm Sign Jobe’s Test Neer Test Hawkins Test	Magnetic Resonance Imaging
Jain, 2018 [32]	Prospective cohort	301	61	52%	RCT	Lift-off Test Bear Hug Test External Rotation Lag Sign Jobe’s Test Neer Test Hawkins Tet Drop Arm Sign	Magnetic Resonance Imaging
Kappe, 2018 [33]	Prospective cohort	106	57	59.4%	RCT	Lift Off Test Internal Rotation Lag Sign Bear Hug Test	Arthroscopy
Yazigi Junior, 2021 [34]	Prospective cohort	733	50.5 (median!)	52.25%	RCT	Jobe’s Test Drop Arm Sign Painful Arc Test Neer Sign Hawkins Test, Patte Test	Magnetic Resonance Imaging
Zou, 2022 [35]	Retrospective cohort	75	Not available	Not available	Not available	Jobe’s Test Drop Arm Sign Neer Test Hawkins Test	arthroscopy

Statistical analysis

Meta-analyses were performed when data was sufficient for pooling. We calculated sensitivities, specificities, DORs, positive LRs, and negative LRs for physical examination tests used in study participants for the diagnostic evaluation of SIS or RCT. We preferred to report our outcomes as DORs since they are representative of both sensitivities and specificities irrespective of disease prevalence and pre-test probability [36]. Bivariate methods were employed to model the sensitivity and specificity of diagnostic tests simultaneously. We presented DORs using traditional univariate meta-analysis methods, and Forest plots were created for the values of these statistics for all individual studies reporting on a particular diagnostic test. Heterogeneity was evaluated using Cochran’s Q and I² statistic and was considered significant at p < 0.05 for the Q statistic or I² > 50%. A random effects model was employed to calculate DOR with 95% confidence intervals (CIs) due to the heterogeneity of the patient populations of our included studies. Sensitivity analysis was conducted using the “leave-one-out” method to estimate the influence of each study on the pooled results. Outlier studies were deleted, and the analysis was revised when funnel plots suggested the potential for publication bias. All analyses were conducted using R version 4.4.0 in R Statistical Software [37]. The meta package was used for all DOR calculations using the random effect model [37, 38]. Package mada was used to calculate positive LR and negative LR as well as bivariate models for sensitivity and specificity [37, 38]. Forest plots and funnel plots were generated by packages meta and mada [37].

Institutional review board statement

The Trinity Health Ann Arbor Institutional Review Board, research compliance department, did not deem our project to involve human subjects and was exempt from further correspondence.

Consent to participate

Not applicable as no human or animal subjects were included in this study.

Results

After screening 5,549 titles and abstracts, 116 studies were selected for full text review (Fig. 1). Twenty studies met the inclusion criteria and had data extracted for statistical analysis from 3,438 patients (Table 1) [16–35, 39]. Study characteristics were recorded as follows: first author and year of publication; study design; number of study participants; mean age of participants; shoulder pathology studied; physical examination tests studied; and reference standard (Table 1). Data was adequate to perform meta-analyses on ten physical examination tests for RCT and five physical examination tests for SIS, which will be discussed below. These will be listed in order of DORs of the highest to lowest magnitude.

Fig. 1.

Prisma flow diagram

Tests for the diagnosis of Rotator Cuff tears

External rotation lag sign at 90 degrees

Our random effects model meta-analysis of three studies with 234 patients resulted in a significant pooled DOR of 12.70 (95% CI, 3.68 – 43.86; P < 0.0001) (Fig. 2A) for the External Rotation Lag Sign at 90 degrees used for diagnosing RCTs [20, 28, 31]. Leave-one-out sensitivity analysis did not show any significant difference in the reported DOR (Supplement 2).

Fig. 2.

DOR Forest plots for tests to diagnose rotator cuff tears

Internal rotation lag sign

A random effects model meta-analysis on four studies which included 295 patients, yielded a significant pooled DOR of 7.03 (95% CI, 2.98 – 16.61; P < 0.0001) for the Internal Rotation Lag sign used for diagnosing RCT (Fig. 2B) [20, 24, 28, 33]. Leave-one-out sensitivity analysis showed no significant difference in the reported DOR (Supplement 2).

External rotation lag sign at zero degrees

Our random effects model meta-analysis of three studies, which included 972 patients, for the External Rotation Lag Sign at Zero Degrees when use for diagnosing RCTs resulted in a significant pooled DOR of 3.56 (95% CI, 1.10 – 11.48; P = 0.03) (Fig. 2C) [24, 31, 34]. Leave-one-out sensitivity analysis showed no significant difference in the reported DOR (Supplement 2).

Jobe’s test

Nine studies with 1,668 patients comprised our random effects model meta-analysis for the Jobe’s test for diagnosing suspected RCT yielding a significant pooled DOR of 3.54 (95% CI, 1.37 -9.19; P = 0.009) (Fig. 2D) [19, 24, 28–32, 34, 35]. Leave-one-out sensitivity analysis showed no significant difference in the reported DOR (Supplement 2).

Hawkins test

Also known as the Hawkins-Kennedy test or Impingement Sign, our random effects model of seven studies, that included 1,195 patients, resulted in a significant pooled DOR of 3.11 (95% CI; 1.25 – 7.75; P = 0.015) for this examination in diagnosing suspected RCT (Fig. 2) [17, 24, 28, 29, 31, 34, 35]. Leave-one-out sensitivity analysis showed no significant difference in the reported DOR (Supplement 2E).

Bear hug test

Employing the Bear Hug Test in diagnosing RCT, our random effects meta-analysis of three studies, that included 437 patients, yielded a significant pooled DOR of 2.77 (95% CI; 1.13 – 6.77; P = 0.03) (Fig. 2F) [31, 33, 35]. Leave-one-out sensitivity analysis showed no significant difference in the reported DOR (Supplement 2F).

Patte test, lift-off test, Neer test, and drop arm sign

These four tests had random effects meta-analyses performed and resulted in non-significant pooled DOR in diagnosing RTC. Three studies which included 1,020 patients resulted in a DOR of 3.08 (95% CI; 0.20 – 47.62; P = 0.42) for the Patte Test (Fig. 2G) [29, 31, 34]. Five studies with 814 total patients resulted in a DOR of 3.00 (95% CI; 0.23 – 39.12; P = 0.40) for the Lift-off Test (Fig. 2H) [19, 28, 31–33]. Six studies which included 514 patients resulted in a DOR of 2.07 (95% CI; 0.87 – 4.93; P = 0.10) for the Neer Test (Fig. 2I) [24, 28, 29, 31, 35]. Five studies with 1,109 patients resulted in a DOR of 1.45 (95% CI 0.07 – 29.58; P = 0.81) for the Drop Arm Sign (Fig. 2J). [7, 20, 28, 31, 34].

Tests for the diagnosis of subacromial impingement syndrome

Yergason’s test

We performed a random effects model meta-analysis on three studies that included 209 patients and resulted in a significant pooled DOR of 4.71 (95% CI; 2.16 – 10.32; P = 0.0001) for the Yergason’s Test in diagnosing SIS (Fig. 3A) [16, 23, 26]. Leave-one-out sensitivity analysis showed no significant difference in the reported DOR (Supplement 2).

Fig. 3.

DOR Forest Plot for physical examination tests to diagnose subacromial impingement syndrome

Neer test

Seven studies which included 960 patients compromised our random effects model meta-analysis resulting in a significant pooled DOR of 4.02 (95% CI; 2.11 -7.69; P < 0.0001) for the Neer Test in diagnosing SIS (Fig. 3B) [16, 18, 21–23, 26, 27]. Leave-one-out sensitivity analysis showed no significant difference in the reported DOR (Supplement 2).

Hawkins test

Eight studies, which included 960 patients, comprised our random effects model meta-analysis resulting in a significant pooled DOR of 3.41 (95% CI; 1.78 – 6.54; P = 0.002) for the Hawkins test when used for diagnosing SIS (Fig. 3C) [16, 18, 21–23, 26, 27]. Leave-one-out sensitivity analysis showed no significant difference in the reported DOR (Supplement 2).

Painful arc test

Six studies with 891 total patients comprised our random effects meta-analysis resulting in a significant pooled DOR of 2.81 (95% CI; 1.31 – 6.06; P = 0.01) for the Painful Arc Test in diagnosing evaluation of suspected SIS (Fig. 3D) [16, 18, 21–23, 26]. Leave-one-out sensitivity analysis showed no significant difference in the reported DOR (Supplement 2).

Jobe’s test

Four studies which included 220 total patients comprised our random effects model metanalysis resulting in a non-significant pooled DOR of 2.03 (95% CI; 0.77 – 5.35; P = 0.15) (Fig. 3E) when used for diagnosing SIS. [21, 23, 25, 26] Leave-one-out sensitivity analysis showed no significant difference in the reported DOR (Supplement 2).

Discussion

Our study successfully identified some new information which should be very useful for clinicians. First, the External Rotation Lag Sign at 90 degrees, with a significant pooled DOR of 12.70 (95% CI, 3.68 – 43.86; P < 0.001) (Table 2), had the highest degree of diagnostic accuracy for the detection of a RCT and higher than reported in a past review [40]. Second, the Internal Rotation Lag Sign was also shown to have significant diagnostic accuracy of moderate degree for diagnosing RCT with a pooled DOR of 7.03 (95% CI, 2.98 – 16.60; P < 0.0001) (Table 2) and also higher than reported in a prior review [40]. Similar to previous studies, we found significant diagnostic accuracy for the External Rotation Lag Sign at Zero Degrees, Jobe’s Test, Hawkins Test, and Bear Hug Test for the detection of RCTs. These last four tests all had less diagnostic accuracy than the External Rotation Lag Sign at 90 degrees and the Internal Rotation Lag Sign (Table 2). Comparable with past systematic reviews and meta-analyses, we did not identify any single test to be very accurate in the diagnosis of SIS (Table 3) [6, 40]. We were surprised to find that Yergason’s test, which is not typically considered to be useful for the diagnosis of SIS, had the highest magnitude of diagnostic accuracy for this condition with a pooled DOR of 4.71 (95% CI; 2.16 – 10.32; P = 0.0001) (Table 3). The implication of the results for Yergason’s Sign for the diagnosis of SIS must be reviewed cautiously, as data was derived from a small number of studies with a small sample size.

Table 2.

Physical exam tests for rotator cuff tears meta-analysis data

Physical examination tests	Number of studies	Sensitivity, (95% CI)	Specificity, (95% CI)	Positive LR (95% CI)	Negative LR (95% CI)	DOR (95% CI)
External Rotation Lag Sign at 90 Degrees	3	0.17 (0.06—0.39)	0.99 (0.93—1.00)	6.91 (2.59—18.45)	0.88 (0.78—0.99)	12.70 (3.68—43.86)
Internal Rotation Lag Sign	4	0.525 (0.10—0.91)	0.92 (0.84—0.96)	4.23 (2.63—6.79)	0.77 (0.57—0.99)	6.76 (3.15—14.54)
External Rotation Lag Sign at 0 Degree	3	0.38 (0.13—0.71)	0.89 (0.46—0.99)	2.80 (0.80—9.74)	0.78 (0.62—0.97)	3.56 (1.10—11.48)
Jobe’s Test	9	0.717 (0.59—0.82)	0.57 (0.44—0.69)	1.58 (1.15—2.17	0.52 (0.31—0.86)	3.54 (1.37—9.19)
Hawkins Test	7	0.75 (0.58—0.87)	0.51 (0.27—0.75)	1.44 (1.10—1.87)	0.56 (0.40—0.77)	3.11 (1.25—7.75)
Bear Hug Test	4	0.84 (0.32—0.98)	0.28 (0.02—0.89)	1.29 (0.98—1.69)	0.74 (0.57—0.94)	2.77 (1.13—6.77)
Patte Test	4	0.33 (0.20—0.48)	0.89 (0.34—0.99)	2.96 (0.57—15.41)	1.02 (0.56—1.84)	3.08 (0.20 -47.62)
Lift Off Test	5	0.19 (0.14—0.26)	0.95 (0.65—1.00)	2.91 (0.59 -14.40	1.04 (0.77—1.39)	3.00 (0.23—39.12)
Neer Test	6	0.56 (0.45—0.65)	0.64 (0.38—0.83)	1.4 0(0.98—2.00)	0.78 (0.62—0.97)	2.07 (0.87—4.93)
Drop Arm Test	5	0.26 (0.11—0.51)	0.85 (0.20—0.99)	1.97 (0.30—12.93)	1.24 (0.77—2.00)	1.45 (0.07—29.58)

Table 3.

Physical exam tests for subacromial impingement syndrome meta-analysis data

Physical examination tests	Number of studies	Sensitivity, (95% CI)	Specificity, (95% CI)	Positive LR (95% CI)	Negative LR (95% CI)	DOR (95% CI)
Yergason’s Test	3	0.45 (0.33—0.58)	0.86 (0.76—0.92)	3.12 (1.73—5.62)	0.69 (0.57—0.84)	4.71 (2.16—10.32
Neer Test	7	0.73 (0.63—0.81)	0.52 (0.27—0.76)	1.54 (1.09—2.18)	0.47 (0.41—0.54)	4.02 (2.11—7.69)
Hawkins Test	8	0.71 (0.60—0.80)	0.59 (0.43—0.73)	1.64 (1.22—2.19)	0.53 (0.39—0.71)	3.41 (1.78—6.54)
Painful Arc Test	6	0.59 (0.43—0.74)	0.61 (0.37—0.81)	1.57 (1.07—2.31)	0.63 (0.46—0.86)	2.81 (1.31—6.06)
Jobe’s Test	4	0.60 (0.50—0.70)	0.57 (0.29—0.82)	1.36 (0.83—2.22)	0.72 (0.52—1.01)	2.03 (0.77—5.35)

The importance of our study is reflected by potential impact of the initial evaluation of patients presenting to their PCP with shoulder pain, given that as high as 65% of these patients have been shown to have rotator cuff tears or tendinopathy [40, 41]. The physical examination is likely the only diagnostic modality available to PCPs in these initial evaluations of patients presenting with shoulder pain. Hence, the results of this evaluation can have important diagnostic and therapeutic implications. Given that as many as 180 physical examination tests are described for evaluating patients with suspected shoulder pathology, the task of an efficient and accurate bedside diagnosis presents a challenge in clinical practice [42]. Unfortunately, data from several past systematic reviews and meta-analyses have not led to an evidence-based bedside diagnostic algorithm or clinical practice guidelines for the evaluation of patients with shoulder pain [6, 7, 42]. In fact, none of these studies concluded that any single physical examination test performed well when used by themselves to have reasonably good diagnostic accuracy to support or exclude the diagnoses of RCT or SIS [6, 7, 42]. Like previous studies, we found a modicum of physical examination tests which could be used alone as positive predictors for the presence of RCT or SIS with the exception of the External Rotation Lag Sign at 90 degrees and the Internal Rotation Lag Sign for the diagnosis of RCTs. Given their high and moderate positive predictive values, respectively, these tests deserve further attention for expanded use in clinical practice. They may also be useful when used with a combination of other physical examination tests, especially given the lower diagnostic accuracy of all the other tests we studied [17, 43]. Since many clinicians are more accustomed to LR than DOR, we also reported them in our meta-analysis (Tables 2 and 3). Negative likelihood ratios can be very beneficial for their negative predictive value; however, none of the tests we studied for either condition was particularly accurate to assist in excluding the diagnosis of either SIS or RCT.

We have recognized several limitations of our study. The main limitation was that all our included studies were observational (primarily prospective cohort studies), and none had a control group for comparison of the examination findings. All the included studies comprised patients who had symptoms related to their shoulders and were evaluated in clinics designed to evaluate these symptoms, as well as to have received a reference standard test to confirm or exclude the suspected diagnosis. This represents selection bias and is inherent in many systematic reviews and meta-analyses which study physical examination findings. Confirmation bias was mitigated by including only studies that performed the physical examination prospective to the reference standard test performance. This ensures that all the study examiners were blinded to the results of the reference standard test. We found a moderate degree of heterogeneity in all our included studies which warranted the use of a random effects model for our meta-analyses (Figs. 2 and 3). Sensitivity analysis was used for each meta-analysis we performed. We were unable to identify any outliers with the “leave-one-out” sensitivity analysis, implying that the influence of heterogeneity on the effect size was low (Supplement 2). Using four variables from a modified version of the Newcastle–Ottawa Scale for assessing the quality of non-randomized studies in meta-analyses, we assessed the risk of bias for all our included studies (Table 4) [13]. Of our twenty included studies eight of them were deemed to be low risk of bias and the remaining twelve were high risk of bias (Table 4). The most significant source of bias was the lack of control groups in eleven of our included studies. The quality of evidence for all our results were low, given the observational nature of all our included studies according to the Grading of Recommendations, Assessment, Development, and Evaluation (GRADE) assessment tool [14].

Table 4.

Modified Newcastle–Ottawa quality assessment score

First Author, year	Representativeness of cases or cohort is consecutive or a series of cases	Cases clearly defined	Selection of control patients from the community	Independent blind assessment of outcome
Calis, 2000 [16]	1	1	1	1
Litaker, 2000 [17]	1	1	0	1
Park, 2005 [18]	1	1	0	1
Kim, 2007 [19]	1	1	0	1
Miller, 2008 [20]	1	1	0	1
Silva, 2008 [21]	1	1	0	1
Fodor, 2009 [22]	1	1	0	1
Michener, 2009 [23]	1	1	0	1
Bak, 2010 [24]	1	1	1	1
Fowler. 2010 [25]	1	1	0	1
Kelly, 2010 [26]	1	1	0	1
Toprak, 2013 [27]	1	1	0	1
Van Kampen, 2014 [28]	1	1	1	1
Villafane, 2015 [29]	1	1	1	1
Penning 2016 [30]	1	1	0	1
Jain, 2017 [31]	1	0	1	1
Jain, 2018 [32]	1	1	1	1
Kappe, 2018 [33]	1	1	1	1
Yazigi Junior, 2021 [34]	1	1	1	1
Zou, 2022 [35]	1	1	1	1

A score of 0 is equal to a high risk of bias and a score of one is equal to a low risk of bias

In conclusion, we present a robust body of low-quality evidence for the diagnostic accuracies of the most commonly used physical examination tests for identifying both RCT and SIS. We have identified novel data for the accuracy of the External Rotation Lag Sign at 90 degrees and the Internal Rotation Lag Sign, which have high to moderate diagnostic accuracy for ruling in RCTs. The clinical implications of accurate bedside diagnostic studies include a potential reduction in disease burden and clinical costs with the improved patient outcomes and satisfaction.

Abbreviations

PCP

Primary care physician

RCT

Rotator cuff tear

SIS

Subacromial impingement syndrome

PICOS

Patient, intervention, comparison, outcomes, and study design

PRISMA

Preferred Reporting Items for Systematic Reviews and Meta-Analysis

PROSPERO

Prospective Register of Systematic Reviews

DOR

Diagnostic odds ratio

LR

Likelihood ratio

Richard A Shellenberger

Richard A Shellenberger, DO is the Associate Program Director in the Internal Medicine Residency Program at Trinity Ann Arbor Hospital in Ann Arbor, MI, USA and can be contacted at Richard.Shellenberger@trinity-health.org.

Authors’ contributions

RS, QZ and CD conceived the study design. RS and QZ designed and conducted the systematic review. RS and CD developed the search strategies. SW and RS conducted the meta-analysis. RS and QI performed the quality assessment. NK, PP, MT, MO and RS extracted study information and data for outcomes measures. SW performed the statistical analysis. All authors assisted in writing and manuscript and approved the present version for submission.

Funding

This study received no funding.

Data availability

Data are available through electronic Supplementary Material and upon request of the Corresponding Author.

Declarations

Ethics approval and consent to participate:

Not applicable.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.