Clinical course of pain and function in subacromial shoulder pain: a systematic review with meta-analysis

Abstract

Objective

To assess the clinical course of pain and function in patients with subacromial pain.

Design

Systematic review.

Methods

We searched Medline, Embase, AMED, Web of Science, Cochrane Library, and Scopus for randomized clinical trials and observational studies recruiting participants with subacromial pain. Pain and function scores were extracted for groups receiving either ‘no intervention’ or ‘usual care’. Changes in pain and function from baseline were pooled at 6 weeks, 3, 6, and 12 months follow-up. Random effects meta-analyses and meta-regression were performed to assess the clinical course of pain and function in subacromial pain.

Results

Nineteen studies were included and 17 studies were pooled. Data from 5 studies in the ‘no intervention’ group were pooled to 3 months, showing minimal improvement for pain (mean difference, MD, 5.3/100, 95%CI −0.8 to 11.4) and function (MD 3.1/100, 95%CI −1.7 to 7.9). Data from 12 studies in ‘usual care’ group were pooled to 12 months, showing significant improvement for pain (MD 32.5/100, 95%CI 22.6 to 42.3) and function (MD 30.3/100, 95%CI 24.4 to 36.1), with approximately 40% of this gain in the first 6 weeks.

Conclusion

With ‘no intervention’, participants with subacromial pain are unlikely to show changes for pain and function within 3 months. For participants receiving ‘usual care’, recovery continued up to 12 months with almost 40% of this change during the first 6 weeks of care. Individuals with subacromial pain who received ‘usual care’ may have moderate changes in terms of minimal clinical important differences up to 12 months.

PROSPERO registration number CRD42016052518.

Keywords:

• Clinical course
• shoulder pain
• usual care
• no intervention
• systematic review

Introduction

Shoulder pain is the third most common musculoskeletal disorder [1]. The annual prevalence of shoulder complaints ranges from 10% to 28% [2–4]. Subacromial shoulder pain is the most common shoulder disorder [5, 6] representing 89% of patients referred to general practitioners and physiotherapists for shoulder pain [7]. The umbrella term of subacromial shoulder pain includes anterolateral shoulder pain, non-specific shoulder pain, rotator cuff tendinopathy, shoulder impingement syndrome, and subacromial pain syndrome [8, 9].

Considerable variation in the rate of recovery has been reported for patients with subacromial shoulder pain [10–13]. Between 23% and 50% of patients with a new episode of subacromial shoulder pain improved after six months [12, 13]. Bonde, Mikkelsen [10] reported that 50% of the construction workers in their study improved after ten months. Cummins, Sasso [11] reported that 23% of patients with subacromial shoulder pain still have persistent shoulder pain after two years. Variable recovery rates and timeframes reported by previous studies highlight the need to increase our understanding of expected rates of improvement for subacromial shoulder pain over time.

Observational studies reporting changes in pain and function scores in subacromial shoulder pain are scarce and limited their recruitment to those participants who received treatment within primary health care [14, 15]. We lack observational studies to report the clinical course of subacromial pain for those patients without treatment, while 50% to 80% of patients with shoulder pain do not seek any treatment in primary care [12, 16]. Recent systematic reviews summarized prognostic factors for subacromial shoulder pain but did not report the clinical course [17–19]. Only one systematic review has reported the natural course of frozen shoulder in patients who did not receive treatment [20]. Summarizing the course of pain and function for research participants who have not received interventions, or for those who received primary care (usual care), may help understand the expected recovery rates in the community.

Pain and function scores are the two most commonly reported variables in the studies assessing shoulder pain. Two previous studies indicated a strong correlation between pain and function scores, however, they had cross-sectional designs and reported the association between pain and function in patients with multiple musculoskeletal diseases [21, 22]. To our knowledge, no study has explored the association between pain and function scores over time in patients with subacromial shoulder pain. The illustration of the patterns of changes between pain and function scores (in ‘no intervention’ and ‘usual care’ groups) would provide us with evidence-based information on these correlations over time.

We aimed to assess the clinical course of pain and function in patients with subacromial shoulder pain who received ‘no intervention’ or ‘usual care’. The second aim was to explore the patterns of change for pain and for function over time for each treatment group (i.e. ‘no intervention’ and ‘usual care’).

Methods

The Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) statement was followed for reporting this systematic review [23]. The protocol was registered at the International Prospective Register of Systematic Reviews (PROSPERO: CRD42016052518) and published in BMJ Open [24].

Study characteristics

Studies were assessed for eligibility based on the following criteria: (1) enrolled participants with subacromial shoulder pain; (2) published in any languages; (3) randomized clinical trials, experimental studies, or prospective cohort studies; (4) enrolled participants receiving ‘usual care’ or ‘no intervention’; (5) reported pain and function scores for follow-ups longer than 2 weeks.

Types of intervention

Studies offering ‘no intervention’ or ‘usual care’ groups were considered eligible. ‘No intervention’ referred to participants allocated to waiting lists or control groups receiving no form of treatment in a trial. ‘Usual care’ was defined as any form of primary healthcare treatment routinely prescribed by healthcare professionals for patients with subacromial shoulder pain [14, 25, 26]. ‘Usual care’ could include education, advice, oral analgesic, non-steroid anti-inflammatory drugs, physiotherapy, corticosteroid injection, massage and exercise prescription, as long as authors clearly stated that these interventions were part of ‘usual care’ (i.e. ‘usual care’ was delivered based on health provider decision making) [25]. For this purpose, studies in which participants followed a specific protocol for intervention were excluded. We excluded placebo or sham interventions from this review. Placebo or sham interventions can have effects on pain and function, thereby potentially influencing outcomes compared with individuals in the community who sought ‘no intervention’ or who received ‘usual care’ [27, 28]. We did not undertake statistical analysis to compare the clinical course for patients receiving ‘no intervention’ with those receiving ‘usual care’. In each category, the pattern of changes in pain and function scores were reviewed over time. We suggest that the clinical course of pain and function with ‘no intervention’ or ‘usual care’ is relevant for people living with subacromial shoulder pain in the community context.

Types of outcome measure

Studies were included if they reported outcome measures of pain or function, further defined under ‘Data Collection’ section. Examples of outcome measures for assessing shoulder pain are Visual Analogue Scale (VAS), Numeric Pain Rating Scale (NPRS) or Shoulder Pain and Disability Index for pain, (SPADI-Pain) for pain scores. For function, we considered outcome measures such as SPADI-Total, or Shoulder Rating Questionnaire (SRQ) for function scores.

Participants

Studies with participants aged 18 years or older with subacromial shoulder pain were included. Subacromial shoulder pain was defined as pain at the shoulder with or without pain in the neck and elbow that worsens during arm movements especially overhead activities [25, 29]. We included diagnostic terms of subacromial shoulder pain, subacromial pain disorder, rotator cuff tendinopathy, rotator cuff syndrome or disease, rotator cuff-related pain syndrome, rupture or tears of rotator cuff muscles, subacromial bursitis, anterolateral shoulder pain or non-specific shoulder pain [8]. Studies with >5% of participants with shoulder pain due to neurological disorders, hemiplegia following stroke, capsulitis, severe trauma, shoulder fracture, instabilities, history of shoulder surgeries, and concomitant systemic disorders (e.g. diabetes, cancer and rheumatoid disorders) were excluded [30, 31].

Search strategy

We conducted electronic searches using Medline, Embase and AMED (via Ovid), Web of Science, Cochrane Library, and Scopus from the date of inception to 22^nd September 2021. Our search strategy included three search categories (i.e. shoulder pain, intervention type, and prognosis). The full search strategy is described in the Appendix.

Study selection

After deleting duplicates, two reviewers (ZJT and DCR) independently screened titles and abstracts and full texts for eligibility criteria. In the case of disagreement between reviewers, the third or fourth reviewers (ASG and GS) were consulted and a final decision was made by consensus.

Data Collection

Two reviewers (ZJT and ASG) independently extracted the following information: design, country, sample size, participant’s characteristics, symptoms’ duration, group allocation (‘no intervention’, and ‘usual care’), follow-up time points, outcome measures, and mean pain and function scores. In the case of insufficient reported data, ZJT contacted authors for the additional data and waited one month for a reply.

To avoid selection bias, we adopted the following criteria for selecting the outcome measures for studies that reported more than one outcome measure for pain or function [32]. First, we identified the most frequently used outcome measure in the included studies. Second, if a study did not use the most frequently used outcome measure but reported more than one outcome measure for pain or function, we selected the outcome measure with the highest correlation based on past research with the most commonly reported outcome measure used by other included studies. With respect to pain, the first option was to use average pain (during 24 h or during the last week) measured with the VAS or the NPRS. If average pain was not reported, ‘pain at rest’ was the second option that was favored over ‘pain during movement’, as the latter is likely to be influenced by the level of function or disability [33]. With respect to function outcome measures, where possible we opted for extracting the following outcome measures: the SPADI-Total, the SRQ or the Disabilities of Arm, Shoulder, and Hand (DASH) questionnaire. We opted for these outcomes as those are strongly correlated with each other [34, 35]. For the studies that reported only one outcome measure for pain or function, we used those data for the meta-analysis. Any uncertainties for prioritizing the outcome measures were discussed and agreed among the team.

Risk of bias within included studies

Two reviewers (ZJT and ASG) independently assessed the risk of bias of the included studies. The risk of bias of RCTs was assessed using the Cochrane Risk of Bias tool [36]. This tool has six domains (i.e. selection bias, performance bias, detection bias, attrition bias, reporting bias, and other sources of bias) and each domain is rated as ‘low’, ‘high’ or ‘unclear’ risk of bias. For the purpose of this review, a drop-out greater than 20% was considered as ‘high risk’ when grading attrition bias [37]. For other domains, we followed the Cochrane instructions for RCT studies [36]. When assessing performance bias domain (blinding of personnel and participants), a low risk of bias was considered for all included studies as 1) blinding was not possible for personnel and participants in the ‘no intervention’ arms, and 2) this review’s aims and design did not include comparisons between two groups. Therefore, as described by the Cochrane Handbook [38], performance bias can be judged low risk when non-blinding of participants was unlikely to influence the outcome measure.

Risk of bias within observational studies was assessed using a tool proposed by Altman [39, 40]. This tool has two items focusing on sampling, two on follow-up, and one item on prognostic outcomes. Each item was categorized with ‘yes’, ‘no’, or ‘unclear’. Disagreements between reviewers were resolved by discussion and, if needed, by consulting with a third reviewer (DCR or GS). The overall risk of bias for RCT and observational studies was reported individually for each domain and considered when interpreting data.

Strategy for data synthesis

We summarized pain and function scores for patients who received ‘no intervention’ or ‘usual care’. When possible, we summarized ‘acute’ or ‘persistent’ subacromial shoulder pain based on mean or median duration of symptom at baseline. Pain symptoms lasting less than three months in duration were considered ‘acute’ [14, 15, 41], whereas, symptoms lasting for three months and longer were considered ‘persistent shoulder pain’ [14, 26]. When sufficient information was not available regarding the duration of symptoms of participants, ZJT contacted the corresponding authors for further information. In the case of no response from authors, that study was allocated to persistent shoulder pain studies. This decision was made as many patients with shoulder complaints seen by GPs have recurrent complaints [42].

To standardize the outcome measures, we converted the dependent outcome (pain and function at baseline and follow-ups) into a common 0–100 scale [38] in which higher scores represented higher levels of pain or functional impairment. The conversions were performed using the following equation:$\mathrm{ }x=\frac{\mathrm{mean}\mathrm{*}100}{\text{full score}}$ (X indicates the converted score, the mean is the reported score of an outcome measure, and the full score refers to the total score of an outcome measure). For outcome measures with inverted scores (e.g., PSFS where greater scores signify better function), we reversed the scores first and then used these in the equation given above. Next, we calculated the changes in pain and function scores. This was performed by detracting follow-up score from baseline (change score = baseline –follow-up) so that change scores with a positive sign referred to clinical improvements in pain and function. We consulted with a biostatistician to define the method of standardizing the outcome measure and calculating change scores. Presenting the pooled estimate based on mean difference is preferable to Standard Mean Difference (SMD) as this method provides clinically relevant difference for pain and function scores that could be potentially used for future studies as comparisons. This is more challenging when using SMDs.

Statistical analysis

Meta-analyses were conducted using RevMan 5.3 (Copenhagen: The Nordic Cochrane Centre, The Cochrane Collaboration, 2014). Meta-analyses were performed to analyze changes in pain or function scores between baseline and follow-ups (expressed as mean and SD), and mean and SD differences for studies that provided only mean and SD differences. Follow-ups were categorized as 6 weeks (2 to 6 weeks), 3 months (more than 6 weeks to 3 months), 6 months (more than 3 to 6 months) and 12 months (more than 6 to 12 months) from study entry. Generic inverse variance was used for calculating changes in pain or function scores at follow-ups from baseline, for each time point category. A random effects model was used to allow for different treatment effects between-studies during meta-analyses (e.g. variance due to differences in participants’ clinical characteristics or sequence of treatment) [43]. Given all scores were converted into one scale ranging from 0 to 100, effect sizes were reported as mean differences and 95% confidence intervals [38, 44, 45]. The percentage of total variability due to between-study heterogeneity was assessed using I² [38] where I² less than ‘30%’ was considered as not important heterogeneity; between ‘30–60%’ considered as moderate; ‘60 to 90%’ considered as substantial; and ‘90 to 100%’ considered as considerable heterogeneity [38]. Subgroup analyses were conducted based on duration of shoulder pain. Baseline pain and function scores in each category (i.e. ‘no intervention’ and ‘usual care’) were pooled. Pooling data was performed in Excel 2013 with a formula described in the Cochrane Handbook for combining data from two or more groups [38].

Meta-regression was conducted only where a minimum of 10 studies were presented for the same score [46, 47]. All available follow-up time points were used in the analysis. A restricted cubic spline meta-regression was used to model the non-linear relationship between time (considered as a continuous variable) and pain scores [48, 49]. We used a restricted maximum likelihood to estimate the heterogeneity variance parameter. Meta-regression was implemented using the ‘rms’ package [50] from R [51].

Results

Study selection

The original search resulted in 4,260 studies. After removal of duplicates, the titles and the abstracts of 3,805 potentially relevant studies were screened and the full text of 93 articles was included and read. Eventually, nineteen studies (15 RCTs and 4 observational studies) were included in the qualitative synthesis (Figure 1). Seventeen studies were included in the meta-analyses and two studies were only assessed qualitatively [52, 53]. Out of those two, one study presented data for participants who received ‘no intervention’. In that study, the follow-up lasted until 6 months [52], while in other studies allocated for meta-analyses for ‘no intervention’ groups, participants were followed up to 3 months [54–58]. Another study reported findings for participants receiving ‘usual care’. In that study, the data were reported as median and interquartile range [53]. There are guidelines to convert the median and interquartile range to mean and SD [59], but we decided to report this data qualitatively as when the data is skewed, it is reported as median and interquartile range [38]. Ten studies reported pain outcome measures for participants receiving ‘usual care’, and were included in the meta-regression [6, 14, 25, 26, 41, 60–64]. All studies were published in English. We contacted the authors of 19 studies, requesting additional information about their methods or data of baseline and follow-ups [6, 14, 15, 26, 29, 52–54, 58, 60–62, 65–71]. Eight authors responded [6, 14, 15, 29, 54, 62, 67, 71]. Two authors informed us that the requested data were not available [14, 15], and we excluded one of those due to lack of data [15].

Figure 1. Flow of studies in the review. A diagram of the stages passed in the systematic review. It includes the number of articles screened (n = 3805), the full text of articles reviewed for eligibility (n = 93), the articles included in the qualitative synthesis (n = 19) and the meta-analysis (n = 17).

Study characteristics

A detailed summary of study characteristics with description of changes in pain and function in each time point is provided at Table 1.

Table 1. A summary of study characteristics.

Study	Country	Participants	Shoulder pain duration	Group allocation	Outcome measure	Follow-up time points	Mean pain at baseline and follow-ups (0-100)	Mean function (0-100)
Bergman, Winters [25]	The Netherlands	N = 71	39% < 6 W	Usual care: information, advice, medication, corticosteroid injections and physiotherapy	Pain (Shoulder Pain Score 0-21)	Baseline	27.1 ± 26.6
		34 men	32% = 6 W			6 weeks	18.5
		37 women	15% = 12 W			3 months	17.6
		Age:	14% > 26 W			6 months	16.7
		47.8 ± 11.8
Bron, De Gast [58]	The Netherlands	N = 35	16% 6-9 M	No intervention	Pain (average pain VAS 0-10)	Baseline 6 weeks 3 months	43.4 ± 17	30.8 ± 11.9
		9 men	26% 9-12 M		Function (DASH 0-100)		40.0 ± 20.7	27.5 ± 15.5
		26 women	19% 1-2 Y				33.2 ± 23.3	26.1 ± 13.8
		Age:	16 % 2-5 Y
		45 ± 13.2	23% > 5 Y
Choi, Lee [54]	South Korea	N = 11	Chronic	No intervention	Pain (VAS 0-10)	Baseline	58.2 ± 26.4
		11 women				4 weeks	58.2 ± 24.8
		Age:
		46 ± 5
Dickens, Williams [52]	The UK	N = 40	Chronic	No intervention	Function (Constant-Murley Score 0-100)	Baseline	56.0
		22 men				6 months	64.3
		18 women
		Age:
		54 (26 to 73)
Geraets, Goossens [26]	The Netherlands	N = 89	Chronic	Usual care:	Pain (VAS 0-10)	Baseline	54.0 ± 18	65.6 ± 19.9
		38 men		Medical information, pain medicine, wait-and-see, corticosteroid injections (PRN), referral to orthopedic surgeon or physiotherapy	Function (Shoulder Disability Questionnaire 0-100)	3 months	39.0	50.3
		51 women
		Age:
		53.3 ± 11.7
Kachingwe, Phillips [60]	The USA	N = 7	70 ± 92.4 M	Usual care: Physician advice – postural awareness, limitation of overhead activities, home exercise without any input from physiotherapist	Pain in the past 24	Baseline	44.0 ± 12	49.4 ± 29.3
		4 men			hours (VAS 0-10)	6 weeks	37.7	32.6
		3 women			Function (SPADI-Total 0-130)
		Age:
		45.6 ± 13
Littlewood, Malliaras [71]	The UK	N = 12	49 (3-168) M	Usual care:	Function (SPADI-Total 0-100)	Baseline		39.7 ± 18.3
		7 men		physiotherapy		3 months		20.7 ± 20.3
		5 women
		Age:
		63.9 (44-79)
Littlewood, Bateman [29]	The UK	N = 44	17.5(3-120) M	Usual care: physiotherapy	Function (SPADI-Total 0-100)	Baseline		49.0 ± 18
		26 men				3 months		30.7 ± 19.7
		18 women				6 months		24.0 ± 19.7
		Age:				12 months		21.4 ± 25.4
		55.6 (23-80)
Lombardi, Magri [93]	Brazil	N = 30	13.9 ± 9.6 M	No intervention	Pain at rest (VAS 0-10)	Baseline	39.0 ± 26	44.8 ± 18.3
		5 men			Function (DASH 0-100)	2 months	43.0 ± 32	43.4 ± 22.8
		25 women
		Age:
		54.8 ± 9.4
Ludewig and Borstad [56]	The USA	N = 33	Chronic	No intervention	Pain (Work Related Pain Score, 0-100)	Baseline	46.0 ± 2.8	38.0 ± 3
		33 men			Function (SPADI-Function 0-100)	2.5 months	41.0 ± 2.9	37.0 ± 3
		Age:
		49.2 ± 1.8
Miller and Osmotherly [94]	Australia	N = 12	17 (12-62.5) M	Usual care: physiotherapy	Pain (SPADI-Pain 0-100)	Baseline	53.3 ± 25.6	48.9 ± 28
		7 men			Function (SPADI-Total (0-100)	6 weeks	26.6 ± 44.4	26.2 ± 43.7
		5 women
		Age:
		54.5 (45.5–62.5)
Ottenheijm, Cals [41]	The Netherlands	N = 53	5.5 ± 3.5 W	Usual care: stepwise wait and see with advise and analgesic for 2 weeks, corticosteroid injection, referral to physiotherapy and hospital specialist	Pain (Shoulder Pain Score 7-28)	Baseline	59.5 ± 20.4	68.3 ± 17.2
		33 men			Function (Shoulder Disability Questionnaire 0-100)	3 months	39.0 ± 26.1	49.4 ± 29.8
		20 women				6 months	30.9 ± 25.7	36.3 ± 27.1
		Age:				12 months	21.4 ± 29.8	31.0 ± 29.7
		49.4 ± 10.9
Van den Dolder and Roberts [55]	Australia	N = 14	30 (23-91) W	No intervention	Pain over last 24 h (VAS 0-100)	Baseline	53.9 ± 21.6	63.6 ± 18.3
		9 men			Function (PSFS 0-130)	2 weeks	53.8 ± 26.3	65.4 ± 18.6
		5 women
		Age:
		65.9 ± 9.2
Zebis, Andersen [64]	Denmark	N = 112	Chronic	Usual care: advise to stay physically active and were consulted once a week with physiotherapists	Pain (Nordic Pain Questionnaire 0-9)	Baseline	53.0 ± 20.1
		20 men				20 weeks	26.4 ± 28.8
		92 women
Zhang, Sun [63]	China	N = 42	8.5 ± 3.9 M	Usual care: conventional orthopedic treatment	Pain, average (VAS 0-100)	Baseline	64.0 ± 1.4	41.0 ± 7.9
		19 men			Function (DASH 0-100)	2 weeks	58.6 ± 8.7	38.3 ± 7.3
		23 women				4 weeks	51.2 ± 15.5	35.6 ± 9.4
		Age:				2 months	51.9 ± 17.8	32. 4 ± 12
		45.6 ± 7.6				4 months	51.7 ± 17.2	26.0 ± 12.2
Laslett, Steele [61]	New Zealand	N = 161	108.4 ± 193 D	Usual care:	Pain, mean (VAS 0-100)	Baseline	36.4 ± 20.6	38.0 ± 20.1
		79 men		corticosteroid injection for the first 3 weeks and physiotherapy, corticosteroid injection, or no treatment as appropriate	Function (SPADI-Total 0-100)	3 weeks	13.2 ± 20	16.1 ± 19.2
		82 women				3 months	14.7 ± 20.4	17.5 ± 21
		Age:				6 months	15.0 ± 21.5	17.0 ± 20.2
		44 ± 13.88				12 months	8.4 ± 17.6	10.1 ± 16.4
Roberts and Li [6]	Australia	N = 88	4.4 ± 4.5 M	Usual care: physiotherapy	Pain (NPRS 0-10)	Baseline	59.0 ± 42
		41 men				6 months	19.0
		47 women
		Age:
		61 ± 15.7
Reilingh, Kuijpers [14]	The Netherlands	N = 205	< 6 w	Usual care:	Pain (NPRS 0-10)	Baseline	43.0 ± 26
		100 men		advice, medication, steroid injection or physiotherapy		6 months	13.0 ± 22
		105 women
		Age:
		49.5 ± 14.7
Winters, Sobel [62]	The Netherlands	N = 58	Chronic	Usual care:	Pain (Shoulder Pain Score 7-28)	Baseline	42.3 ± 22.2
		Age:		Two weeks NSAID, and usual therapeutic option including physiotherapy, injection, and manipulation therapy, NSAID and Paracetamol		4 weeks	35.8 ± 21.2
		47.3 ± 15.4				26 weeks	12.0 ± 15.8

N = Number; D = days; W = weeks; M = Months; DASH = Disability of Arm Shoulder and Hand; SPADI = Shoulder Pain and Disability Index; PSFS = patient specific functional scale; NPRS = numeric pain rating scale.

Nineteen studies yielded a total of 1,117 participants with subacromial shoulder pain (45.5% men and 54.5% women). Six studies included 163 participants who received ‘no intervention’ [52, 54–58]. Thirteen studies recruited 954 participants who received ‘usual care’ [6, 14, 25, 26, 29, 41, 53, 60–64, 71]. Mean ages for all participants ranged between 44.0 and 65.9 years. The data related to shoulder pain duration was not reported in six studies so we considered them as persistent shoulder pain [26, 52, 54, 56, 62, 64].

Six studies included participants with persistent shoulder pain who received ‘no intervention’ or who were allocated to ‘wait and see’ groups, of which five of these studies were pooled for meta-analyses [54–58]. Thirteen studies included participants who received ‘usual care’, twelve of these were pooled for meta-analyses [6, 14, 25, 26, 29, 41, 60–64, 71]. Of these thirteen studies, eleven studies recruited participants with persistent pain [6, 25, 26, 29, 53, 60–64, 71], and two studies recruited participants with acute pain [14, 41].

‘Usual care’ consisted of the following: (1) ‘multimodal usual care’ which included advice, medication, corticosteroid injection and referral to physiotherapy or specialist [14, 25, 26, 41, 61, 62]; (2) ‘exercise and advice’, which included advice to stay physically active and limit overhead activities, home exercise without any input from physiotherapists, or heat or cold therapy [60, 63, 64]; and (3) ‘conventional physiotherapy intervention’, which included advice, stretching, specialized exercise, manual therapy, electrotherapy, and taping at the discretion of physiotherapists [6, 29, 53, 71].

Outcome measures

The outcome measures used in our meta-analyses are described in Table 1. The most common outcome measures used for assessing pain were VAS and NPRS, and SPADI-Total was the most commonly outcome measure used for assessing function.

Risk of bias within included studies

The risk of bias scores of RCTs and observational studies are presented in Tables 2 and 3. Seven studies were evaluated with low risk of bias [25, 26, 55, 57, 64] or unclear risk of bias in one domain [41, 60]. Eight studies were evaluated as high risk of bias in incomplete outcome data [29, 52, 63], blinding outcome assessment [29, 56, 71], selective reporting [58], and other biases [53, 54]. All observational studies were rated as low risk of bias [6, 14, 61, 62].

Table 2. Risk of bias for RCTs.

Study	Random sequence generation	Allocation concealment	Blinding of participants and personnel	Blinding of outcome assessment	Incomplete outcome data	Selective reporting	Other bias
Bergman, Winters [25]	Low	Low	Low	Low	Low	Low	Low
Bron, De Gast [58]	Low	Unclear	Low	Low	Low	High	Low
Choi, Lee [54]	Unclear	Unclear	Low	Unclear	Low	Low	High
Dickens, Williams [52]	Low	Unclear	Low	Low	High	Low	Low
Geraets, Goossens [26]	Low	Low	Low	Low	Low	Low	Low
Kachingwe, Phillips [60]	Low	Unclear	Low	Low	Low	Low	Low
Littlewood, Malliaras [71]	Low	Low	Low	High	Low	Low	Low
Littlewood, Bateman [29]	Low	Low	Low	High	High	Low	Low
Lombardi, Magri [93]	Low	Low	Low	Low	Low	Low	Low
Ludewig and Borstad [56]	Unclear	Low	Low	High	Low	Low	Low
Miller and Osmotherly [94]	Low	Low	Low	Low	Low	Low	High
Ottenheijm, Cals [41]	Low	Unclear	Low	Low	Low	Low	Low
Van den Dolder and Roberts [55]	Low	Low	Low	Low	Low	Low	Low
Zebis, Andersen [64]	Low	Low	Low	Low	Low	Low	Low
Zhang, Sun [63]	Low	Unclear	Low	Low	High	Low	Low

Table 3. Risk of bias for observational studies.

Study	Defined sample^a	Representative sample^b	Follow-up rate^c	Follow-up time^d	Prognosis^e
Laslett, Steele [61]	Yes	Yes	Yes	Yes	Yes
Roberts and Li [6]	Yes	Yes	Yes	Yes	Yes
Reilingh, Kuijpers [14]	Yes	Yes	Yes	Yes	Yes
Winters, Sobel [62]	Yes	Yes	Yes	Yes	Yes

^a Description of source of population, inclusion and exclusion criteria.

^b Participants were selected as consecutive.

^c Outcome data were available for at least 80% of participants at one follow-up point.

^d At least one prognostic outcome was followed up at three months or later.

^e Raw data, percentages, survival rates or continuous outcome reported.

Meta-analysis

‘No intervention’

The baseline mean and SD for pain and function scores for participants receiving ‘no intervention’ were 45.5 ± 39.0 and 41.1 ± 36.6 (out of a maximum score of 100), respectively. For participants with persistent pain receiving ‘no intervention’, the pooled mean differences for pain and function were analyzed at two-time point categories: up to 6 weeks, and from 6 weeks to 3 months after study entry. For both time points, no significant improvements were observed for pain or function scores. The heterogeneity in all meta-analyses was minimal (I² < 15%) (Figures 2–5). Only one study reported six-month outcomes, thus could not be pooled [52]. The data from that study suggested an improvement of 0.6 (SD = 7.5) on the Constant-Murley score at 6 months [52]. The patterns of changes between pain and function scores were similar over three months (Figure 6).

Figure 2. Mean difference in pain up to 6 weeks after study entry for participants receiving ‘no intervention’. Forest plots showing pain improved to 0.76/100 from baseline to six weeks in participants with ‘no intervention’.

Figure 3. Mean difference in pain more than 6 weeks to 3 months after study entry for participants receiving ‘no intervention’. Forest plot showing pain improved to 5.31/100 from baseline to three months in participants with ‘no intervention’.

Figure 4. Mean difference in function up to 6 weeks after study entry for participants receiving ‘no intervention’. Forest plot showing function improved to 2.3/100 from baseline to six weeks in participants with ‘no intervention’.

Figure 5. Mean difference in function more than 6 weeks to 3 months after study entry for participants receiving ‘no intervention’. Forest plot showing function improved to 3.1/100 from baseline to three months in participants with ‘no intervention’.

‘Usual care’

The baseline mean and SD for pain and function scores for participants receiving ‘usual care’ were 46.3 ± 43.1 and 49.5 ± 44.4 (out of a maximum score of 100), respectively. For participants with acute and persistent pain receiving ‘usual care’, pooled mean differences for pain scores from baseline showed a change of 12.3/100 at 6 weeks, 15.7/100 up to 3 months, 24.3/100 up to 6 months and 32.5/100 at 12 months (Figures 7–10). Pooled mean differences for function scores from baseline, showed a change of 13.4/100 at 6 weeks, 16.2/100 up to 3 months, 23.8/100 up to 6 months and 30.3/100 at 12 months (Figures 11–14). Heterogeneity for these analyses ranged from moderate (I² = 39%) to considerable (I² = 89%). One study reported outcomes for pain and function using median and interquartile range (IQR) [53] and showed improvements in pain score of 26 (IQR = 36) and in function score of 22 (IQR = 37) at three months for participants receiving ‘conventional physiotherapy intervention’. Overall, participants receiving ‘usual care’ presented a similar pattern of improvement between pain and function at each time category (Figure 15). A summary of changes of pain and function in participants with ‘no intervention’ and ‘usual care’ is provided in Table 4.

Figure 6. Clinical course of pain and function scores (mean differences between baseline and follow-up, 95% confidence intervals) for participants receiving ‘no intervention’. Higher scores indicate greater improvements. A linear graph showing similar patterns of improvements between pain and function over three months in participants who remained with ‘no intervention’. Overall, together the changes were minor for pain and function scores with maximum improvement of 5/100 scores at three months.

Figure 7. Mean difference in pain up to 6 weeks after study entry in participants receiving ‘usual care’. Forest plot showing pain improved to 12.3/100 from baseline to six weeks for participants receiving ‘usual care’.

Figure 8. Mean difference in pain, 6 weeks to 3 months after study entry in participants receiving ‘usual care’. Forest plot showing pain improved to 15.7/100 from baseline to three months for participants receiving ‘usual care’.

Figure 9. Mean difference in pain, 3 to 6 months after study entry in participants receiving ‘usual care’. Forest plot showing pain improved to 24.3/100 from baseline to six months for participants receiving ‘usual care’.

Figure 10. Mean difference in pain, 6 to 12 months after study entry in participants receiving ‘usual care’. Forest plot showing pain improved to 32.5/100 from baseline to 12 months for participants receiving ‘usual care’.

Figure 11. Mean difference in function up to 6 weeks after study entry in participants receiving ‘usual care’. Forest plot showing function improved to 13.4/100 from baseline to six weeks for participants receiving ‘usual care’.

Figure 12. Mean difference in function, more than 6 weeks to 3 months after study entry in participants receiving ‘usual care’. Forest plot showing function improved to 16.2/100 from baseline to three months for participants receiving ‘usual care’.

Figure 13. Mean difference in function, more than 3 to 6 months after study entry in participants receiving ‘usual care’. Forest plot showing function improved to 23.8/100 from baseline to six months for participants receiving ‘usual care’.

Figure 14. Mean difference in function, more than 6 to 12 months after study entry in participants receiving ‘usual care’. Forest plot showing function improved to 30.3/100 from baseline to 12 months for participants receiving ‘usual care’.

Figure 15. Clinical course of pain and function scores (mean differences between baseline and follow-up, 95% confidence intervals) for participants receiving ‘usual care’. Higher scores indicate greater improvements. A linear graph showing the similar patterns of improvement between pain and functions for participants receiving ‘usual care’ throughout 12 months follow-up. In this graph, pain and function improved rapidly to approximately 15/100 at 6 weeks, remained constant to 3 months and improved gradually to around 30/100 at 12 weeks.

Table 4. A summary of changes in pain and function from baseline in ‘no intervention’ and ‘usual care’ (score ranging from 0 to 100).

‘No intervention’
	Baseline (mean ± SD)	Changes up to 6 weeks (95%CI)	Changes up to 3 months (95% CI)	Changes up to 6 months (95%CI)	Changes up to 12 months (95%CI)
Pain	45.5 ± 39	0.76 (-3.6 to 5.2)	5.31 (-0.8 to 11.4)	____________	____________
Function	41.1 ± 36.6	2.3 (-3.8 to 8.4)	3.11 (1.7 to 7.9)	____________	____________
‘Usual care’
	Baseline (mean ± SD)	Changes up to 6 weeks (95%CI)	Changes up to 3 months (95% CI)	Changes up to 6 months (95%CI)	Changes up to 12 months (95%CI)
Pain	46.3 ± 43	12.3 (5.4 to 19.1)	15.7 (10.6 to 20.9)	24.3 (18.0 to 30.5)	32.5 (22.6 to 42.3)
Function	49.5 ± 44.4	13.4 (0.6 to 26.3)	16.2 (11.5 to 21.0)	23.8 (16.6 to 31.0)	30.3 (24.4 to 36.1)

The changes of pain and function in ‘no intervention’ group were pooled from participants with persistent shoulder pain. The changes of pain and function in ‘usual care’ group were pooled from participants with acute and persistent shoulder pain together.

Subgroup analysis

Subgroup analyses were performed for participants receiving ‘usual care’ based on the duration of symptoms (Table 5). For studies of participants with acute subacromial shoulder pain, the mean pain at baseline was 46.5 ± 36.3 and for participants with persistent shoulder pain, the mean pain at baseline was 45.2 ± 40.1 (out of a maximum score of 100). We pooled data from two studies [14, 41] that reported data from participants with acute pain, and ten studies that reported participants with persistent subacromial shoulder pain [6, 25, 26, 29, 60–64, 71]. The two included studies reporting acute shoulder pain did not report function scores. The heterogeneity was minimal for subgroup analysis of acute shoulder pain (I² = 0%) but varied from minimal to considerable for subgroup analyses of persistent shoulder pain (I² = 0% to 93%). Results showed pain improved to 29.7 (25.3 to 34.7) points in patients with acute shoulder pain and 22.6 (15.1 to 30.2) points in patients with persistent shoulder pain at 6 months.

Table 5. Subgroup analyses based on duration of shoulder pain for participants receiving ‘usual care’ (score ranging from 0 to 100).

Duration of symptoms	Baseline Mean ± SD	Changes to 6 W (95% CI)	Changes to 3 M (95% CI)	Changes to 6 M (95% CI)	Changes to 12 M (95% CI)
Acute shoulder pain
Pain	Participants (n)	–	–	260 [14, 41]	–
	Pain score 46.5 ± 36	–	–	29.7 (25.3 to 34.7)	–
	Heterogeneity	–	–	I^2 = 0%	–
Persistent shoulder pain
Pain	Participants (n)	339 [25, 60–63]	363 [25, 26, 61, 63]	532 [6, 25, 61–64]	–
	Pain score 45.2 ± 40	12.3 (5.4 to 19.1)	14.8 (8.9 to 20.7)	22.6 (15.1 to 30.2)	–
	Heterogeneity	I^2 = 81%	I^2 = 75%	I^2 = 86%	–
Function	Number of participants	210 [60, 61, 63]	348 [26, 29, 61, 63, 71]	247 [29, 61, 63]	205 [29, 61]
	Function score 46.6 ± 44	13.4 (0.6 to 26.2)	15.7 (10.1 to 21.3)	16.2 (10.6 to 21.8)	27.9 (23.9 to 31.8)
	Heterogeneity	I^2 = 93%	I^2 = 70%	I^2 = 64%	I^2 = 0%

N = number of participants, CI = confidence intervals.

Meta-regression

A meta-regression was performed for ten studies reporting pain scores for patients who received ‘usual care’ [6, 14, 25, 26, 41, 60–64]. We used the model to estimate pooled means and 95% CIs of expected pain outcomes at 2, 3, 4, 6, 12, 16, 20, 26 and, 52 weeks. The pooled estimate for average pain scores was 47.2/100 points (95%CI; 41.2–53.2) at baseline, reducing to 28.2 (95%CI; 22.1–34.4) at 6 weeks, 28.5 (95%CI; 22.3–34.8) at 3 months, 23.8 (95%CI; 17.7–29.8) at 6 months and 19.8 (95%CI; 13.4–26.6) at 12 months. Figure 16 shows the association between pain scores and time.

Figure 16. Meta regression showing average pain scores over time. A linear regression of average pain scores over one year in participants receiving usual care. The average pain was 47.2/100 at baseline, reduced nearly 20 scores at 6 weeks, remained constant to 3 months before dropping gradually down to about 20/100 at 12 months.

Discussion

Our findings suggested that participants with persistent shoulder pain who received ‘no intervention’ did not experience significant changes in pain and function scores over time (up to 6 months following study entry). Participants with acute or persistent shoulder pain who received ‘usual care’ presented 30/100 points improvement in pain and function scores over 12 months and approximately 40% of that improvement in pain and function occurred during the first six weeks. Changes between pain and function scores over time were similar for those participants who received either ‘no intervention’ or ‘usual care’. The subgroup analysis showed pain improved 30/100 points for patients with acute shoulder pain and 22/100 for patients with persistent shoulder pain at 6 months.

Meta-analysis of studies reporting ‘no intervention’ and ‘usual care

The meta-analysis for studies reporting data of participants with persistent shoulder pain who received ‘no intervention’ presented, on average, no change in pain and function from baseline to three months follow-up. Findings from one study were not included in that meta-analysis as that study reported outcomes only for 6 months follow-up [52]. In that study, participants also reported no changes in pain and function scores to 6 months. All studies reporting data from participants receiving ‘no intervention’ recruited participants with persistent shoulder pain. It is possible that pain and function do not improve at 3 months or even up to 6 months [52] if participants with persistent subacromial shoulder pain receive ‘no intervention’.

Participants who received ‘usual care’ improved 30/100 points in pain and function scores over 12 months. Pain and function scores both improved approximately 13/100 points during the first 6 weeks after entering the study (around 40% reduction). The clinical course for pain and function almost plateaued by 3 months (16/100 points) and improved gradually to 24/100 at 6 months and 30/100 at 12 months. Similarly, the findings of the meta-regression indicated an overall change of 28/100 in pain score at 12 months with about 40% of these changes occurring in the first 6 weeks following ‘usual care’ intervention. The meta-regression analyses presented high heterogeneity, suggesting large uncertainties on these estimates.

The pooled estimates of changes in pain and function scores for studies offering ‘usual care’ showed high heterogeneities. Possible causes for high heterogeneity include wide individual range of pain and function scores (evident with relatively large SDs at baseline, Table 4), shoulder pain duration, the low number of sample size in some studies which may diversify the course of improvement, and a wide range of interventions covered by the term ‘usual care’. Individual person’s response to interventions likely add to the heterogeneity, particularly as the presence of lesions such as rotator cuff tears do not predict symptoms and level of disability in most people with shoulder pain [72, 73]. Additionally the difference in the number of participants in studies could contribute towards heterogeneity; it is possible that the low number of participants lead to higher variability due the underestimating or overestimating of research outcomes [74].

Heterogeneity varied according to the subgroup being analyzed. We found low heterogeneity (I²=0%) when analyzing participants with acute pain [14, 41]. The low heterogeneity could be attributed to similar duration of symptoms and follow-up time point (i.e. 6 months) [14, 41]. On the other hand, heterogeneity was high for analysis of participants with persistent pain (I² =86%). This may have been caused by a wide range of pain intensity, pain duration, different follow-up time points and interventions received by participants in those studies. For example, two studies included participants with acute and persistent pain [25, 61]. Patients with acute shoulder pain may have a more predictable recovery rate for pain and function compared to those with persistent pain as the latter is likely influenced by multiple contributing factors, potentially adding to the high heterogeneity for that group [14, 75].

The selection of interventions offered to participants with persistent pain may also have impacted the heterogeneity. In the study by Laslett, Steele [61], for instance, participants were treated with corticosteroid injection up to 3 first weeks. In comparison, in the study by Geraets, Goossens [26] participants were treated with analgesic medication for the first two weeks. It has been shown that, in the short term, corticosteroid injection can significantly reduce pain compared to analgesic medication or physiotherapy intervention; however, this effect is not evident in the longer term [74, 76–78]. Therefore, the differences in the type of intervention would contribute to participants pain report during follow-ups particularly in short term.

The risk of bias within randomized controlled trials varied. Five randomized controlled were rated low risk of bias in all domains [25, 26, 55, 57, 64]. Randomization or allocation concealment was unclear in seven studies [41, 52, 54, 56, 58, 60, 63]. The lack of blinding in these domains can contribute to bias in unpredicted distribution of participants in groups and bias in effect estimate [79]. Assessor blinding was considered to be unclear or high risk in four studies [29, 54, 56, 71]. Given outcome measures were based on patients’ reported questionnaire, it is unlikely that the non-blinding of assessors affected participants’ reporting [80]. Three studies presented high risk of attrition bias [29, 52, 63]. Those studies analyzed their data based on intention-to-treat principles and used multiple imputation to minimize the risk of reporting for missing data. The imputation makes unconfirmed assumptions that may bias the treatment effect [81]. One study scored high risk of bias in selective reporting [58]. The high risk of bias in that domain did not affect the results of this systematic review as the data missing was irrelevant to the outcome of interest for the present review [58]. Two studies were rated as high risk in other bias domain due to limited information presented [53, 54]. All four observational studies were rated with low risk of bias in all domains [6, 14, 61, 62].

Subgroup analysis

Our systematic review showed similar pain scores at baseline between acute and persistent shoulder pain. These findings are confirmed by findings from other studies showing that pain intensity does not correlate with the duration of the symptom [40, 82]. We found moderate clinical differences in pain in both acute and persistent subacromial shoulder pain at 6 months. There is strong evidence suggesting that patients with longer pain duration present poorer clinical outcomes [17, 19]. We showed participants with acute pain presented improvements of 29.7/100 (95% CI from 25.3 to 34.7) scores over 6 months, and participants with persistent shoulder pain presented improvements of 22.6/100 (95% CI 15.1 to 30.2) over 6 months. However, because of the limitation in the meta-analysis, it was not possible to assess the statistical differences between acute and persistent shoulder pain.

Patterns of improvement between pain and function

Our findings suggest a similar pattern of changes between pain and function over time. Similar findings were reported in a previous systematic review in patients with low back pain [40]. Pain and function scores were found to have a moderate to strong correlation [21, 83], which may explain our findings for patients with subacromial pain syndrome. This finding highlights the effect of pain on functional presentation [84] and illustrates how patients often rate their pain as function, or vice versa [21].

Implications for clinical practice

The findings of this study can be interpreted using minimal clinically important differences (MCID) for pain and function scores as a reference. For the NPRS, a change of 20 points (0–100) represents minimal improvement [85–87], 20 to 30 represents ‘meaningful/moderate improvement, and ≥35 to 40 points represents ‘large/substantial’ clinical improvement [88, 89]. As a group, participants receiving ‘usual care’ may not have met the MCIDs at three months but due to gradual changes, they reported moderate clinical change at six months (mean: 24.3 scores) and at 12 months (mean: 32.5 scores) from baseline (Figures 9 and 10). During the subgroup analyses, both participants with acute or persistent shoulder pain presented moderate clinical changes at 6 months (Table 4).

Regarding the clinical improvement in function for patients who received ‘usual care’, the median MCID for function scores (i.e. SPADI, DASH, and Shoulder Function Questionnaires) was 15 [86]. To our knowledge, there is no information of minimum changes required for patients to achieve moderate or large improvements in function. Based on the MCID’s, patients with persistent subacromial shoulder pain receiving ‘usual care’ reported minimal improvement in function (mean difference = 15.7/100 scores) at three months (Table 4). We did not pool function scores in any time points for patients with acute shoulder pain, given the limited data available.

We observed gradual changes in pain and function over one year for participants who received ‘usual care’. The gradual changes are likely to be influenced by the individual patients’ expectations, baseline characteristics, treatment effect as well as placebo effects [28]. A recent systematic review concluded that it is not possible to separate the real treatment effect from placebo effects as the treatment effect is integrated with positive contextual factors such as expectation of the treatment, belief, previous experience, and, for those receiving manual therapy, therapeutic touch [90]. The pooled data reported in this review for changes of pain and function scores in the ‘usual care’ group can be used as a reference for interventional studies to determine whether an intervention is superior to current usual care or not.

Limitations

This review has a few limitations. 1) The first limitation is the wide diversity of outcome measures used for pain and function scores. A previous systematic review highlighted the importance of developing a core domain of outcome measurement sets for trials recruiting patients with shoulder pain, a need that was also highlighted by the results of this study [91]. 2) Another limitation includes the fact that ‘usual care’ interventions included a wide range of interventions (e.g. ‘multimodal intervention’, ‘advise and exercise’, and ‘conventional physiotherapy intervention’), which resulted in high clinical heterogeneity between studies. 3) We could not undertake subgroup analyses for different types of interventions due to limited number of studies in each treatment category. 4) A final limitation is that there is no widely accepted tool for assessing the quality of observational studies [92]. In this review, we used an adapted version of the tool designed by Altman [39]. 5) We did not document the agreement between two reviewers for study selection and risk of bias.

Conclusions

Our findings suggest that patients with persistent subacromial shoulder pain who receive ‘no intervention’ did not improve their pain or function scores when followed up at three and/or six months. The findings of the meta-analysis and meta-regression for participants receiving ‘usual care’, both with high heterogeneity, may indicate moderate clinical improvement at 12 months following the initial treatment period for those with acute and persistent subacromial shoulder pain. The review included a small number of observational studies in addition to fifteen randomized controlled trials. Some of the randomized controlled trials had a low number of participants which can lower the study power increasing risk of underestimating or overestimating the outcomes. In light of the limitations, we recommend interpreting these results with caution. The clinical courses for pain and function scores were similar when receiving either ‘no intervention’ or ‘usual care’, suggesting that the pain intensity may influence the level of function. Subgroup analysis based on the duration of symptoms in participants receiving ‘usual care’ suggests moderate clinical improvement in pain at six months, regardless of whether they presented with acute or persistent shoulder pain. The findings of this systematic review and meta-analyses add to the growing literature information on the expected recovery rate at different time points. Future studies may refer to these expected changes as a reference for interpreting their findings.

Disclosure statement

No potential conflict of interest was reported by the authors.

Additional information

Funding

The author(s) reported there is no funding associated with the work featured in this article.

Appendix:

Clinical course of pain and function in patients with subacromial shoulder pain: a systematic review with meta-analysis & meta-regression

Keywords used for data search

Table

Population	Intervention	Prognosis
Shoulder	No intervention	Prognosis
Musculoskeletal disorders	Waiting list No exercise	Clinical course
Shoulder pain	Wait and see	Natural course
Shoulder impingement syndrome	Usual care	Course of pain
Shoulder problems	Standard care	Course of shoulder
Rotator cuff	Ergonomics	Inception
Subacromial pain syndrome	Usual therapeutic approach	Follow-up
Subacromial impingement		Recovery
Painful arc
Shoulder complaints
Subacromial injection
Shoulder disorders
Non-specific shoulder

Search strategy used in each database

Terms that used for population:

Shoulder, musculoskeletal disorders, shoulder pain, shoulder impingement syndrome, shoulder problems, rotator cuff, subacromial pain syndrome, subacromial impingement, painful arc, shoulder complaints, shoulder disorders, non-specific shoulder

Terms that used for the intervention:

no intervention, no treatment, standard care, usual care, waiting list, wait and see, wait-and-see, no exercise, ergonomics, usual therapeutic approach

Terms that used for clinical course:

Prognosis, clinical course, natural course, course of pain, course of shoulder, inception, follow-up, recovery

Search strategy in Web of Science:

TS=(‘rotator cuff’ OR ‘painful arc’ OR ‘shoulder pain’ OR ‘shoulder impingement syndrome’ OR ‘shoulder problems’ OR ‘subacromial pain syndrome’ OR ‘subacromial impingement’ OR ‘shoulder complaints’ OR ‘shoulder disorders’ OR ‘non-specific shoulder’ OR (shoulder AND ‘musculoskeletal disorders’)

AND

TS=(‘no intervention’ OR ‘no treatment’ OR ‘standard care’ OR ‘usual care’ OR ‘wait and see’ OR ‘waiting list’ OR ‘no exercise’ OR ergonomic OR ‘usual therapeutic approach’)

AND

TS=(‘follow-up’ OR recovery OR inception OR prognosis OR ‘clinical course’ OR ‘natural course’ OR ‘course of pain’ OR ‘course of shoulder’)

NOT

TS=(animal OR rat OR rabbit OR cancer OR rheumatology OR dislocation OR fracture OR arthroplasty OR hemiplegic OR frozen OR stroke OR surgery)

Search strategy in Scopus:

#1 ‘shoulder pain’ OR ‘shoulder impingement syndrome’ OR ‘shoulder problems’ OR ‘rotator cuff’ OR ‘subacromial pain syndrome’ OR ‘subacromial impingement’ OR ‘painful arc’ OR ‘shoulder complaints’ OR ‘subacromial injection’ OR ‘shoulder disorders’ OR ‘non-specific shoulder’ OR ‘nonspecific shoulder’ OR (shoulder AND ‘musculoskeletal disorders’)

#2 ‘no intervention’ OR ‘no treatment’ OR ‘standard care’ OR ‘usual care’ OR ‘wait and see’ OR ‘waiting list’ OR ‘no exercise’ OR ergonomic OR ‘usual therapeutic approach’

#3 = ‘follow-up’ OR recovery OR prognosis OR ‘clinical course’ OR ‘natural course’ OR ‘course of pain’ OR ‘course of shoulder’ OR inception

#4 animal OR rat OR rabbit OR cancer OR rheumatology OR dislocation OR fracture OR arthroplasty OR hemiplegic OR frozen OR stroke OR surgery

#5: #1 AND #2 AND #3

#6: #5 AND NOT #4

Search strategy in Cochrane Library, Medline, Embase, and AMED (via OVID)

#1 ‘shoulder pain’ OR ‘shoulder impingement syndrome’ OR ‘shoulder problems’ OR ‘rotator cuff’ OR ‘subacromial pain syndrome’ OR ‘subacromial impingement’ OR ‘painful arc’ OR ‘shoulder complaints’ OR ‘subacromial injection’ OR ‘shoulder disorders’ OR ‘non-specific shoulder’ OR ‘nonspecific shoulder’ OR (shoulder AND ‘musculoskeletal disorders’)

#2 ‘no intervention’ OR ‘no treatment’ OR ‘usual care’ OR ‘standard care’ OR ‘wait and see’ OR ‘wait-and-see’ OR ‘waiting list’ OR ‘no exercise’ OR ergonomic OR ‘Usual therapeutic approach’

#3 prognosis OR ‘clinical course’ OR ‘natural course’ OR inception OR ‘course of pain’ OR ‘course of shoulder’ OR Follow-up OR recovery

#4 animal OR rat OR rabbit OR cancer OR rheumatology OR dislocation OR fracture OR arthroplasty OR hemiplegic OR frozen OR stroke OR surgery

#5: 1 AND 2 AND 3

#6: 5 NOT 4