The influence of exercise therapy dosing on pain and functional outcomes in patients with subacromial pain syndrome: A systematic review

Abstract

Background

The objective was to identify exercise therapy dosing parameters for subacromial pain syndrome (SAPS) associated with improved pain and function outcomes (via effect sizes) and determine the extent of exercise intervention reproducibility.

Methods

An electronic search of PubMed/MEDLINE, Cumulative Index to Nursing and Allied Health Literature, EMBASE, Cochrane Database of Systematic Reviews, and SportDiscus identified studies that used exercise therapy exclusively to treat SAPS. Exercise therapy dosing parameters were extracted and within-group effect sizes were calculated for all pain and functional outcomes. Template for Intervention Description and Replication and Consensus on Exercise Reporting were used to record intervention reporting. The risk of bias and Grading of Recommendations, Assessment, Development, and Evaluation were assessed by two reviewers.

Results

Twenty-one trials with 674 subjects were included. Effect sizes for pain and function were large in 18 studies, medium in six studies, and small or no effect in four studies, despite the type of supervision. With moderate certainty, effect sizes of pain and function were not influenced by exercise therapy dosing parameters. Intervention reporting was generally poor.

Discussion

Exercise therapy for SAPS was associated with large effect sizes for improvement in pain and function but optimal exercise therapy dosing parameters could not be identified. Strong recommendations conditionally suggest treating SAPS with a variety of exercise therapy dosing parameters.

Introduction

Subacromial pain syndrome (SAPS) includes rotator cuff disease and associated dysfunction and also carries the label of shoulder impingement.^1,2 SAPS accounts for 44% to 65% of all care-seeking visits for shoulder pain,^3,4 and includes non-traumatic, unilateral shoulder problems with localized acromion pain that increases with overhead activity. ¹ Exercise therapy is recommended for the management of SAPS in many clinical practice guidelines and systematic reviews,^1,5–7 and recent evidence suggests that exercise therapy may result in outcomes comparable to surgery.^8,9

The optimal dose of exercise therapy for the management of shoulder disorders has not yet been determined and varies greatly between practitioners.^6,7,10–12 Exercise dosing is most commonly defined as the total repetitions and sets of a specified number of exercises, and the frequency, intensity, and duration of the exercises. ¹³ Recent systematic reviews attempting to describe these treatment parameters for lower extremity musculoskeletal disorders have also included the total number of sessions, frequency of these sessions, and the amount of time spent in each session.^14,15 Establishing optimal exercise dosing guidelines could help clinicians maximize clinical outcomes, assist in standardizing care, and aid researchers when designing study protocols.

The efficacy of exercise therapy for SAPS has been investigated,^6,7 but conclusions about specific exercise therapy parameters have not been determined due to poor exercise intervention reporting. ¹¹ Defining exercise therapy dosing parameters was considered a “subsidiary” aim in the aforementioned review, ¹¹ and no formal intervention reporting guidelines were available for the reviewers to reference at that time. Also, a lack of transparency in intervention methodology and reporting has made reproduction of interventions in clinical settings challenging.^16,17 The Template for Intervention Description and Replication (TIDieR) checklist was developed to improve intervention reporting, improve methodological transparency, and may help with the replication of interventions used in trials.^18,19 The Consensus on Exercise Reporting (CERT) was subsequently developed to address suboptimal guidance from the TIDieR checklist to replicate exercise interventions, adding much greater specificity. ²⁰ While there is overlap in seven categories, the CERT is meant to be an extension of TIDieR specifically for exercise interventions.

Because these checklists are relatively new, it is unclear how much impact they have had on intervention reporting in trials. Strong recommendations have been made for the use of exercise therapy to address pain and function in patients with SAPS, but an emphasis on dosing parameters or thoroughness of reporting is lacking. ⁷ Poor reporting of important details can substantially limit reproducibility. Therefore, the primary objective of this review was to describe effect sizes for exercise therapy dosing parameters and to highlight the parameters associated with the largest effect sizes for pain and/or function in trials for SAPS. This knowledge can help improve guidance for real-world implementation of trial findings. ⁷ The secondary objective was to assess the reproducibility of exercise interventions using the TIDieR and the CERT checklists.

Methods

The Preferred Reporting Items for Systematic Reviews and Meta-Analysis guidelines were followed in the reporting of this systematic review, and the review was registered (PROSPERO: CRD42020185598).

Data sources and eligibility criteria

An electronic search of the PubMed/MEDLINE, Cumulative Index to Nursing and Allied Health Literature (CINAHL), EMBASE, Cochrane Database of Systematic Reviews, and SportDiscus databases was performed to identify clinical trials which used exercise therapy as a primary intervention for SAPS. Systematic reviews were also manually searched for relevant clinical trials. Studies were eligible for inclusion if the study: (1) was a randomized controlled trial; (2) included human subjects 18 to 65 years old with shoulder pain defined as SAPS or rotator cuff disease based on clinical findings; (3) used exercise therapy as a primary intervention without adjunct interventions, such as a modality or manual therapy, in at least one treatment arm; (4) included an outcome measure for pain and/or function; (5) was published between January 2000 and August 2021. Studies were excluded if the: (1) investigation included participants with full-thickness rotator cuff tears or (2) study was not published or translated into the English language. Full-thickness tears were excluded because these individuals may respond differently to exercise therapy, and previous definitions of SAPS have only included partial thickness tears.^1,21

Search strategy

The search strategies used database-dependent controlled descriptors such as medical subject headings (MeSH terms), CINAHL headings, and recommended synonyms or keywords. The controlled descriptors and keywords were used in combination with the Boolean operators “AND” and “OR.” The appendix provides an example search strategy.

Two reviewers (RM and MK) independently screened the titles and abstracts for potential full texts to review within Covidence, a web-based data management tool for systematic reviews. Eligible full-text studies were retrieved and screened by the same reviewers for inclusion. Reference lists in systematic reviews were assessed manually for relevant studies not identified during the database searches. Disagreements between reviewers were resolved by consensus, and a third author (JY) was available if consensus could not be met.

Data extraction and quality assessment

Data were extracted by two reviewers (RM and MK) and included exercise type and supervision provided (unsupervised, supervised, or a combination), single session duration, frequency of intervention, total number of sessions, duration of care, and follow-up time frame. Operational definitions of these exercise therapy dosing parameters have been described in other studies (Table 1).^14,15 If data were missing, corresponding authors were contacted. The names of outcome measures assessing pain and/or function and the means and standard deviations of the measures were also captured.

Table 1.

Definition of dosing variables.

Dosing variable	Operational definition
Exercise type	Activity performed by a patient that was prescribed by a healthcare provider and required physical effort with the intention of improving overall health and fitness. Can include aerobic activity, strengthening, stretching, or proprioceptive exercises
Single session duration	Amount of time spent in one single exercise session, either supervised by a healthcare provider or as a HEP
Frequency	How often does the individual perform supervised exercise intervention or HEP
Total number of sessions and supervision provided	Number of exercise sessions supervised or unsupervised that were performed over the duration of the study
Duration of care	Number of weeks or months an individual performed supervised exercise or a HEP
Follow-up time frame	Length of time, in weeks, months, or years between the initial exercise intervention and the final follow-up time frame

Two reviewers (RM and MK) independently scored the risk of bias in trial design, conduct, and reporting using the Cochrane Collaboration’s revised Risk of Bias (RoB-2) tool. Each study’s risk of bias was rated as “low risk,” “some concerns,” or “high risk,” and an overall consensus was reached through discussion. The RoB-2 assesses the risk of bias in studies across five domains: bias arising from the randomization process, bias due to deviation from the intended interventions, missing outcome data, bias in the measurement of the outcome, bias in the selection of reported result, as well as the overall risk of bias. Cohen’s kappa was used to measure reviewer agreement when evaluating the risk of bias. ²²

The TIDieR checklist and the CERT^18,20 were used to record components of the exercise interventions reported in each trial by two independent reviewers (RM and MK). A third reviewer (JY) was available to resolve disagreements. The TIDieR checklist describes 12 domains that are recorded as included or not included. The CERT includes six sections with 16 items and records information as included or not included.

Data synthesis and analysis

The within-group means and standard deviations of the reported outcomes at baseline and the end of the trial were used to calculate within-group effect sizes. Within-group effect sizes have been used in other exercise training studies,^23,24 and are appropriate for use with this type of comparison to determine if exercise has an effect greater than zero and if specific exercise dosing parameters influence the magnitude of the effect.^25,26 Cohen’s benchmarks (<0.2 no effect; 0.2 small; 0.5 medium; 0.8 large) were used to classify the magnitude of the effect. ²⁷ Either control or experimental groups were included in the analysis if the intervention included exercise therapy exclusively.

Two reviewers (RM and MK) judged the certainty of evidence using the Grading of Recommendations Assessment, Development, and Evaluation (GRADE) approach. ²⁸ Two questions relating to the primary aim were considered for making overall narrative synthesis recommendations: (1) do different exercise therapy dosing parameters influence the effect size of pain outcomes? and (2) do different exercise therapy dosing parameters influence the effect size of functional outcomes? Evidence was assessed according to the GRADE handbook where

• “High certainty” was defined as being very confident that the true effect lies close to that of the estimate of the effect.

• “Moderate certainty” was defined as being moderately confident in the effect estimate where the true effect is likely to be close to the estimate of the effect, but there is a possibility that it is substantially different.

• “Low certainty” is defined as limited confidence in the effect estimate. (The true effect may be substantially different from the estimate of the effect.)

• Very low confidence in the effect suggests very little confidence in the effect estimate. (The true effect is likely to be substantially different from the estimate of effect.) ²⁹

Because the focus of this review was to summarize exercise dosing parameters that showed a significant treatment effect, rather than assess the treatment effect for exercise therapy, a meta-analysis was not relevant or appropriate. The goal was to better understand which specific dosing parameters should be used in clinical settings if similar treatment effects are expected.

Results

Database searches identified 6481 studies before the removal of duplicates. After title/abstract and full-text screening, 21 studies with 674 subjects were included and effect sizes were calculated (Figure 1). Agreement prior to consensus was substantial (k = 0.67). Both the control and intervention groups were included in seven studies.^30–36 The exercise therapy dosing variables and calculated effect sizes for SAPS can be found in Table 2. None of the individual dosing parameters were consistently associated with a specific effect size.

Figure 1.

Flow diagram of search results.

Table 2.

Exercise therapy dosing variables and calculated effect sizes.

Study	Exercise type	Subjects: (n); sex (M:F)	Mean age (years) (SD)	Single session duration	Frequency	Total number of sessions and supervision provided	Duration of care/ follow-up time frame	Effect size (calculated for a time frame of study)	Magnitude of effect
Ager et al. 37	Group supervised neuromuscular training (multi-station)	n = 16; 16:0	33 (9.5)	35–45 min	3×/week	18 supervised	6 weeks/12 weeks	DASH (−1.14) WORC (1.39)	Large Large
Arias-Buria et al. 38	Eccentric exercises for rotator cuff and scapular stabilizing musculature	n = 19; 5:14	57 (6)	25–30 min	1×/week supervised; 2×/day unsupervised	4 supervised 56 unsupervised	4 weeks/5 weeks	DASH (−4.41) NPRS Current (−2.36) NPRS worst (−1.83) NPRS best (−3.35)	Large Large Large Large
Arias-Buria et al. 39	Concentric/eccentric exercise for rotator cuff and scapular stabilizing musculature	n = 25; 19:6	48 (6)	20–25 min	1×/week supervised; 2×/day unsupervised	5 supervised 70 unsupervised	5 weeks/12 months	DASH (−4.79) NPRS Current (−3.33) NPRS worst (−4.70)	Large Large Large
Bang and Deyle 40	Standardized flexibility and strengthening program	n = 24; 12:12	45 (8.4)	Supervised 30 min; unsupervised 5 min	2×/week supervised; 1×/day unsupervised	6 supervised 21 unsupervised	3 weeks/8 weeks	Pain composite score (−0.77) Functional assessment questionnaire (0.70)	Medium Medium
Boudreau et al.a 31	RCEx: Rotator cuff exercises RCEx + coactivation: Rotator cuff exercises plus coactivation of glenohumeral adductors	RCEx: n = 21; 8:13 RCEx+: n = 21; 12:9	RCEx: 49.6 (13.2) RCEx+: 50.2 (10.9)	RCEx: Not reported RCEx + coactivation: Not reported	RCEx: 1×/day unsupervised RCEx + coactivation: 1×/day unsupervised	RCEx: 2 supervised 42 unsupervised RCEx + coactivation: 2 supervised 42 unsupervised	6 weeks/6 weeks	DASH for RCEx (−0.08) WORC for RCEx (0.51) VAS at rest for RCEx (0.43) DASH for RCEx+ (−0.24) WORC for RCEx+ (0.65) VAS at rest for RCEx+ (−0.02)	No effect Medium Small Small Medium No effect
Camargo et al. 41	Stretching and strengthening exercises	n = 23; 14:9	32.65 (10.73)	Not reported	Not reported	Unclear	4 weeks/4 weeks	DASH-Brazilian (−0.91) VAS at rest (−0.62) VAS during shoulder movement (−1.47) VAS greatest pain in last week (−1.30) VAS least pain in the week (−0.08)	Large Medium Large Large No effect
Dejaco et al.a 33	Concentric: Rotator cuff strengthening exercises, stretching exercises Eccentric: Eccentric rotator cuff exercises, stretching exercises	Concentric: n = 16; 9:7 Eccentric: n = 20; 10:10	Concentric: 48.6 (12.3) Eccentric: 50.2 (10.8)	Concentric: Not reported Eccentric: Not reported	Concentric: 1×/week supervised for 6 weeks then 3× over the next 6 weeks; 1×/day unsupervised Eccentric: 1×/week supervised for 6 weeks then 3× over next 6 weeks; 2×/day unsupervised	Concentric exercise: 9 supervised 84 unsupervised Eccentric exercise: 9 supervised 168 unsupervised	12 weeks/26 weeks	Constant Murley score concentric (1.19) VAS Concentric (−0.96) Constant Murley score eccentric (0.83) VAS Eccentric (−0.92)	Large Large Large Large
de Oliveira et al. 42	Sensorimotor training, strengthening exercises	n = 26; 15:9	29.4 (7.5)	30–45 min	2×/week for 4 weeks, then 1×/week for 2 weeks supervised; 3×/day unsupervised	10 supervised 126 unsupervised	6 weeks/24 weeks	DASH (−2.24) Brief Pain Inventory (−1.70) WORC (2.18)	Large Large Large
Dupuis et al. 43	Isometric exercises then graduated strengthening after 2 weeks	n = 20; 13:7	33 (7)	15 min	3×/day × 2 weeks, then 1×/day for 4 weeks, unsupervised	70 unsupervised	6 weeks/ 6 weeks	DASH (−1.05) WORC (1.37) Brief Pain Inventory (−0.84)	Large Large Large
Farfaras et al. 44	Phased shoulder exercise for strengthening	n = 21; 13:8	49.9 (9.3)	60 min	2×/week supervised; unsupervised “rest of the days”	Unclear	3–6 months/2–3 years	SF-36 (PF) (−0.10) SF-36 (RP) (0.81) SF-36 (BP) (0.83) SF-36 (GH)(−0.06) SF-36 (VT) (0.28) SF-36 (SF) (0.24) SF-36 (RE) (0.47) SF-36 (MH) (0.18) Constant Murley (0.27)	No effect Large Large No effect Small Small Small No effect Small
Giombini et al. 45	Prone pendulum swings and passive stretches	n = 11; 9:2	26.3 (6.2)	5 min	1×/week supervised; 2×/day unsupervised	4 supervised 56 unsupervised	4 weeks/ 10 weeks	VAS at night/ with movement/ at rest (−1.36) Constant Murley (0.88)	Large Large
Granviken and Vasseljena 32	Unsupervised and Supervised groups: Neuromuscular re-education, strengthening, stretching	Unsupervised: n = 23; 12:11 Supervised: n = 23; 12:11	Unsupervised: 48.2(9.8) Supervised: 47.6(10)	Unsupervised Not reported Supervised: Not reported	Unsupervised: 1 × supervised, then 2×/day unsupervised Supervised: 10 × supervised and 2×/day unsupervised	Unsupervised: 1 supervised 84 unsupervisedb Supervised: 10 supervised 84 unsupervisedb	6 weeks/26 weeks	SPADI-unsupervised (−1.32) NPRS-average in the last week-unsupervised (−1.11) SPADI-supervised (−1.46) NPRS-average in the last week supervised (−0.84)	Large Large Large Large
Gutiérrez-Espinoza et al.a 35	Control: Conscious control for proprioception; scapular and glenohumeral control exercises Intervention: As above, plus “unilateral corner stretch” (pectoralis minor) Both groups also: AROM and stretching cervical spine and shoulder	Control: n = 40; 25:15 Intervention: n = 40; 27:13	Control: 44.5 (5.4) Intervention: 45.2 (4.3)	Control: Not reported Intervention: Not reported	Control: 3×/week supervised; 2×/day unsupervised Intervention: 3×/week supervised; 2×/day unsupervised	Control: 36 supervised 168 unsupervised Intervention: 36 supervised 168 unsupervised	12 weeks/12 weeks	Constant Murley control (3.32) DASH control (−3.41) VAS at rest control (−0.16) VAS during movement-control (−2.74) Constant Murley intervention (3.46) DASH intervention (−2.69) VAS at rest intervention (0.16) VAS during movement intervention (−3.02)	Large Large No effect Large Large Large No effect Large
Letafatkar et al. 46	Stretching and strengthening exercises	n = 40; 18:22	40.5 (5.5)	60 min	3×/week supervised	24 supervised	8 weeks/8 weeks	NRPS (−1.05) DASH-Iranian (−0.88)	Large Large
Ludewig and Borstad 47	Resisted strengthening (serratus anterior, glenohumeral ER), stretching (pectoralis minor and posterior shoulder), relaxation (upper trapezius)	n = 34; 34:0	48(1.8)	Not reported	1x/day stretching; 5x/day relaxation; 3x/week strengthening, all unsupervisedc	56 unsupervised	8 weeks/ 8–12 weeks	Shoulder rating questionnaire (5.65) Work-related disability component of SPADI (−5.42) Work-related pain (−7.02)	Large Large Large
Moslehi et al. 48	Scapular-focused exercise (isometric stretching, intrinsic, and eccentric isotonic exercises)	n = 25; 5:20	37.5 (8.1)	Not reported	Not reported	22 unclear	8 weeks/8 weeks	VAS (−1.76) DASH (−0.99)	Large Large
Østerås et al.a 36	Control: Low dose exercise therapy, progressive resistive shoulder strengthening, and aerobic exercise Intervention: High dosage medical exercise therapy, progressive resisted shoulder strengthening, and aerobic exercise	Control: n = 30; 17:13 Intervention: n = 31; 19:12	Control: 41.8 (14.5) Intervention: 46.1 (11.2)	Control: 20–30 min Intervention: 70–90 min	Control: 3×/week Intervention: 3×/week	Control: 36 supervised Intervention: 36 supervised	12 weeks/12 months	VAS control (−0.95) SRQ control (0.67) VAS Intervention (−3.16) SRQ intervention (3.68)	Large Medium Large Large
Senbursa et al. 49	AROM, flexibility, and strengthening exercises (rotator cuff and scapular musculature)	n = 15; NR	49.5 (7.9)	10–15 min	1×/day unsupervised	28 unsupervised	4 weeks/12 weeks	VAS night pain (−2.79) VAS pain with motion (−1.74) VAS pain at rest (−0.77)	Large Large Large
Shah et al.a 30	Control: Strengthening and stretching Intervention: Strengthening, stretching, and scapular stability	Control: n = 30; 12:18 Intervention: n = 30; 19:11	Control: 20–60d Intervention: 20–60d	Control: Not reported Intervention Not reported	Control: 6×/week Intervention: 6×/week	24 unclear	4 weeks/4 weeks	VAS control (−4.72) SPADI control (−3.93) VAS intervention (−6.90) SPADI intervention (−4.36)	Large Large Large Large
Turgut et al.a 34	Control: Shoulder girdle stretching and strengthening exercises Intervention: Shoulder girdle stretching and strengthening with additional scapular stabilization exercises	Control: n = 15; 8:7 Intervention: n = 15; 8:7	Control: 39.5 (8.2) Intervention: 33.4 (9.3)	Control: Not reported Intervention Not reported	Control: Not reported Intervention: Not reported	Control: Unclear Intervention: Unclear	12 weeks/12 weeks	SPADI pain-control (−1.24) SPADI disability-control (−1.03) SPADI total control (−1.16) VAS pain at rest control (−0.46) VAS pain during activity control (−1.41) VAS pain at night control (−0.65) SPADI pain-intervention (−2.23) SPADI disability-intervention (−1.68) SPADI total-intervention (−2.02) VAS pain at rest-intervention (−0.71) VAS pain during activity-intervention (−2.51) VAS pain at night-intervention (−0.99)	Large Large Large Medium Large Medium Large Large Large Large Large Large
Vinuesa-Montoya et al. 50	Stretching and strengthening of the shoulder girdle	n = 20; 13:7	38–58d	30 min	2×/day unsupervised	70 unsupervised	5 weeks/5 weeks	VAS (−0.52) DASH (−.50) SDQ (−0.62)	Medium Medium Medium

AROM: active range of motion; BP: body pain; DASH: Disability of Arm, Shoulder, and Hand; ER: external rotation; GH: general health; MH: mental health; NPRS: numeric pain rating scale; PADI: shoulder pain and disability index; PF: physical functioning; RE: role-functioning emotional; RP: role-functioning physical; SDQ: strengths and difficulties questionnaire; SF: social functioning; SF-36: 36-Item Short Form Survey; SRQ: self-reporting questionnaire; VAS: Visual Analogue Scale; VT: vitality; WORC: Western Ontario Rotator Cuff Index.

^a Intervention and control groups included.

^b Medians reported for this study. In order to be comparable to other studies, the calculated total number of sessions were used in analysis.

^c Omitted 5×/day relaxation of upper trapezius muscle from total number of sessions.

^d Only range provided.

Methodological quality assessment

Six studies were rated as “high risk,” four studies “some concerns,” and 11 studies “low risk” for risk of bias. The initial agreement was fair (k = 0.33). The consensus bias judgments for each domain are provided in Supplementary Data and the overall judgment is represented in Figure 2. The initial agreement was substantial (k = 0.79) for consensus with the TIDieR checklist and the CERT (k = 0.66). See Supplementary Data for TIDieR and CERT tables for details. Using the GRADE approach, the level of certainty was downgraded by one level for indirectness due to the heterogeneity of the populations included in the various trials, resulting in moderate certainty in the overall results. See Supplemental Data for explanatory tables.

Figure 2.

Cochrane Risk of Bias-2 Reporting.

Effect sizes

The effect sizes were calculated for 28 exercise therapy groups from 21 trials for a variety of outcome measures. All of the included studies used strength, mobility/flexibility, sensorimotor, or a combination of these types of exercise therapy. The most reported outcome measure for pain was the Visual Analogue Scale (VAS) and the Disability of Arm, Shoulder, and Hand (DASH) for function. Large effect sizes for pain and function were found for 18 studies,^{30,32–39,41–49} medium effect sizes for six studies,^{31,34,36,40,41,50} and small or no effects for four studies.^31,35,41,44 Table 2 includes the effect sizes for all studies. Due to the heterogeneity of the data, a meta-analysis of effect sizes was not possible.

Number of exercise therapy sessions and supervision provided

The total number of exercise therapy sessions for studies assessing functional outcomes ranged from 18 to 204^{30–33,35–40,42,43,45–47,50} and from 24 to 204 for pain outcomes^{30–33,35,36,38–40,43,45–47,49,50} over a duration of three to 12 weeks, and included both supervised and unsupervised sessions.

Supervised sessions only

Four groups included supervised exercise therapy sessions only (range 18–36; average 28.5), with large effect sizes calculated for pain, and large and medium effect sizes for function.^36,37,46 Two groups had large effect sizes for pain and function (range 24–36; average 30).^36,46 One group had a large effect size for function (18 sessions), ³⁷ and one group had a medium effect size for function and a large effect size for pain (36 sessions). ³⁶

Unsupervised sessions only

Four groups included only unsupervised exercise therapy sessions (range 28–70; average 56) with medium and large effect sizes for pain and function.^47,49,50 Two groups had large effect sizes for pain and function (range 56–70; average 63).^43,47 One group had a large effect size for pain only (28 sessions) ⁴⁹ and one group had medium effect sizes for both pain and function (70 sessions). ⁵⁰

Both supervised and unsupervised sessions

Twelve groups included both supervised and unsupervised exercise therapy sessions (supervised sessions: range 4–36; average 8.17/unsupervised sessions: range 21–168; average 83.42/total sessions: range 27–204; average 91.58) with no effect on large effect sizes for pain and function.^{31–33,35,38–40,42,45} Eight groups had large effect sizes for both pain and function (supervised sessions: range 1–10; average 6.5/unsupervised sessions: range 56–169; average 91/total visits: range 60–177; average 97.5).^{32,33,38,39,42,45} One study had no effect and large effect sizes for various pain-related outcomes and a large effect size for functional outcomes (supervised sessions: 36/unsupervised sessions: 168/total sessions: 204). ³⁵ Medium effect sizes were calculated for one group for pain and function (supervised sessions: 6/unsupervised sessions: 21/total sessions: 27). ⁴⁰ One study with two groups had no effect to small effect sizes for pain and no effect to medium effect sizes for function (supervised sessions: 2; unsupervised sessions: 42; total sessions: 44). ³¹

Supervision level not reported

Six groups did not report, or were unclear, in describing whether exercise delivery was supervised or unsupervised.^{30,34,41,44,48} Three groups had large effect sizes for pain and function,^30,34,48 one group had medium and large effect sizes for pain and large effect sizes for function. ³⁴ One group had no effect, medium and large effect sizes for various pain outcomes, and large effect sizes for functional outcomes. ⁴¹ One study, which did not report pain outcomes, had no effect, small and large effect sizes for functional outcomes. ⁴⁴

Single session duration

Single session durations were reported in 12 of the 21 studies, ranging from 5 to 90 min.^{36–40,42–46,49,50} Three of the four studies^31,35,41 with small to no effect sizes did not report this information, and three of the six studies^31,34,41 with medium effect sizes did not report this information.

Frequency of intervention

Frequencies ranged from three to 21 exercise therapy sessions per week, at home or in a clinic, for treatment arms with pain-related outcomes. Two times per day (14 times per week) was the most common frequency.^{32,33,35,38,39,45,50} In the 23 treatment arms, where functional outcomes were reported, the frequency of intervention ranged from three to 21 sessions per week.^{30–33,35–40,42–45,47,48,50} Four studies did not report, or were unclear, on the frequency of exercise therapy sessions prescribed.^34,41,44,48 Large effect sizes for pain and functional outcomes were determined with a frequency of sessions ≥14 times per week regardless of supervision.^{32,33,35,38,39,42,45,50} Medium effect sizes for pain and function were determined with frequencies ranging from three to 14 times per week regardless of supervision.^31,36,40,50 Small effect sizes for pain and function were calculated in one group with combined (supervised/unsupervised) exercise therapy sessions with a frequency of seven times per week. ³¹ One study indicated no effects for pain with a frequency of 14 times per week. ³¹ Only one combined supervision study indicated no effects for pain and reported a frequency of seven times per week. ³⁵

Duration of care

The duration of care for studies assessing functional outcomes ranged from three to 24 weeks and three to 12 weeks for pain outcomes. The most common durations of care were four, six, and 12 weeks. Large effect sizes for functional outcomes were observed in five groups with a total duration of care of four weeks,^30,38,41,45 one with five weeks, ³⁹ five with six weeks,^32,37,42,43 three with eight weeks,^46–48 and seven with 12 weeks.^33–36 One outlier study, with a reported 12–24 week duration, had small effect sizes using the 36-Item Short Form Survey (SF-36) and Constant Murley (CM) functional scales. ⁴⁴ Large effect sizes were found for pain outcomes in six groups at four weeks,^{30,38,41,45,49} four groups at six weeks,^32,42,43 three at eight weeks,^46–48 and six at 12-week durations.^33–36 Pain outcomes were not investigated beyond 12 weeks of intervention in any of the studies.

Exercise reproducibility

TIDieR checklist

All studies reported on two TIDieR domains: “brief name” of the intervention and the “why” behind the rationale for the intervention. All other items lacked reporting from one or more studies. As noted in Figure 3, the least reported checklist items were in the domains of “modifications,” or changes to the intervention on a study design level (n = 2; 10.0%), “tailoring,” or how exercise personalized or adapted for each participant (n = 9; 43.0%), and “how well treatment fidelity was actually monitored” (n = 10; 48.0%). Most studies provided sufficient information for “what materials,” physical or informational were provided or required by the patients (n = 18; 86.0%), “what procedures, activities, or processes” were used in the intervention (n = 17; 81.0%), and “who provided” the intervention to include discipline and expertise (n = 17; 81.0%).

Figure 3.

Reported Items on the Template for Intervention Description and Replication (TIDieR).

The CERT

Only one study reported all of the items on the CERT. As noted in Figure 4, the least frequently reported checklist item was “description of non-exercise components” (item 10; n = 4; 19%). “Description of motivation strategies” (item 6; n = 6; 29%) and “description of any home exercise component” were also poorly reported (item 9; n = 6; 29%). Most studies reported “description of the setting where exercise was performed” (item 12; n = 15; 71%), “whether exercises were supervised or not” (item 4; n = 16; 76%), and “whether exercises were generic or tailored” (item 14(a); n = 14; 67%). Further breakdown of the checklist may be viewed in the Supplemental Data .

Figure 4.

Reported Items on the Consensus on Exercise Reporting (CERT) Checklist.

Modifications from registration

The original search date was extended to include studies from July 2020 to August 2021 to include the most recent evidence during the manuscript review process. The CERT was added to provide further detail about exercise therapy intervention reporting beyond the TIDieR checklist, which was originally planned in isolation.

Discussion

Most studies included in this review exhibited large effect sizes for pain and/or function and suggest that exercise therapy may be beneficial for the treatment of SAPS.

The findings of this review include: (1) the use of both supervised and unsupervised exercise therapy sessions was associated with large effect sizes, (2) trends for other optimal exercise therapy session dosing parameters could not be identified, and (3) exercise therapy intervention reporting lacks sufficient detail for clinical replication.

Exercise therapy session dosing parameters

When all exercise therapy sessions were performed under supervision, effect sizes were almost always large.^36,37,46 Because only a small number of studies used this supervised exercise therapy only approach, these results should be interpreted with caution. When unsupervised exercise therapy was used in isolation,^43,47,49,50 a greater number of sessions were reported to obtain similar results. If both supervised and unsupervised exercise sessions were included, fewer supervised visits, but more unsupervised visits resulted in large effects sizes for pain and function. This may suggest the importance of the role of unsupervised exercise when a combination of supervision types is used.

Making definitive recommendations about the optimal type of supervision and number of sessions is complicated by the fact that adherence to home exercise programs was poorly reported, and the actual number of sessions may not have been the same as the prescribed number of sessions. Using a combination of both supervised and unsupervised exercise therapy sessions is a more pragmatic patient care approach, and the results from this review found this approach to be associated with large effect sizes. These effects could be achieved with an average of 6.5 (range 1–10) supervised sessions and 91 (range 27–204) unsupervised sessions (including multiple sessions a day).^{32,33,38,39,42,45} There is a currently a dearth of evidence to support any one type of supervision, or combination of supervision types, in the treatment of musculoskeletal disorders and further research into optimal supervision levels will shine light on this clinical question. Interestingly, a recent study found that best practice advice (including instruction on self-guided exercises) compared to six supervised physical therapy sessions did not produce significantly different outcomes for SAPS. ⁵¹ This reiterates that the level of exercise supervision may not be important.

Three main types of exercise therapy were included in this review (strength, flexibility/mobility, sensorimotor, or a combination). One study ³¹ demonstrated a small effect size for shoulder-specific functional outcomes measures with the use of both supervised and unsupervised eccentric rotator cuff exercise therapy sessions. Other studies which used general rotator cuff strengthening programs, or a combination of scapular strengthening, flexibility, or sensorimotor training often yielded larger effect sizes. It could be tempting to suggest that adding a larger variety of exercises to the patient’s treatment plan may lead to better outcomes; however, this review does not provide sufficient data to be able to suggest that one type of exercise is superior to another.

A variable range of single session durations was associated with large effect sizes, suggesting clinicians may have successful outcomes with exercise therapy sessions of varying durations (5–90 min). Specific approaches such as 5 min of prone pendulum swings and passive stretching ⁴⁵ or 60 to 90 min of phased shoulder complex strengthening could be considered given the large effects associated with both these approaches.^36,44,48 Individual patient considerations, such as acuity, irritability or the specific patient population may play a role and should be considered when planning the duration of exercise therapy sessions. Many studies did not report this information and future trials should further highlight single session duration.

Some variability in effect sizes for pain may be related to the measurement of “resting level of pain” versus “other pain measures” (worst/best/current/composite/night). Research has suggested outcomes, such as movement-evoked pain and resting level of pain, may have different mechanisms of action which could explain varying responses to therapeutic interventions.^52,53 Likewise, variability observed in functional outcomes may be related to the intent of the outcome measure, where some measure “general function” (36-Item Short Form Survey (SF-36) and Functional Assessment Questionnaire) and others measure “shoulder specific function” (DASH, CM, etc.).

There was significant variability across studies for the duration of care, limiting the ability to create an optimal recommendation. Duration of care should not be considered in isolation. For example, a higher number of exercise therapy sessions (higher frequency) within a defined duration of time may have different outcomes than a lower number of exercise therapy sessions in the same time frame. Findings from this review further suggest that optimal trends for frequency of exercise therapy sessions could not be established due to mixed outcomes associated with various prescribed frequencies.

Exercise reproducibility

TIDieR checklist

The overall reporting of plans to assess fidelity or adherence to exercise therapy interventions, as well as actual follow-up of intervention adherence, was not consistently reported. This could imply that the true number of exercise therapy sessions may actually be lower, if compliance was just assumed for everyone. This makes concrete prognostic information regarding exercise adherence difficult to discern. “Tailoring” of the exercise program was another domain with partial reporting. Describing the details on “when and how much,” “what procedures,” “what materials,” “why,” and “how” may require more contemplation during study design, and reporting these details may provide practicing clinicians with better information to be able to tailor exercise based on patients’ unique needs during a treatment session.

It was surprising that many studies omitted information on “when and how much” exercise was delivered, including the number of sessions, the schedule and duration of visits and its intensity or dose.^{30,34,39,41,46,48,49} Attaching descriptive exercise appendices with these details may better equip clinicians with tools to improve therapeutic outcomes of pain and function. The actual adherence (described as the extent to which the delivered intervention varied from the intended intervention) ¹⁸ was infrequently described with sufficient detail. Only two studies included in this review with large effect sizes for pain and function reported short term exercise adherence rates of 80% to 90%.^32,42 Rates of self-reported exercise adherence have been recorded as 25% to 41% within different patient populations. ⁵⁴ Longer term adherence, up to one year after discharge from the clinic, has been estimated as low as 20%.^55–57 Because the actual number of exercise sessions performed is unknown without accurate monitoring of adherence, clinicians may consider the real possibility that these dosing parameters reflect only prescribed exercise and not what was actually completed. It is important for researchers to include a method to track and describe exercise adherence to report actual treatment fidelity.

The CERT

Further investigation of exercise reporting via the CERT also demonstrated significant issues with the reporting of critical components of exercise therapy. The “how” of delivery of exercise was the item category most insufficiently reported. First, the “description of any non-exercise components,” such as education, was most often omitted. Education before, during, or after exercise may influence the ultimate treatment outcome and is important to consider as it is often included with exercise.^58,59 Second, the “detailed description of the home exercise program” was infrequently reported. Home exercises contribute to dosing frequency and a thorough understanding of this intervention is necessary to be able to replicate a study’s intervention. Third, the “motivational strategies” used for the intervention were insufficiently reported in the majority of the studies, and these details could influence overall patient engagement and assist the therapist in improving patient adherence. Fourth, transparency with “actual exercise therapy intervention adherence reporting” may provide an explanation for the effect or lack of effect of an intervention and should be reported. Lastly, most of the studies did not provide a “detailed description of the decision rules for determining exercise progression,” and more specifically, “how the overall program was progressed.” Starting rules and clear progression are necessary for replication, ²⁰ and may also improve treatment efficiency.

Consistent with the TIDieR checklist, detailed “descriptions of tailored exercises” were often omitted, and “determining how individuals would start their program” were lacking. Details on appropriate starting rules and individual tailoring of activities are sensible so that individuals are not under or over dosed. This may also influence patient engagement and participation ²⁰ and may ultimately result in better clinical outcomes. The CERT was consistent in identifying that “how well the intervention actually was delivered as planned” was not reported sufficiently. The current literature supports exercise therapy as a first-line defense for SAPS, ⁷ but most of the included studies using exercise therapy insufficiently report the necessary details to be able to replicate the intervention in the clinic.

The TIDieR checklist and the CERT were created to improve the focus and reporting of interventions, including exercise therapy dosing, in published trials. The deficiencies noted in intervention reporting highlight the deviation of several studies from the most commonly used definition of exercise therapy dosing, which includes the number of repetitions, sets, frequency, intensity, and duration of exercise prescription. ¹³ Increased use of these reporting tools may lead to the recording of dosing parameters that are more compatible with accepted definitions. If researchers incorporate these tools more frequently, the definition of exercise therapy dosing would be more relevant and follow-up studies investigating exercise therapy dosing parameters may provide more robust findings.

Other considerations influencing exercise therapy

The timing when exercise therapy is implemented may be an important consideration with evidence suggesting that earlier use of exercise may result in better outcomes. ⁶⁰ In a secondary analysis, Kromer et al. ⁶¹ concluded that duration of complaints and baseline disability were the main factors influencing disability change scores in patients with SAPS.

Patient beliefs about exercising through pain may have influenced some outcomes. ⁶² Contemporary research in the field of musculoskeletal pain suggests that psychosocial factors can contribute to patients’ pain presentations and their response to treatment. ⁶³ Studies included in this review did not account for the influence of pain beliefs or perceptions on outcomes.

The studies included in this review described a variety of exercise approaches, and large effect sizes for improvements in pain and/or function were identified across the various examples of exercise types used. Variables outside this investigation, such as rest intervals, effort levels, training tempo, and magnitude of mechanical tension may account for some variability in effect sizes. ⁶⁴ Also, optimal exercise therapy dosing parameters for use in physical therapy may not be the same as the parameters used in non-rehabilitation settings (e.g. strength and conditioning). Manipulation of training frequency, intensity, rest and recovery days, order, session format, movement velocity, load, and intensity may be relevant strength and conditioning methodology to consider in future studies designed to investigate exercise therapy for its effect size.^65,66

Study limitations

There are several limitations that warrant discussion. Definitive conclusions were challenging due to a lack of detailed reporting in most studies. Therefore, conclusions were drawn based on the limited available information from these studies. Second, defining what constitutes an exercise therapy session and/or exercise in general proved challenging. Some exercise prescriptions only included one exercise whereas others included multiple exercises. One study included a brief upper trapezius relaxation exercise five times per day, which was not classified as an exercise therapy session and excluded from frequency reporting, whereas other authors may have chosen to include it. ⁴⁷ Third, the diversity of the outcome measures used to assess pain and function across the 21 studies makes it challenging to draw direct comparisons. Fourth, supervised versus unsupervised exercise adherence remains an important element of any exercise program. More than half of the studies did not report on the actual fidelity of the home exercise interventions when it was included.^{30,35,36,38,39,41,44,45,48–50} Thus, total actual sessions could be lower or higher than what was formally prescribed and could have substantially impacted the observations and conclusions of this review. Fifth, most of the studies in this review followed participants for 12 weeks or less, making longer term recommendations impossible. Sixth, the use of within-group differences to calculate effect sizes may be controversial despite previously being used in other professional publications.^25,26 It has been disputed that within-group effect size calculations do not measure a true effect and may misrepresent findings by including regression to the mean, natural recovery, nonspecific effects, and treatment effects. ⁶⁷ Seventh, participant numbers within the subgroups were small and likely led to an underpowered analysis of effect size, which could bias our findings. Lastly, because a meta-analysis was not possible, we were not able to make strong recommendations for one dosing parameter over another, but instead were limited to describing dosing parameters associated with the largest effect sizes.

Conclusions

Most exercise therapy groups in this review, regardless of supervision, were related to large effect sizes for pain and functional outcomes, but no specific individual dosing parameters were associated with these effects. The use of both supervised and unsupervised sessions was associated with large effect sizes for pain and function, the most common design was a combination of supervised and unsupervised sessions. The results suggest, with moderate certainty, that the effect size of pain and function was not influenced by exercise therapy dosing parameters. Based on the included studies, a conditional, strong recommendation is made for treating SAPS with a variety of exercise therapy dosing parameters. These findings highlight the importance of exercise therapy and extending the care plan outside of the traditional supervised setting, to ensure the patient is an active participant in their outcomes. Overall reporting of exercise therapy parameters was poor, limiting conclusions about optimal exercise dosing and likely affecting reproducibility. Further research, with emphasis on detailed exercise intervention reporting, may improve our understanding of the impact of dosing on outcomes for individuals with SAPS.