Physiotherapy for patients with soft tissue shoulder disorders: a systematic review of randomised clinical trialsBMJ 1997; 315 doi: https://doi.org/10.1136/bmj.315.7099.25 (Published 05 July 1997) Cite this as: BMJ 1997;315:25
- Geert J M G van der Heijden, senior researchera,
- Daniëlle A W M van der Windt, research fellowb,
- Andrea F de Winter, research fellowb
- a Institute for Rehabilitation Research, PO Box 192, 6430 AD Hoensbroek, Netherlands
- b Institute for Research in Extramural Medicine, Vrije Universiteit, Amsterdam, Netherlands
- Correspondence to: Dr van der Heijden
- Accepted 15 April 1997
Objective: To assess the effectiveness of physiotherapy for patients with soft tissue shoulder disorders.
Design: A systematic computerised literature search of Medline and Embase, supplemented with citation tracking, for relevant trials with random allocation published before 1996.
Subjects: Patients treated with physiotherapy for disorders of soft tissue of the shoulder.
Main outcome measures: Success rates, mobility, pain, functional status.
Results: Six of the 20 assessed trials satisfied at least five of eight validity criteria. Assessment of methods was often hampered by insufficient information on various validity criteria, and trials were often flawed by lack of blinding, high proportions of withdrawals from treatment, and high proportions of missing values. Trial sizes were small: only six trials included intervention groups of more than 25 patients. Ultrasound therapy, evaluated in six trials, was not shown to be effective. Four other trials favoured physiotherapy (laser therapy or manipulation), but the validity of their methods was unsatisfactory.
Conclusions: There is evidence that ultrasound therapy is ineffective in the treatment of soft tissue shoulder disorders. Due to small trial sizes and unsatisfactory methods, evidence for the effectiveness of other methods of physiotherapy is inconclusive. For all methods of treatment, trials were too heterogeneous with respect to included patients, index and reference treatments, and follow up to merit valid statistical pooling. Future studies should show whether physiotherapy is superior to treatment with drugs, steroid injections, or a wait and see policy.
Because of the small sample sizes and unsatisfactory methods of most trials, only a few randomised clinical trials of methods of physiotherapy in patients with soft tissue shoulder disorders allow firm conclusions on effectiveness of treatment
When compared with placebo or another treatment, ultrasound therapy seems ineffective in patients with shoulder disorders
Evidence is insufficient to support the effectiveness of low level laser therapy, heat treatment, cold therapy, electrotherapy, exercise, and mobilisation in such patients
Future trials should focus on the effectiveness of exercise and mobilisation in comparison to analgesics, non-steroidal drugs, steroid injections, and advice and a wait and see policy
Special attention should also be given to the principles of adequate design and conduct of trials and the standards of reporting
Pain is the primary symptom in most patients with shoulder disorders affecting the soft tissue. In many patients, painful restriction of the range of shoulder movement limits the ability to perform daily activities. Estimates of the cumulative annual incidence of shoulder disorders vary from 7 to 25 per 1000 general practice consultations.1 2 3 Five per cent of all general practice consultations are reported to be related to shoulder disorders.4 5 Half of all presented episodes resolve within six months, but some last a year or more. Most patients with such disorders are treated in primary care. Their management includes advice, analgesics, non-steroidal anti-inflammatory drugs, steroid injections, and physiotherapy. Evidence from randomised clinical trials on shoulder disorders shows small effects favouring the effectiveness of non-steroidal drugs6 and steroid injections.7 A wide array of physiotherapy methods is used to treat shoulder disorders.8 9
Patients are often referred for physiotherapy10 11; in the Netherlands as many as a third of all patients with shoulder disorders are referred.2 3 12 So far, little effort has been invested in establishing the effectiveness of management with physiotherapy. We examined whether certain methods in physiotherapy are effective for patients with soft tissue shoulder disorders by reviewing reports of 20 randomised clinical trials.
Selection of studies
Relevant trial reports were harvested from Medline (Index Medicus January 1966 to December 1995) and Embase (Excerpta Medica January 1984 to December 1995) according to the computerised search strategy of Dickersin et al.13 This strategy was supplemented with citation tracking of relevant publications. GH identified trial reports that met the following five conditions: firstly, patients had shoulder pain at inclusion; secondly, treatments were allocated by a random procedure; thirdly, at least one of the treatments included physiotherapy; fourthly, success rate, pain, mobility, or functional status were included as outcome measures; and, finally, results were published as a full report before January 1996. From this selection DW and AW independently selected the trials that included patients with soft tissue shoulder disorders.
Assessment of methods
To assess trial methods, eight criteria for internal validity were used (box). These criteria are based on generally accepted requirements of methods for design and conduct of intervention research.14 15 16 17 In addition, five data display and extraction criteria (box) were used to provide information on the feasibility of statistical pooling.18
Validity criteria for assessment of methods of trials
Enrolment—Restriction to a homogeneous population with respect to prognosis and susceptibility to allocated interventions by explicit selection criteria. Used as prognostic indicators: age, duration of complaint, painful arc, pain at night, number of relapses, radiating pain, previous treatment
Randomisation—Adequate procedure for generation of random numbers list and concealed allocation of interventions
Similarity at baseline—Similarity of intervention groups at baseline with respect to prognosis and susceptibility to allocated interventions. Used as prognostic indicators: baseline scores for outcome measures, age, duration of complaint, painful arc, pain at night, number of relapses, radiating pain, previous treatment
Withdrawals from treatment—No patients withdrew from treatment or number of patients was <10% in each group, with comparable reasons for withdrawal
Missing values (for example, loss to follow up)—Number of randomised patients minus number of reported patients at main moment of effect measurement for main outcome measure—if not stated according to reviewers—divided by all randomised patientsx100 was <10% in each group
Cointerventions—Either standardised or excluded in trial design
Blinded application of interventions—Therapists: blinding by credible placebo. Patients: blinding by credible placebo or by enrolment of patients who were naive to allocated interventions
Blinded assessment of outcome—Assessor of effect of variables (for example, patient, therapist, physician, or research staff) blinded for allocated interventions
Data display and extraction criteria
Sample size of groups
Standardisation of allocated interventions—Adequate description of type, method, application of technique, intensity, duration, number, and frequency of sessions for all allocated interventions
Reported outcome variables—Success rate (for example, proportion of patients cured or improved); pain; functional state (activities of daily living); mobility (range of movement); non-trial cointerventions (for example, drugs or surgery)
Outcome assessments—Identical timing of assessment for all intervention groups: immediately after last treatment or over three months
Actual data for outcome variables—An adequate point estimate is presented for each intervention group (with corresponding distribution measure) for success rate or improvement for pain or most important outcome measure on most important moment of effect measurement
We independently analysed the completeness of information from the selected trial reports. For each criterion we logged whether incomplete information had hampered the assessment of methods. If sufficient information was given we judged and logged whether bias was likely or not. For criteria for which consensus could not be reached, the presented results are based on agreement of two reviewers. Subsequently, the trials were ranked according to the number of validity criteria for which bias was considered to be unlikely.
Success rates were determined for each intervention group by dividing the number of documented successes at the end of the intervention period by the number of patients randomly allocated to the intervention (that is, intention to treat analysis). When success rates could not be calculated, we determined change in scores for pain and mobility ratings. Missing values for outcome measures were assumed to represent failures (that is, worst case assumption). Next, to judge the effectiveness of treatments we calculated the differences between groups for outcome measures, with 95% confidence intervals. Finally, to draw conclusions we related these confidence intervals to the number of satisfied validity criteria.
GH identified 47 trial reports that met the five conditions for further selection. DW and AW excluded 24 trials: seven in which the results of patients who received physiotherapy for shoulder disorders were not presented separately, one in which similar physiotherapy was given as a cointervention to all patients, four on exercise therapy after mastectomy, four on physiotherapy for shoulder pain after fracture, seven on physiotherapy for shoulder pain in hemiplegic subjects, and one trial on rheumatoid arthritis. The methods of the remaining 23 trial reports were assessed.19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 Information was combined for three trials that were reported twice.23 33 23 Hence, the systematic review included 20 trials on the effectiveness of physiotherapy for patients with soft tissue shoulder disorders.
Assessment of methods
Table 1 lists for each trial the validity criteria for which bias was considered likely. This table also presents the validity and data display and extraction criteria for which incomplete information had hampered the assessment of methods. The trials are ranked according to the number of validity criteria that were satisfied. Equally ranked trials are ordered alphabetically.
Validity criteria–Eleven of the 20 trials satisfied at least four of the eight validity criteria. One trial satisfied all eight19; three other trials satisfied six.20 21 22 Three trials seemed to be flawed by a large proportion of withdrawals from treatment28 31 37; two trials by a large proportion of missing values26 37; nine trials by insufficient blinding of intervention25 27 28 31 35 36 37 38 39 and three trials by a insufficient blinding of outcome assessment.25 30 39 Many reports lacked adequate information on several validity criteria. The randomisation procedure was adequately reported for one trial19 and prognostic status at baseline for four trials19 25 28 30; information on cointerventions was often insufficient.
Data display and extraction criteria–In general the sample sizes of the studies were small: six trials compared groups of 25 or more patients25 26 29 33 35 39 and six trials compared groups of 15 to 25 patients.21 27 28 30 36 40 All other trials included smaller study populations. Data on outcome measures were poorly reported. Of the 11 trials with acceptable methods,19 20 21 22 23 25 26 27 28 29 30 five provided sufficient data for the calculation of 95% confidence intervals.19 23 25 26 30 Such calculation was possible for six of the nine remaining trials with unsatisfactory methods.
Characteristics of trials
Table 2 outlines the study population, intervention, follow up, and reported results of the assessed trials. Again, the trials are ordered by the number of fulfilled validity criteria. In nine trials participation was restricted to narrowly defined diagnostic categories (for example, rotator cuff tendinitis),19 20 22 23 25 26 28 30 33 whereas other trials included a wide variety of soft tissue disorders (for example, painful shoulder, periarthritis humeroscapularis). In eight trials, duration of symptoms at baseline was not specified as an entry criterion.23 27 28 30 35 37 38 40 Another eight trials included patients who, at baseline, had had their symptoms for less than three months,19 21 22 29 31 32 33 36 whereas in the four remaining trials duration of symptoms at baseline exceeded three months.
Ultrasound therapy was studied in six trials,19 23 28 29 38 39 different methods of thermotherapy in three trials,37 38 40 and low level laser therapy in four trials.22 30 32 33 Three trials concerned magnetotherapy,20 21 26 three concerned manipulations or mobilisations,27 31 36 two trials involved electrotherapy28 35 or cold therapy,31 39 and one trial evaluated an exercise programme.25
Six trials compared various methods of physiotherapy,26 28 31 35 38 39 nine trials compared physiotherapy with placebo treatment,19 20 21 22 25 29 30 32 33 and 10 trials compared physiotherapy with another intervention (mainly analgesics, non-steroidal drugs, and steroid injections).23 25 27 31 32 37 38 39 40 41 Furthermore, two trials included a control group without any treat- ment.21 31 Results from long term follow up (at least two months after randomisation) were available from four trials.21 25 36 39 Follow up in all other trials was restricted to assessment of outcome directly after completion of treatment, usually three or four weeks after randomisation.
Effectiveness of treatment
The validity of four of the six trials that studied the effect of ultrasound therapy was acceptable, but none of these trials showed evidence in its favour.19 23 28 29 Ultrasound therapy was no better than cold therapy and steroid injections,39 non-steroidal anti-inflammatory drugs and acupuncture,23 transcutaneous electrical stimulation,28 and analgesics and iontophoresis.38 Moreover, ultrasound therapy did not seem to be effective in placebo controlled trials.19 23 29 The validity of two of the four trials that studied the effectiveness of low level laser therapy was acceptable.22 30 Saunders could not find significant differences between active and placebo laser.22 Our calculations of the results of Vecchio et al showed very small differences in favour of active low level laser therapy, though the authors, using different statistical methods, did not find significant differences.30 The two other trials with unsatisfactory methods reported effects in favour of the short term effectiveness of low level laser therapy compared with placebo32 33 or with non-steroidal drugs.32
Transcutaneous electrical stimulation did not seem to be more effective than ultrasound therapy28 or than other electrical methods.35 We could not find any placebo controlled trial on electrotherapy. The two placebo controlled trials on pulsed electromagnetic fields had acceptable validity and reported favourable results for treatment.20 26 The results of Chard et al, however, were non-significant when they were analysed according to the intention to treat principle.26 Magnetic treatment seemed to be ineffective when it was compared with no treatment.21
Cold therapy was no more effective than ultrasound therapy,39 steroid injection,31 39 mobilisations, or no intervention.31 Different methods of thermotherapy were not more effective than placebo37 38 or steroid injections and analgesics.40
Exercises were as effective as surgery in patients with a stage II impingement syndrome and were more effective than placebo laser therapy.25 When they were compared to no intervention,31 36 mobilisations and manipulations did not contribute to recovery nor were they superior to steroid injections27 31 or cold therapy.31
This systematic review, based on the reports of 20 randomised clinical trials, evaluated whether physiotherapy contributes to the extent and speed of recovery for patients with soft tissue shoulder disorders. It used an assessment of methods to minimise bias.
The validity of the methods of 11 of the 20 assessed trials was satisfactory. One trial reported all the information needed for assessment of validity and data display and extraction.19 Many trials did not provide sufficient information for at least two validity criteria. This poor reporting might hide flaws; thus it hinders the interpretation of trial results. This lack of information was most prominent for the randomisation procedure, baseline similarity of treatment groups, and cointerventions.
Schulz et al provided empirical evidence of bias for trials with inadequate concealment of treatment and lack of blinding.42 Lack of prognostic comparability at baseline, withdrawals, and missing data are also related to success of treatment and therefore represent major sources of bias.43 44
Effectiveness of treatment
Deficiencies in the presentation of data often hampered calculation of 95% confidence intervals. When we could calculate confidence intervals they were wide and included zero, even when trials had acceptable methods.19 23 26
Few of the assessed trials favoured the effectiveness of physiotherapy. The type of control treatment seemed unrelated to the study results. Because there were many small trials with negative results, statistical pooling of the results of trials with acceptable methods would have been useful. However, we considered that the few valid trials on the same methods of treatment (for example, ultrasound therapy or low level laser therapy) were too heterogeneous with respect to administration (for example, intensity, duration, and frequency of administration), the compared treatment (for example, placebo, no treatment, or alternative control treatment), the selection of study populations (for example, regarding specific soft tissue disorders or symptom duration at baseline), and follow up (for example, timing of outcome assessment and choice of outcome measures) to merit statistical pooling.
Given the adequate methods of placebo controlled trials on ultrasound therapy, this method does not seem to be effective in treating patients with shoulder disorders. One placebo controlled trial with adequate methods reported superior effectiveness of pulsed electromagnetic fields. All other trials that reported significant results were small and had unsatisfactory methods. Thus there is insufficient evidence to draw conclusions on the effectiveness of low level laser therapy, heat treatment, cold therapy, electrotherapy, exercise, and mobilisations.
The purpose of treating patients with shoulder disorders is to increase the extent and speed of recovery. As ultrasound therapy is not effective, any further application in patients with shoulder disorders should be discouraged. This can be done by updating treatment guidelines or by withholding reimbursement for its use.
Future trials should show whether other methods of physiotherapy for shoulder disorders are effective. This should be particularly interesting for exercise and mobilisations, which have rarely been subjected to scientific scrutiny in randomised clinical trials despite being commonly used in patients with shoulder disorders. Priority should be given to a comparison of exercise and mobilisations with analgesics and advice and a wait and see policy. As there are some indications for their effectiveness, steroid injections and non-steroidal drugs are other relevant comparative treatments. During the design and execution of future trials specific attention should be given to the control of prevalent flaws, such as many withdrawals, many missing results, and a lack of blinding during treatment and assessment of outcome. Standards of reporting trials should prevent confusion about the validity of trial methods and ensure adequate data analysis and presentation of pertinent data.16 45
We thank Pieter Leffers and Paul Knipschild (department of epidemiology, Maastricht University, Netherlands) and Lex Bouter (Institute for Research in Extramural Medicine, Vrije Universiteit, Amsterdam, Netherlands) for their comments on the draft of this paper.
Funding: No external funding.
Conflict of interest: None.