Australian Physiotherapy Journal

Journal of Physiotherapy 61 (2015) 155 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Trial Protocol A biopsychosocial primary care intervention (Back on Track) versus primary care as usual in a subgroup of people with chronic low back pain: protocol for a randomised, controlled trial Reni MA van Erp a, Ivan PJ Huijnen a,b,c, Jeanine A Verbunt a,b,c, Rob JEM Smeets a,b,c a Department of Rehabilitation Medicine, CAPHRI, Maastricht University; b Department of Rehabilitation Medicine, Maastricht UMC+, Maastricht; c Adelante, Centre of Expertise in Rehabilitation and Audiology, Hoensbroek, The Netherlands Abstract measures are: credibility and expectancy, anxiety and depression, catastrophising, pain intensity, kinesiophobia, self-efficacy, partici- Introduction: Multidisciplinary biopsychosocial interventions are pant’s global perceived effect, cost-effectiveness, and cost-utility effective at improving functional disability in people with chronic estimated with cost diaries and quality-adjusted life years. Analysis: low back pain. However, these interventions are often expensive Linear mixed models using an intention-to-treat principle. Incre- and have long waiting times before treatment starts. Therefore, mental cost-effectiveness and cost-utility ratios will be calculated implementing biopsychosocial interventions in primary care and plotted on a cost-effectiveness plane. Discussion: This study settings may be of interest. Because people with chronic low back will provide useful information on a biopsychosocial intervention pain show different biopsychosocial profiles, they might respond for chronic low back pain in primary care settings. differently to specific interventions. Research questions: This study will investigate the difference in (cost) effectiveness Trial registration: Clinicaltrials.gov. Registration number: between a biopsychosocial primary care intervention, Back on NCT02220543. Was this trial prospectively registered: Yes. Track, and primary care physiotherapy as usual in a subgroup of Funded by: Adelante, Centre of Expertise in Rehabilitation and adults with chronic low back pain. Design: Double-blind, multi- Audiology Hoensbroek, The Netherlands; the Province of Limburg centre (n = 8), randomised, controlled trial. Participants: Eighty- and CZ Foundation. Funder approval number: SAS-2012-01300, six adults with chronic low back pain, aged 18 to 65 years, AFVV12-205. Anticipated completion: January 2017. Prove- experiencing low to moderate levels of disability and in whom the nance: Invited. Not peer-reviewed. Correspondence: Reni MA contributing role of psychosocial factors to this disability is van Erp, Department of Rehabilitation Medicine, CAPHRI, Maas- restricted. Intervention: The Back on Track intervention: four tricht University, Maastricht, The Netherlands. Email: reni. individual and eight group sessions, based on biopsychosocial [email protected] approaches from multidisciplinary pain rehabilitation programs and provided by trained physiotherapists. Control: Primary care Full protocol: Available on the eAddenda at doi:10.1016/j.jphys. physiotherapy as usual. Measurements: The primary outcome is 2015.03.003 functional disability (Quebec Back Pain Disability Scale) at post treatment, and 3-month and 12-month follow-up. Secondary http://dx.doi.org/10.1016/j.jphys.2015.03.003 1836-9553/ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 61 (2015) 160 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers A web-based therapy program enhances occupational performance and visual perception in children with unilateral cerebral palsy Synopsis Summary of: James S, Ziviani J, Ware R, Boyd R. Randomized Melbourne Assessment of Unilateral Upper Limb Function. controlled trial of web-based multimodal therapy for unilateral Secondary outcome measures were: the Canadian Occupational cerebral palsy to improve occupational performance. Dev Med Child Performance Measure and the Test of Visual Perceptual Skill. Neuro. 2015; doi:10.1111/dmcn.12705. Results: Ninety-two participants completed the study. The intervention group completed 32 hours (95% CI 4 to 75 hours) of Question: Does a web-based therapy program improve the program. At the end of the intervention, the Assessment of occupational performance, upper limb function and visual Motor and Process Skills motor scores were significantly greater in perception in children with unilateral cerebral palsy? Design: the treatment group by 0.28 logits (95% CI 0.17 to 0.39). The Randomised controlled trial with concealed allocation and blinded Jebsen-Taylor Test of Hand Function for the dominant limb outcome assessment for some, but not all, outcome measures. improved significantly more in the intervention than the control Setting: Home environment in Australia. Participants: Children group (–DTFI43[_]$ .7 seconds, 95% CI –[4_D$7]FTI .4 to –2_[T]FD$5I .0), but not for the impaired with spastic-type unilateral cerebral palsy, Manual Ability limb (–]D$_6[TF2I 2.0 seconds, 95% CI –4_F]D$[I7T 4.8 to 0DT_]F[I$8 .7). The intervention group Classification System levels I to III and Gross Motor Function had higher scores on the Canadian Occupational Performance Classification System levels I or II. Children were aged 8 to 18 years, Measure performanceD]F[I1T_$ (I_]$9F[D1T .3 points, 95% CI $D1_[T00]FI .7 to 1TD$[IF1_] .9) and had sufficient understanding and co-operation to perform the satisfaction (21[TD$IF]1_ .5 points, 95% CI 1DIF[$03]T_ .4 to TD1]_4I[0F$ .8) FIDs$]T[15_ cales, and the Test of program, and D$[2i_T]FI nternet access at home. Exclusion criteria were: Visual Perceptual Skill (D$F61T[6I]_ .8 points, 95% CI I$DF_71[]T2.8 to ID$T_8F1]1[ 0.8). The groups upper or lower limb surgery in the previous 6 months and unstable did not differ for other outcomes. Conclusion: A web-based epilepsy. Randomisation of 102 participants allocated 51 to the therapy program at home has the potential to enhance occupa- web-based therapy group and 51 to a waitlist control group. tional performance and visual perception for children with Interventions: The intervention group participated in a web-based cerebral palsy and to increase therapy dose.DFI[9_$T]1 therapy program comprising upper limb, cognitive and visual perceptual training (60%) and gross motor activities (40%). The Provenance: Invited. Not peer-reviewed. dose was 20 to 30 minutes, 6 days per week for 20 weeks (maximum 60 hours). For each child, therapists devised an Nora Shields individual program selected from 14 modules and adjusted it School of Allied Health, La Trobe University, Australia weekly. The control group received usual care. Outcome measures: The primary outcomes at baseline and after the intervention were: http://dx.doi.org/10.1016/j.jphys.2015.05.005 the Assessment of Motor and Process Skills, Assisting Hand Assessment, the Jebsen-Taylor Test of Hand Function and the Commentary The efficacy of high-intensity therapy programs in improving The efficacy of home programs for children with cerebral palsy upper limb function in children with unilateral spastic cerebral has been demonstrated3 and is a central component of service palsy has previously been established.1,2 However, these programs provision for this client population. Web-based programs afford can be labour-intensive and are not feasible for all families. The use another mode of service delivery for therapists that may, by their of interactive computer play to augment face-to-face therapy is tech-based nature, be of greater appeal to children and adolescents appealing as it may afford a low-cost means of intensive therapy than more traditional approaches to home program prescription. delivery, typically in the home setting. Web-based programs facilitate repetitive practice and can improve the quality of performance of activities of living, motor and This trial was well designed, conducted and reported. The processing skills. Clinicians should consider the intensity, duration observed differences between the intervention and control groups in and specificity of training when considering implementing such motor and processing skills, occupational performance and visual programs in practice. perception validates the proposal that web-based training permits repetitive practice of the perceptual, cognitive and motor control Provenance: Invited. Not peer-reviewed. components of a task. The absence of improvement in the Assisting Hand Assessment and the Melbourne Assessment of Unilateral Claire Kerr Upper Limb function scores for the impaired limb could be due to the School of Allied Health, Australian Catholic University, lack of specificity of the training: it was neither goal-directed nor task-specific. Alternatively, the program may not have been Australia sufficiently intense to achieve experience-dependent neuroplastic changes. Participants completed an average of 32 hours (of the References recommended 40 to 60 hours) over 20 weeks, with six participants completing  10 hours. These training hours were substantially less 1. Tinderholt Myrhaug H, et al. BMC Pediatr. 2014;14:292. and were distributed over a longer period than typically advocated 2. Sakzewski L, et al. Pediatr. 2014;133:e175–e204. for other efficacious ‘high-dose’ therapy programs.2 3. Novak I, et al. Pediatr. 2009;124:e606–e614. http://dx.doi.org/10.1016/j.jphys.2015.05.004 1836-9553/ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 61 (2015) 160 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers A web-based therapy program enhances occupational performance and visual perception in children with unilateral cerebral palsy Synopsis Summary of: James S, Ziviani J, Ware R, Boyd R. Randomized Melbourne Assessment of Unilateral Upper Limb Function. controlled trial of web-based multimodal therapy for unilateral Secondary outcome measures were: the Canadian Occupational cerebral palsy to improve occupational performance. Dev Med Child Performance Measure and the Test of Visual Perceptual Skill. Neuro. 2015; doi:10.1111/dmcn.12705. Results: Ninety-two participants completed the study. The intervention group completed 32 hours (95% CI 4 to 75 hours) of Question: Does a web-based therapy program improve the program. At the end of the intervention, the Assessment of occupational performance, upper limb function and visual Motor and Process Skills motor scores were significantly greater in perception in children with unilateral cerebral palsy? Design: the treatment group by 0.28 logits (95% CI 0.17 to 0.39). The Randomised controlled trial with concealed allocation and blinded Jebsen-Taylor Test of Hand Function for the dominant limb outcome assessment for some, but not all, outcome measures. improved significantly more in the intervention than the control Setting: Home environment in Australia. Participants: Children group (–DTFI43[_]$ .7 seconds, 95% CI –[4_D$7]FTI .4 to –2_[T]FD$5I .0), but not for the impaired with spastic-type unilateral cerebral palsy, Manual Ability limb (–]D$_6[TF2I 2.0 seconds, 95% CI –4_F]D$[I7T 4.8 to 0DT_]F[I$8 .7). The intervention group Classification System levels I to III and Gross Motor Function had higher scores on the Canadian Occupational Performance Classification System levels I or II. Children were aged 8 to 18 years, Measure performanceD]F[I1T_$ (I_]$9F[D1T .3 points, 95% CI $D1_[T00]FI .7 to 1TD$[IF1_] .9) and had sufficient understanding and co-operation to perform the satisfaction (21[TD$IF]1_ .5 points, 95% CI 1DIF[$03]T_ .4 to TD1]_4I[0F$ .8) FIDs$]T[15_ cales, and the Test of program, and D$[2i_T]FI nternet access at home. Exclusion criteria were: Visual Perceptual Skill (D$F61T[6I]_ .8 points, 95% CI I$DF_71[]T2.8 to ID$T_8F1]1[ 0.8). The groups upper or lower limb surgery in the previous 6 months and unstable did not differ for other outcomes. Conclusion: A web-based epilepsy. Randomisation of 102 participants allocated 51 to the therapy program at home has the potential to enhance occupa- web-based therapy group and 51 to a waitlist control group. tional performance and visual perception for children with Interventions: The intervention group participated in a web-based cerebral palsy and to increase therapy dose.DFI[9_$T]1 therapy program comprising upper limb, cognitive and visual perceptual training (60%) and gross motor activities (40%). The Provenance: Invited. Not peer-reviewed. dose was 20 to 30 minutes, 6 days per week for 20 weeks (maximum 60 hours). For each child, therapists devised an Nora Shields individual program selected from 14 modules and adjusted it School of Allied Health, La Trobe University, Australia weekly. The control group received usual care. Outcome measures: The primary outcomes at baseline and after the intervention were: http://dx.doi.org/10.1016/j.jphys.2015.05.005 the Assessment of Motor and Process Skills, Assisting Hand Assessment, the Jebsen-Taylor Test of Hand Function and the Commentary The efficacy of high-intensity therapy programs in improving The efficacy of home programs for children with cerebral palsy upper limb function in children with unilateral spastic cerebral has been demonstrated3 and is a central component of service palsy has previously been established.1,2 However, these programs provision for this client population. Web-based programs afford can be labour-intensive and are not feasible for all families. The use another mode of service delivery for therapists that may, by their of interactive computer play to augment face-to-face therapy is tech-based nature, be of greater appeal to children and adolescents appealing as it may afford a low-cost means of intensive therapy than more traditional approaches to home program prescription. delivery, typically in the home setting. Web-based programs facilitate repetitive practice and can improve the quality of performance of activities of living, motor and This trial was well designed, conducted and reported. The processing skills. Clinicians should consider the intensity, duration observed differences between the intervention and control groups in and specificity of training when considering implementing such motor and processing skills, occupational performance and visual programs in practice. perception validates the proposal that web-based training permits repetitive practice of the perceptual, cognitive and motor control Provenance: Invited. Not peer-reviewed. components of a task. The absence of improvement in the Assisting Hand Assessment and the Melbourne Assessment of Unilateral Claire Kerr Upper Limb function scores for the impaired limb could be due to the School of Allied Health, Australian Catholic University, lack of specificity of the training: it was neither goal-directed nor task-specific. Alternatively, the program may not have been Australia sufficiently intense to achieve experience-dependent neuroplastic changes. Participants completed an average of 32 hours (of the References recommended 40 to 60 hours) over 20 weeks, with six participants completing  10 hours. These training hours were substantially less 1. Tinderholt Myrhaug H, et al. BMC Pediatr. 2014;14:292. and were distributed over a longer period than typically advocated 2. Sakzewski L, et al. Pediatr. 2014;133:e175–e204. for other efficacious ‘high-dose’ therapy programs.2 3. Novak I, et al. Pediatr. 2009;124:e606–e614. http://dx.doi.org/10.1016/j.jphys.2015.05.004 1836-9553/ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 61 (2015) 164 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Clinical Practice Guidelines Congenital muscular torticollis Physical therapy management of congenital muscular torticollis: an evidence-based clinical practice guideline Date of latest update: December 2013. Date of next update: Description: These guidelines describe the evidence support- Not indicated. Patient group: Infants with congenital muscular ing the physiotherapy management of congenital muscular torticollis. Intended audience: The guidelines are primarily torticollis. They provide information on the latest evidence for: intended for paediatric physiotherapists, but could also be useful infant screening; examination (eg, active and passive range of for family members, other clinicians, educators and researchers. motion, posture, positioning, pain, palpation, asymmetries, Additional versions: There are no additional versions. Expert activity, developmental status and positioning during participa- working group: The expert working group comprised a three- tion activities such as feeding); evaluation; diagnosis; reasons for member committee of paediatric physical therapists from the USA. referral; prognosis; intervention (eg, passive and active range of Funded by: The Section on Pediatrics, American Physical Therapy motion, development of symmetrical movement patterns, Association. Consultation with: Invited stakeholders representing environmental adaptations, parental education); discharge; medicine, surgery, nursing, midwifery, physical therapy clinicians and long-term assessment of outcomes. The authors have and researchers, and a parent representative reviewed the developed 16 action statements or recommendations, which guidelines in three rounds. The guidelines were also posted online are centred on: identification and referral (eg, who should be for public consultation. Approved by: The Section on Pediatrics, monitored, treated or referred); physiotherapy examination; American Physical Therapy Association. Location: The guidelines intervention; and discharge/follow-up. The guidelines include a are published as: Kaplan S, Coulter C, Fetters L. Physical Therapy referral flow diagram, congenital muscular torticollis classifica- Management of Congenital Muscular Torticollis: An Evidence- tion grades, and a decision tree with which to assist clinicians. Based Clinical Practice Guideline: From the section on Pediatrics of The references for all recommendations are supplied at the end of the American Physical Therapy Association. Pediatric Physical the document. Therapy Winter 2013; Vol 25 (4) p 348-394. The guidelines are available for free at the Pediatric Physical Therapy website: http:// Sandra Brauer journals.lww.com/pedpt/pages/default.aspx The University of Queensland, Australia http://dx.doi.org/10.1016/j.jphys.2015.03.002 1836-9553/ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 61 (2015) 158 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers Early physical rehabilitation may improve physical quality of life domains in patients admitted to ICU with sepsis syndromes Synopsis Summary of: Kayambu G, Boots R, Paratz J. Early physical hospital discharge. Secondary outcome measures included: the rehabilitation in intensive care patients with sepsis syndromes: a Physical Function ICU Test; muscle strength measured using the pilot randomised controlled trial. Intensive Care Med. 2015; Medical Research Council muscle score; psychological outcome Published online F4DI8T$]_[ April 2015. measured using the anxiety subscale of the Hospital Anxiety and Depression Scale; pro-inflammatory and anti-inflammatory activity Question: Does early physical rehabilitation improve physical measured using cytokines. Results: Fifty participants completed the function in intensive care patients with sepsis syndromes? Design: study. At discharge from ICU, there was no between-group Randomised controlled trial with concealed allocation and blinding difference in the Acute Care Index of Function (mean difference of participants and outcome assessors. Setting: A quaternary FTI_6]6D$[ points, 95% CI –12 to 24IDF$_[1T)] . Six months following hospital discharge, hospital general intensive care unit (Royal Brisbane and Women’s compared with the standard care group, those in the intervention Hospital, Australia). Participants: Inclusion criteria were: aged  group reported better physical function (FIT[D$_72] 2, 95% CI T18IDF]_$[ to ID]FT_94[$ 3 points) 18 years, mechanically ventilated  48 hours and diagnosed with and physical role (]F1I4D[T0_$ 4, 95% CI DF1]$I[1T_ 5 to T_]7D[1IF$2 4 points) domains on the SF-36. sepsis, severe sepsis or septic shock. Exclusion criteria were: the Over the ICU stay, there was a greater change in interleukin-10 (in presence of head injury, burns, spinal injury, multiple fractured lower limbs, unresponsive to maximal treatment or expected to die favour of the intervention group) and tumour necrosis factor-a (in  48 hours. Randomisation of the 50 participants allocated 26 to an intervention group and 24 to a standard care group. Interventions: favour of the standard care group). There were no other between- The intervention group commenced individualised, targeted physi- group differences. Conclusion: Early physical rehabilitation im- cal rehabilitation within 48 hours of sepsis diagnosis. Rehabilitation proved physical quality of life domains in patients admitted to ICU comprised electrical muscle stimulation, active and passive range of with sepsis syndromes. motion, sitting out of bed and other mobilisation techniques, as appropriate, for 30 minutes, one to two times daily until discharge [95% CIs calculated by the CAP Editor.] from ICU. The standard care group received physical therapy strategies provided by the ICU physiotherapist. Outcome measures: Provenance: Invited. Not peer-reviewed. The primary outcomes were: physical function, measured using the Acute Care Index of Function at discharge from ICU; and quality of Elizabeth H Skinner life, measured using the Short Form-36 (SF-36) 6T[]F5DI$_ months following Physiotherapy, Western HealthD$T_F2[I], Melbourne, Australia Commentary http://dx.doi.org/10.1016/j.jphys.2015.05.002 Survivors of sepsis syndromes commonly have long-lasting Also, the multiple cytokine assessments and lack of an a priori functional impairments,1 with few proven treatments. Attempting analysis plan reduces confidence in a causal relationship between to improve these outcomes, Kayambu and colleagues performed a the intervention and either decreasing inflammation (interleukin- pilot randomised controlled trial of physical rehabilitation early in 6) or increasing anti-inflammation (interleukin-10). Neither the course of sepsis. The authors concluded that their intervention association was ‘significant’ after correction for multiple compar- ‘improve(s) self-reported physical function and induce(s) systemic isons. The finding of a greater decline in TNF-alpha in the control anti-inflammatory effects.’ group further confuses the interpretation. _D$W1]FI[T hile these authors are to be lauded for conducting an There is important work yet to be done to prove and achieve important trial, it is difficult to agree that this study supports 2]t$_FIDT[ heir benefit from early rehabilitation in critically ill patients, with few conclusions.F[3_TD$I] There are several reasons for this assertion.D$[]FI4_T RCTs demonstrating efficacy to date.2,3 Unfortunately, this new study does not offer the level of evidence needed to move the There is confusion about which of the many outcomes is needle and change current clinical practiceD._$][T5IF ‘primary’. The first primary outcome was a measure of function (the Acute Care Index of Function) assessed at discharge from ICU. The Provenance: Invited. Not peer-reviewed. second ‘primary’ outcome was the SF-36 by telephone at 6 months following hospital discharge. Among many secondary outcomes, Catherine L Hough three cytokines were measured at multiple times. With many Department of Medicine, University of Washington, Seattle, USA outcomes assessed, the chances of false-positive findings increase significantly. E-mail address: [email protected] (C.L. Hough). The primary outcome was negative – the Acute Care Index of References Function was equal among groups at discharge from ICU. The intervention group had high 6-month mortality (46% compared 1. Iwashyna TJ, et al. JAMA. 2010;304:1787–1794. with 17% among controls). The SF-36 was only assessed for 2. Schweickert WD, et al. Lancet. 2009;373:1874–1882. 11 subjects in the intervention group, compared with 19 controls. 3. Burtin C, et al. Crit Care Med. 2009;37:2499–2505. This introduced two additional threats (bias of differential follow- up and lack of statistical stability); therefore, we should not http://dx.doi.org/10.1016/j.jphys.2015.05.001 conclude that the intervention improved quality of life. 1836-9553/ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 61 (2015) 158 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers Early physical rehabilitation may improve physical quality of life domains in patients admitted to ICU with sepsis syndromes Synopsis Summary of: Kayambu G, Boots R, Paratz J. Early physical hospital discharge. Secondary outcome measures included: the rehabilitation in intensive care patients with sepsis syndromes: a Physical Function ICU Test; muscle strength measured using the pilot randomised controlled trial. Intensive Care Med. 2015; Medical Research Council muscle score; psychological outcome Published online F4DI8T$]_[ April 2015. measured using the anxiety subscale of the Hospital Anxiety and Depression Scale; pro-inflammatory and anti-inflammatory activity Question: Does early physical rehabilitation improve physical measured using cytokines. Results: Fifty participants completed the function in intensive care patients with sepsis syndromes? Design: study. At discharge from ICU, there was no between-group Randomised controlled trial with concealed allocation and blinding difference in the Acute Care Index of Function (mean difference of participants and outcome assessors. Setting: A quaternary FTI_6]6D$[ points, 95% CI –12 to 24IDF$_[1T)] . Six months following hospital discharge, hospital general intensive care unit (Royal Brisbane and Women’s compared with the standard care group, those in the intervention Hospital, Australia). Participants: Inclusion criteria were: aged  group reported better physical function (FIT[D$_72] 2, 95% CI T18IDF]_$[ to ID]FT_94[$ 3 points) 18 years, mechanically ventilated  48 hours and diagnosed with and physical role (]F1I4D[T0_$ 4, 95% CI DF1]$I[1T_ 5 to T_]7D[1IF$2 4 points) domains on the SF-36. sepsis, severe sepsis or septic shock. Exclusion criteria were: the Over the ICU stay, there was a greater change in interleukin-10 (in presence of head injury, burns, spinal injury, multiple fractured lower limbs, unresponsive to maximal treatment or expected to die favour of the intervention group) and tumour necrosis factor-a (in  48 hours. Randomisation of the 50 participants allocated 26 to an intervention group and 24 to a standard care group. Interventions: favour of the standard care group). There were no other between- The intervention group commenced individualised, targeted physi- group differences. Conclusion: Early physical rehabilitation im- cal rehabilitation within 48 hours of sepsis diagnosis. Rehabilitation proved physical quality of life domains in patients admitted to ICU comprised electrical muscle stimulation, active and passive range of with sepsis syndromes. motion, sitting out of bed and other mobilisation techniques, as appropriate, for 30 minutes, one to two times daily until discharge [95% CIs calculated by the CAP Editor.] from ICU. The standard care group received physical therapy strategies provided by the ICU physiotherapist. Outcome measures: Provenance: Invited. Not peer-reviewed. The primary outcomes were: physical function, measured using the Acute Care Index of Function at discharge from ICU; and quality of Elizabeth H Skinner life, measured using the Short Form-36 (SF-36) 6T[]F5DI$_ months following Physiotherapy, Western HealthD$T_F2[I], Melbourne, Australia Commentary http://dx.doi.org/10.1016/j.jphys.2015.05.002 Survivors of sepsis syndromes commonly have long-lasting Also, the multiple cytokine assessments and lack of an a priori functional impairments,1 with few proven treatments. Attempting analysis plan reduces confidence in a causal relationship between to improve these outcomes, Kayambu and colleagues performed a the intervention and either decreasing inflammation (interleukin- pilot randomised controlled trial of physical rehabilitation early in 6) or increasing anti-inflammation (interleukin-10). Neither the course of sepsis. The authors concluded that their intervention association was ‘significant’ after correction for multiple compar- ‘improve(s) self-reported physical function and induce(s) systemic isons. The finding of a greater decline in TNF-alpha in the control anti-inflammatory effects.’ group further confuses the interpretation. _D$W1]FI[T hile these authors are to be lauded for conducting an There is important work yet to be done to prove and achieve important trial, it is difficult to agree that this study supports 2]t$_FIDT[ heir benefit from early rehabilitation in critically ill patients, with few conclusions.F[3_TD$I] There are several reasons for this assertion.D$[]FI4_T RCTs demonstrating efficacy to date.2,3 Unfortunately, this new study does not offer the level of evidence needed to move the There is confusion about which of the many outcomes is needle and change current clinical practiceD._$][T5IF ‘primary’. The first primary outcome was a measure of function (the Acute Care Index of Function) assessed at discharge from ICU. The Provenance: Invited. Not peer-reviewed. second ‘primary’ outcome was the SF-36 by telephone at 6 months following hospital discharge. Among many secondary outcomes, Catherine L Hough three cytokines were measured at multiple times. With many Department of Medicine, University of Washington, Seattle, USA outcomes assessed, the chances of false-positive findings increase significantly. E-mail address: [email protected] (C.L. Hough). The primary outcome was negative – the Acute Care Index of References Function was equal among groups at discharge from ICU. The intervention group had high 6-month mortality (46% compared 1. Iwashyna TJ, et al. JAMA. 2010;304:1787–1794. with 17% among controls). The SF-36 was only assessed for 2. Schweickert WD, et al. Lancet. 2009;373:1874–1882. 11 subjects in the intervention group, compared with 19 controls. 3. Burtin C, et al. Crit Care Med. 2009;37:2499–2505. This introduced two additional threats (bias of differential follow- up and lack of statistical stability); therefore, we should not http://dx.doi.org/10.1016/j.jphys.2015.05.001 conclude that the intervention improved quality of life. 1836-9553/ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 61 (2015) 135–141 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Research Home exercises and supervised exercises are similarly effective for people with subacromial impingement: a randomised trial Fredrik Granviken a, Ottar Vasseljen b a Department of Physical Medicine and Rehabilitation, St. Olav’s University Hospital; b Department of Public Health and General Practice, Faculty of Medicine, Norwegian University of Science and Technology, Trondheim, Norway K]F[1I7_TD$ E Y W O R D S ABSTRACT Randomised controlled trial Question: Are there different effects of home exercises and supervised exercises on pain and disability Shoulder for people with subacromial impingement? Design: Randomised trial with two treatment arms, Subacromial impingement syndrome concealed allocation, blinded assessment of some outcomes, and intention-to-treat analysis. Exercise therapy Participants: Forty-six patients with subacromial impingement were recruited from an interdisciplin- Rehabilitation ary outpatient clinic of physical medicine and rehabilitation at a university hospital in Norway. Intervention: The home exercise group had one supervised exercise treatment followed by exercises at home for 6 weeks. The supervised exercise group had up to 10 supervised exercise treatments in addition to home exercises for 6 weeks. Outcome measures: The primary outcome was the Shoulder Pain and Disability Index (SPADI). Secondary outcome variables were: average pain during the past week, the Fear Avoidance Beliefs Questionnaire, participant satisfaction with treatment, active range of motion, work status and clinical shoulder tests. Pain was assessed weekly and all outcomes were assessed at 6 weeks. Participants were free to seek ongoing treatment of their choice until 26 weeks, when the SPADI was assessed again. Results: While both groups improved considerably, the groups did not differ significantly on the SPADI after the intervention at 6 weeks (0 points, 95% CI –14 to 14) or when followed up at 26 weeks (–2 points, 95% CI –21 to 17). There were no between-group differences for pain at any time. The remaining outcomes also did not differ significantly, except for the clinical tests of shoulder impingement. In the supervised exercise group, 11 out of 23 participants had two or more positive tests, compared to 18 out of 21 in the home exercise group. Conclusion: Supervision of more than the first session of a 6-week exercise regimen did not cause significant differences in pain and disability in people with subacromial impingement. Trial registration: NCT01257113. [Granviken F, Vasseljen O (2015) Home exercises and supervised exercises are similarly effective for people with subacromial impingement: a randomised trial. Journal of Physiotherapy 61: 135–141] ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Introduction Kuhn11 reported that exercise therapy had statistically and clinically significant effects on pain and disability, but supervised The shoulder is one of the most frequent sites of musculoskel- exercises were no better than home exercises. Walther and etal pain, exceeded only by back and knee pain.1 The incidence of colleagues12 compared standardised self-training, conventional shoulder pain in primary care patients is estimated to be 11.2 per physiotherapy and a functional brace, which all showed significant 1000 per year.2 The course varies, but a considerable number of reduction in pain levels and improvement in disability. However, people with shoulder pain (41%) show persistent symptoms after no differences among the three groups were found. Senbursa and 1 year.3 Many people with shoulder pain have signs of subacromial colleagues13 also included three groups: a supervised exercise impingement,2,4 which is characterised by pain and disability, group, a supervised exercise group combined with mobilisation, mainly in activities above shoulder height. Subacromial impinge- and a home-based rehabilitation group. All groups experienced ment is reported in 30 to 86% of shoulder pain patients in primary significant decreases in pain and increases in shoulder muscle care, 2,4,5 and 36% in secondary care.6 strength and disability, but no differences between groups were found. None of these studies had any form of blinding. The efficacy of physiotherapy is debated, and some passive treatments are not recommended.7,8 There is strong evidence that In the clinic, patients with subacromial impingement receive extracorporeal shock-wave therapy is ineffective and moderate guidance in different training principles. Guidance is believed to be evidence that ultrasound is ineffective for subacromial impinge- particularly important in the early rehabilitation phase where the ment.7 Brox and colleagues reported that surgical treatment and patients need help and support to deal with pain and dysfunction, supervised exercises were equally effective in the treatment of and to perform the exercises correctly. It remains unclear as to subacromial impingement.9,10 In a published systematic review, whether supervised exercises provide any additional benefit over http://dx.doi.org/10.1016/j.jphys.2015.05.014 1836-9553/ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/).

136 Granviken and Vasseljen: Home versus supervised exercises for subacromial impingement home-based exercises. Therefore, the main research question in serious mental illness. Earlier treatment, but no other treatment this study was: during the study period, was allowed. Are there different effects of home exercises and supervised Interventions exercises on pain and disability for people with subacromial impingement? Before any intervention, all participants took part in a theory lesson with other people with shoulder problems. The course was Method physiotherapist-led and focused on shoulder anatomy and the rehabilitation process. Design The home exercise group had one supervised treatment session In this randomised trial, people with subacromial impingement with a physiotherapist in order to set up a tailored home-exercise were randomised to home exercises or supervised exercises. They program. The supervised exercise group was offered 10 treatments received oral and written information about the study and of supervised exercise therapy, in addition to home exercises. informed consent was obtained before baseline measurements Exercises and overall training dose were the same for both groups. were taken. Allocation was concealed. The participants were The intervention period was 6 weeks. randomised via online access to the randomisation program at the Unit for Applied Clinical Research at Norwegian University of For both groups, established training principles were used.11,20 Science and Technology. Randomisation was stratified by gender The main goal for all exercises was to re-establish normal shoulder to obtain gender-balanced groups because symptoms and pain movement patterns through awareness, which the participants intensity may differ between women and men.14,15 Randomisation could transfer to daily activities. To normalise shoulder motion, a also used variable block sizes to assign participants to the two mirror was used at the start of the rehabilitation for visual treatment groups. Data were obtained before randomisation and at stimulation. All participants started with training of correct scapular the end of the 6-week intervention period by an examiner blinded placement. An example of this was to depress the shoulder during to the participants’ group assignment. The participants were shoulder flexion and abduction movements to avoid pulling the instructed not to discuss their treatment with the examiner who shoulder towards the ear and upward rotation of the scapulae. Focus performed the testing. Twenty-six weeks after randomisation, was on scapular stabilising exercises, rotator cuff exercises, and pain- participants were also assessed without blinding via a mailed free range of motion exercises. Exercises were individually adapted. questionnaire. Based on their symptoms, participants were free to choose whether they wanted to continue treatment, or not, During the training, a thin rubber band was used as a training between 6 and 26 weeks. tool for many of the exercises, either to reduce the arm load, control movement or provide resistance. The exercises were Participants, therapists and centres performed with as little pain as possible, and the choice of exercises, starting position and range of motion were decided with Participants were recruited from patients who had been this in mind. Participants used three sets of 30 repetitions for most referred for shoulder problems to the Interdisciplinary Outpatient exercises. For both groups the same exercises were performed at Clinic of Physical Medicine and Rehabilitation Department at St. home with four to six exercises twice a day every day. The home Olav’s Hospital, Norway, between January 2011 and August training group was also instructed in the progression opportunities 2012. As part of the standard procedures, both a doctor in physical for the appropriate exercises. medicine and an orthopaedic surgeon examined all referrals in order to determine further examination and treatment in the Based on individual needs, participants were later given physical medicine or orthopaedic department. Patients ineligible stretching exercises for tight structures in addition to the other for consideration for the study were surgery candidates with exercises. Stretches were held for 30 seconds and repeated twice fractures, full thickness ruptures/total ruptures, or prosthesis for each exercise. All participants were given written home candidates. A doctor in physical medicine examined all of the other exercises and they registered their training in a training diary. patients who were considered to be suitable for non-operative treatment at the outpatient clinic. From this pool, patients were Outcome measures screened for inclusion in the current study. Baseline data included age, gender, dominant arm, painful arm, To be eligible for the study, patients had to be between 18 and education, duration of symptoms, treatment during the last 3 years 65 years old and have unilateral shoulder pain lasting more than and work status. 12 weeks. Furthermore, they underwent three diagnostic clinical tests based on criteria in previous recommendations.16 The painful Primary outcome arc test17 was positive if pain was present in any parts of the The primary outcome was the Shoulder Pain and Disability motion path between 60 and 120 deg either on the way up or down during active abduction. A positive infraspinatus test18 was Index (SPADI).21 This is a self-reported questionnaire for people indicated by pain and/or weakness in isometric external rotation with shoulder pain. The SPADI contains 13 items that assess two against force performed with 90 deg of elbow flexion and the upper domains: a five-item subscale that measures pain and an eight- arm in neutral position along the side of the body. The Kennedy- item subscale that measures disability. Items are scored on a visual Hawkins test19 was positive if pain was experienced when the arm analogue scale. The total score ranges from 0 to 100 points, where was passively positioned at 90 deg of flexion and internally rotated 0 is no pain/disability and 100 is the worst pain/disability. The by the therapist. For a patient to be included in the study, all three questionnaire was scored as originally described21 and a version tests had to be positive. In addition, they had to have normal adapted to the Norwegian language and culture was used.22 passive glenohumeral physiological range of motion. Secondary outcomes Exclusion criteria were: glenohumeral instability, acromiocla- Secondary outcome variables were: average pain in the past vicular joint pathology, labrum pathology on imaging, proven full thickness ruptures/total ruptures of the rotator cuff, or signs of week, scored on a numerical rating scale; clinical tests (painful arc, glenohumeral osteoarthritis. Patients were also excluded if they infraspinatus and Kennedy-Hawkins tests); the Fear Avoidance had: undergone shoulder surgery, insufficient language capability, Beliefs Questionnaire (FABQ); active range of motion; work status; cervical spine problems (if the patient reported more pain in the and participant satisfaction. neck than the shoulder), rheumatoid arthritis, or other physical or The painful arc, infraspinatus and Kennedy-Hawkins tests are designed for diagnostic purposes, but the tests were repeated at 6 weeks to see if they had changed over the intervention period. Active range of motion was measured using a digital inclino- meter.a Maximum ranges for active flexion, abduction, external and

Research 137 internal rotation were obtained. The inclinometer was placed along these, 46 were found to be eligible and agreed to participate the forearm. Flexion and abduction range of motion were obtained in (Figure 1). The groups were well matched at baseline in terms of sitting with a straight elbow. Participants touched the wall with age, gender, duration of symptoms, dominant arm affected, their hand during the entire movement in abduction. Rotations were education, treatment in the last 3 years, sick leave, SPADI scores tested in supine position with a starting position of the arm abducted or secondary outcome measures (Table 1 and the first two columns to 90 deg, the elbow in 90 deg of flexion and the forearm pointing of Table 2). towards the ceiling. The participant moved the arm as far as possible, regardless of pain, for all movements. The movements were Compliance with the trial protocol performed three times for each direction and averaged values were used for the data analyses. Participants randomised to supervised exercise therapy had a median of [T_2D$8]FI (IQR 7 to 10) treatments. One participant had Fear avoidance may have the potential to negatively affect ]F[ID2$T23_ treatments, three participants had between 44[]_F2TD$I and DIF]$T_652[ treatments, outcomes for people with musculoskeletal disorders. To quantify and 19 participants had 62TD$I_F]7[ or more treatments. fear avoidance in the participants, a modified version of the original FABQ was used,23 where the word ‘back’ was replaced with Participants in the home exercise group completed 88% of the ‘shoulder’.24 The questionnaire consists of 16 items, and each item total planned exercise sessions and the supervised group is scored on a seven-point Likert scale, where 0 is strongly disagree completed 80%. There was a median of 74 (IQR 58 to 81) workouts and 6 is strongly agree. The first five questions are related to for the home exercise group and a median of 67 (IQR 56 to 83) physical activity, the next ]F1D8_1T[$I 1 questions are related to work. workouts for the supervised group during the 6-week intervention Questions two, three, four and five are used for summing physical period. Two participants in the home exercise group dropped out activity score and questions six, seven, nine, ten, eleven, twelve and for unknown reasons during the intervention period. They did not fifteen are used for the work score. Higher scores represent higher differ from the other participants in baseline scores. fear of movement. Scores range from 0 to 24 for physical activity and from 0 to 42 for work. Self-reported work status (working, One other secondary outcome – the quality of life question- sick-listed, other) was also obtained at 6 and 26 weeks follow-up. naire, SF-36 – was registered but not reported because when it later became apparent that the questionnaire comes with a licence After 6 weeks, the participants reported how satisfied they were fee it was not included. There was no budget for this questionnaire. with the treatment. This was measured with two separate scales. First, perceived benefit of the treatment was rated as one of seven Effect of the interventions possibilities: completely recovered, much improved, slightly im- proved, no change, slightly worsened, much worsened1[]T9,$_FDI and worse Group mean outcome scores and between-group differences than ever. Second, satisfaction with treatment was rated as one of are given in Table 2 and Figure 2. Individual participant data are five possibilities: satisfied, somewhat satisfied, mixed (neither presented in Table 3, which is available on the eAddenda. There satisfied nor dissatisfied), somewhat dissatisfiedD[$]20IFT_, and dissatisfied.25 were no significant differences between home exercise and supervised exercise on the SPADI at 6 weeks (MD 0 points, 95% Participants recorded all training in a training diary, and once a CI –14 to 14) or at 26 weeks follow-up (MD –2 points, 95% CI –21 to week they also registered their average pain level during the past 17). There were no significant between-group differences for pain week on a numerical rating scale from 0 (no pain) to 10 (worst at any time (Figure 3), the FABQ physical activity, the FABQ work, possible pain).26 or active range of motion (Table 2). Data analysis The within-group improvement for pain and disability after the intervention was 3DI_72TF[]$ 0 to 40% in both treatment arms. A greater A change of 20 points on the Shoulder Pain and Disability Index improvement was observed for the FABQ work than physical has been defined as the minimum clinically important change.27 activity, while there were small changes for active range of motion. This study was thus designed to detect a between-group difference At the end of the 6-week intervention period, 18/21 in the home of 20 points on the SPADI as statistically significant, with alpha of exercise group still had at least two positive clinical tests for 0.05 and power of 0.80. The standard deviation was set to 20, shoulder impingement compared to 11/23 in the supervised according to a previous study.28 This resulted in a sample size of exercise group (RR 0.55; 95% CI 0.35 to 0.88). This statistically 17 in each group, using standard software.b In order to account for significant difference means thatID[F2_T$] for every DT2F][I_38$ (95% CI 2 to 10) dropouts during the study, the sample size was increasedF[1_TD$]I to a total patients whose exercise regimen is supervised, one who would of 23 participants in each group. otherwise have had two or more positive clinical signs if they had not received the supervision will have one or zero positive clinical Baseline data were assessed for normal distribution. Outcome signs after 6 weeks. variables were analysed with linear mixed-effects models with random slope (time), and the group*time interaction term was Chi-squared tests showed no significant between-group included for comparative analysis of group effects over time. differences for perceived benefit or satisfaction with treatment Estimates of marginal group effects for the primary outcome (SPADI) at 6 weeks. For perceived benefit, none of the participants and pain (numerical rating scale) were adjusted for age and gender. reported complete recovery. In the home versus supervised For the remaining outcome variables, age, gender and baseline pain exercise groups, 29I]FD_[$T2 4 versus 52% reported being much improved, level were adjusted for. Post hoc pairwise comparisons of group 5]FID$T_03[ 7 versus 30% reported being slightly improved, I]13DF_$1T[ 9 versus 9% mean values were performed using the pwcompare command in reported no change, and none reported being much worse. In the Stata,c with the Bonferroni method to adjust for multiple compar- supervised group, one participant reported being slightly worse isons. The precision of the estimates was assessed with 95% CI. The and one reported being worse than ever after the intervention. For Fisher’s exact test was used for the clinical tests and work status. The satisfaction with treatment in the home versus supervised 2D1[_$IF]cT hi-squared test was used for the satisfaction with treatment. Data exercise groups, F]T32$_ID[52 versus 83% reported being satisfied, were analysed according to the intention-to-treat principle. I2]F[3_TD$ 9 versus 4% reported being somewhat satisfied, 1F[D]3_$IT1 9 versus 9% reported being neither satisfied nor dissatisfied, and none Results reported being somewhat dissatisfied, respectively. One patient from the supervised exercise group reported being dissatisfied Flow of participants, therapists and centres through the study with the treatment. A total of 509 patients were assessed, with further examination Among the participants for whom work status was available at for entrance into the physical medicine outpatient programs. Of 6 weeks, D$ITF[]73_4 /21 in the home exercise group and 10/23 in the supervised exercise group were on sick leave. At 26 weeks, FI]$DT4_5[3 /18 in the home exercise group and 3$_DI6FT[3] /21 in the supervised exercise group

[(Figure_1)TD]GI$138 Granviken and Vasseljen: Home versus supervised exercises for subacromial impingement Patients referred from general practitioners and found to be ineligible for surgery between January 2011 and August 2012 Assessed for shoulder impingement (n = 509) Excluded (n = 463) • ineligible (n = 446) • declined to participate (n = 2) • other reasons (n = 15) Week 0 Measured SPADI, pain, FABQ, range of motion, clinical shoulder tests, work status Randomised (n = 46) (n = 23) (n = 23) Lost to follow-up Home exercise Supervised exercise Lost to follow-up (n = 2) group group (n = 0) • uncontactable • 1 supervised • 10 supervised (n = 2) treatment treatments • home exercises • home exercises Measured SPADI, pain, FABQ, range of motion, clinical shoulder tests, work status, perceived benefit and participant satisfaction Week 6 (n = 21) (n = 23) Lost to follow-up Participants in both groups were free to seek Lost to follow-up (n = 3) further treatment during this period (n = 2) • uncontactable • uncontactable (n = 3) (n = 2) Measured SPADI and work status Week 26 (n = 18) (n = 21) Figure 1. Design and flow of participants through the trial. FABQ = Fear Avoidance Beliefs Questionnaire, SPADI = Shoulder Pain and Disability Index. were on sick leave. In the home exercise group, one participant 26-week follow-up questionnaire. Between 6 and 26 weeks, one reported receiving a disability pension and one reported having participant in the home exercise group and two in the supervised retired at both 6 and 26 weeks. One participant reported being group received surgery. Seventeen participants in the home unemployed in the supervised exercise group at 26 weeks. Fisher’s exercise group received a mean of 4.4 (SD 2.0) additional exact tests showed no significant differences between groups for treatment sessions during this period, while 15 participants work status at 6 or 26 weeks. received a mean of 3.3 (SD 1.6) additional treatments in the supervised group. Three participants, in addition to the two dropouts in the home exercise group and two in the supervised group, did not return the Discussion Table 1 Home Supervised In this comparative study of home exercises and supervised Baseline characteristics of participants. exercises exercises exercises for shoulder impingement, no differences were found in the primary outcome, the SPADI. Furthermore, no differences were Characteristic (n = 23) (n = 23) found in the secondary outcomes of pain, the FABQ (physical activity and work), participant satisfaction or active range of Age (yr), mean (SD) 48.2 (9.8) 47.6 (10.0) motion after the intervention period. A significant difference was Gender, n female (%) 11 (48) 11 (48) found in favour of supervised exercise in reduced positive clinical Duration of symptoms (mth), median (IQR) tests for shoulder impingement at 6 weeks, where 18/21 in the Dominant arm affected, n (%) 12 (6 to 36) 17 (10 to 48) home exercise group still had two or more positive clinical tests 12 yr of school or less, n (%) 15 (65) 13 (57) compared to only 11/23 in the supervised exercise group. Although Treatment for symptoms in past 3 yr, n (%) 13 (57) 12 (52) this finding indicates that supervised exercises can reduce pain in Exercise treatment, n (%) 18 (78) 19 (83) specific testing positions of the arm, this does not carry over into Cortisone injection, n (%) 10 (44) 10 (44) Sick leave, n (%) 11 (48) 5 (22) 10 (44) 9 (39)

Research 139 Between-group differencesa Week 26 minus HE minus SE –2 (–21 to 17) any benefits in reported shoulder function, average pain over the Week 0 past week, fear avoidance, range of motion or satisfaction. Therefore, physiotherapists should be reluctant to interpret this Week 6 minus HE minus SE 0 (–14 to 14) reduction in the number of positive diagnostic tests as an Week 0 4]_$FTD[I–0.1 ( ]I$[_DTF5–1.8 to $6T]FID[_1.6) important clinical benefit of using supervision with the exercise 2.8 (–T7_]$DFI[1.0 to [IF]$DT_86.5) regimen. 0.0 (–9$DIF][_T7.0 to T$1]0[_FID6.9) Strengths of this study include: a randomised design, concealed 0 (–]$F1[_IDT16 to []I$DT_1F16) allocation, blinded assessment at baseline and 6 weeks, few drop- –14 (–I$[1]F2DT_43 to [31_T]FID$15) outs and an intention-to-treat analysis. There were no between- group differences in the overall number of training sessions. The 2 (–]F[13$_TDI14 to IT_4$]F1[D18) fact that both groups had high (> 80%) and similar exercise 0 (–]FIT5$D_[110 to I16F_[]T$D11) adherence strengthens the results of this study. Week 26 minus HE (n = 23) SE (n = 23) HE (n = 21) SE (n = 23) HE (n = 18) SE (n = 21) HE (n = 21) SE (n = 23) HE (n = 18) SE (n = 21) –26 (28) Little is known about natural recovery in people with impinge- Week 0 ment. Superior effects have been reported for both surgery and Within-group differences –27 (26) FABQ = Fear Avoidance Beliefs Questionnaire, HE = home exercises, SE = supervised exercises, SPADI = Shoulder Pain and Disability Index. Shaded cells = primary outcome. supervised exercises compared to placebo laser at 6 months and a Mixed-effects models for SPADI and pain were adjusted for age and gender. The remaining outcome variable analyses were adjusted for age, gender and baseline pain level. 2.5 years follow-up;9,10 this was the rationale for not including a Table 2 Week 6 minus –15 (17) placebo group in this study. Another study with two active Mean (SD) for continuous outcomes at each assessment time for each group, mean (SD) difference within groups, and mean (95% CI) difference between groups. Week 0 –1.8 (1.9) interventions reported higher improvement with exercise (40 to –1.7 (4.8) 50%) than with shockwave therapy (20 to 30%) after 6 weeks.28 The –3.1 (7.8) present study found within-group improvements of 30 to 40% for pain and disability after the intervention in both treatment arms – 6 (13) presumably the combined result of exercise and natural recovery. A 12 (23) group with no or sham treatment was not included and, thus, the 4 (19) natural recovery or placebo effects in this study cannot be assessed. 0 (12) A large number of patients were screened for enrolment (n = 509) –17 (15) but only 46 were randomised for participation. The main reason for –2.1 (2.0) this was that there was no pre-selection of patients before the doctor –3.5 (4.1) visit (ie, they were referred with various shoulder symptoms from –3.2 (5.5) general practitioners). Therefore, all patients were screened as part of the standard hospital routine and considered as potential 4 (17) participants. Many did not fit the criteria for study participation. 11 (23) The stringent selection criteria are the main explanation for the low –2 (16) proportion of selected participants relative to patients available for 2 (13) enrolment. Shoulder impingement diagnosed by less stringent selection criteria may have given other results. This affects the Week 26 21 (18) external validity of the study and caution should be shown in generalising the results to all people with shoulder impingement. 24 (24) Local anaesthetics or imaging were not used to verify the diagnosis. However, subacromial impingement is a clinical diagnosis and a Groups Week 6 32 (20) recommended combination of clinical tests was used to confirm 4.1 (2.1) impingement in the participants.16 12.8 (5.8) 16.2 (13.1) It may be argued that 6 weeks is too short an intervention 156 (23) period to detect an effect of supervision. However, 6 weeks was 121 (42) chosen because it was believed that participants randomised to 67 (23) home exercise would not be motivated for a longer intervention on 65 (12) their own. Also, the most improvement was expected within the first few weeks. Engebretsen and colleagues28 studied the effect of 32 (15) supervised exercises in people with shoulder impingement and 4.3 (2.2) found that the largest improvement was within 6 weeks. The 10.6 (5.3) baseline symptom level for the present participants was similar to 17.4 (7.6) that of Engebretsen and colleagues.28 Another study of supervised 154 (23) exercises also foundFD$I[]_3T thatIF]TD$_73[ the largest improvement was within the 128 (42) first 6 weeks, and the authors stated that this time period might be 64 (22) sufficient to detect clinical improvement.29 In the present study, 63 (14) participants with small effects on the SPADI during the first 6 weeks also showed little improvement at 26-weeks follow-up. Week 0 48 (19) 5.9 (2.2) After the 6-week intervention period, the participants in either 14.4 (5.0) group who did not receive full recovery were free to continue the 19.3 (12.4) exercises, with some supervision at the clinic. Consequently, the 151 (26) effects at 26 weeks cannot be ascribed to the intervention alone, 109 (42) since the majority in both groups sought treatment in the period 63 (23) from 6 to 26 weeks. 65 (15) These results support previous research in the area, with no 49 (12) differences between home exercises and supervised exercise in 6.3 (1.3) groups for subacromial impingement.12,13 This study differed from 14.0 (4.0) the other comparable studies in some important aspects. An 20.6 (7.1) independent blinded assessor was used, where neither Walther 150 (21) and colleagues12 nor Senbursa and colleagues13 had any form of 119 (39) blinding. The present study design was also prospectively 65 (20) registered. 62 (15) Outcomes SPADI (0 to 100), mean (SD) Pain (0 to 10), mean (SD) FABQ physical activity (0 to 24), mean (SD) FABQ work (0 to 42), mean (SD) Flexion (deg), mean (SD) Abduction (deg), mean (SD) External rotation (deg), mean (SD) Internal rotation (deg), mean (SD)

SPADI (0 to 100)r_e][GIu(F$DiT)g2140 Granviken and Vasseljen: Home versus supervised exercises for subacromial impingement Average pain in previous week (0 to 10) 60 Home exercise 50 Supervised exercise 40 30 20 10 0 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 Time (weeks) Figure 2. Mean (95% CI) Shoulder Pain and Disability Index (SPADI) scores for the two groups at baseline, 6 weeks and 26 weeks. GIF$DT)3_erugi([] 8 7 Home exercise 6 Supervised exercise 5 4 3 2 1 0 0123456 Time (weeks) Figure 3. Mean (95% CI) scores for average pain over the past week on a 0-to-10 numerical rating scale for the two groups weekly from baseline to 6 weeks. The emphasis in this study was on contrasting the groups _TD$FI83]i[ n the What is already known on this topic: Subacromial impinge- amount of therapist guidance and attention (supervision) rather ment is a common cause of shoulder pain. Exercise improves than on differences in the content or dosage of exercises. Others pain, disability and range of movement. Previous trials did not have also reported the lack of effect of supervision. Andersen and identify a substantial benefit from supervision of the exercise, colleagues,30 investigating supervised exercise relative to home but limitations in the design and quality of these trials mean exercise after subacromial decompression, found no difference that the effect of supervision remains unclear. between groups. The amount of supervision was similar to the What this study adds: People with subacromial impinge- present study. Supervision of exercises for shoulder impingement ment syndrome obtain similar improvements in pain, disability beyond a single session, with or without surgery, may thus be and range of movement after a 6-week exercise regimen, questioned. whether regular supervision is maintained or only the first session is supervised. It cannot be disregarded that certain patient subgroups may experience greater benefit from supervision than others. In post hoc Footnotes: aAcumar, Lafayette Instrument Company, Lafayette, analyses, it was observed that those in the supervised group with USA. bMinitab 15, Minitab Inc, State College, Pennsylvania, USA. high baseline scores on the SPADI (ie, above the mean score of 49) cStata v12, Stata Corp, College Station, TX, USA. had considerably larger improvement in pain and disability after the intervention than those with similarly high baseline scores in the eAddenda items: Table 3 can be found online at doi:10.1016/j. home exercise group. Subgroups with higher symptoms levels jphys.2015.05.014. should be explored in more detail in future studies. Ethics approval: Central Regional Ethics Committee of Norway In this comparative study, no differences were found between approved this study. All participants gave written informed home exercises and supervised exercises on pain and disability for consent before data collection began. people with subacromial impingement. The results question whether extending supervision of exercises beyond an initial Competing interests: None. session is necessary for all people with subacromial impingement, Source(s) of support: The study was funded by St. Olav’s as some may have similar effects of home exercises and supervised University Hospital, Department of Physical Medicine and exercises when the training dose is the same. Rehabilitation and St. Olav’s University Hospital project funds.

Research 141 Acknowledgements: We are grateful to PT Ingunn Kregnes and 12. Walther M, Werner A, Stahlschmidt T, Woelfel R, Gohlke F. The subacromial PT Anders Bakken for their help with measurements, Dr Ulrich impingement syndrome of the shoulder treated by conventional physiotherapy, Schattel and colleagues for recruiting participants and to St. Olav’s self-training, and a shoulder brace: results of a prospective, randomized study. J University Hospital, Department of Physical Medicine and Shoulder Elbow Surg. 2004;13:417–423. Rehabilitation, The Outpatient Clinic for back-neck-shoulder for help and support. 13. Senbursa G, Baltaci G, Atay OA. The effectiveness of manual therapy in supraspi- natus tendinopathy. Acta Orthop Traumatol Turc. 2011;45:162–167. Provenance: Not invited. Peer-reviewed. Correspondence: Fredrik Granviken, Department of Physical 14. Unruh AM. Gender variations in clinical pain experience. Pain. 1996;65:123–167. Medicine and Rehabilitation, St. Olav’s University Hospital, 15. Zheng X, Simpson JA, van der Windt DA, Elliott AM. Data from a study of Trondheim, Norway. Email: [email protected] effectiveness suggested potential prognostic factors related to the patterns of References shoulder pain. J Clin Epidemiol. 2005;5:823–830. 16. Park HB, Yokota A, Gill HS, El Rassi G, McFarland EG. Diagnostic accuracy of clinical 1. Urwin M, Symmons D, Allison T, Brammah T, Busby H, Roxby M, et al. Estimating tests for the different degrees of subacromial impingement syndrome. J Bone Joint the burden of musculoskeletal disorders in the community: the comparative Surg Am. 2005;87:1446–1455. prevalence of symptoms at different anatomical sites, and the relation to social 17. Kessel L, Watson M. The painful arc syndrome. Clinical classification as a guide to deprivation. Ann Rheum Dis. 1998;57:649–655. management. J Bone Joint Surg Br. 1977;59:166–172. 18. Leroux JL, Thomas E, Bonnel F, Blotman F. Diagnostic value of clinical tests for 2. van der Windt DA, Koes BW, de Jong BA, Bouter LM. Shoulder disorders in general shoulder impingement syndrome. Rev Rhum Engl Ed. 1995;62:423–428. practice: incidence, patient characteristics, and management. Ann Rheum Dis. 19. Hawkins RJ, Kennedy JC. Impingement syndrome in athletes. Am J Sports Med. 1995;54:959–964. 1980;8:151–158. 20. Bohmer A, Staff PH, Brox JI. Supervised exercises in relation to rotator cuff disease 3. van der Windt DA, Koes BW, Boeke AJ, Deville W, De Jong BA, Bouter LM. Shoulder (impingement syndrome, stages 1 and 2): a treatment regimen and its rationale. disorders in general practice: prognostic indicators of outcome. Br J Gen Pract. Physiother Theory Pract. 1998;14:93–105. 1996;46:519–523. 21. Roach KE, Budiman-Mak E, Songsiridej N, Lertratanakul Y. Development of a shoulder pain and disability index. Arthritis Care Res. 1991;4:143–149. 4. Ostor AJ, Richards CA, Prevost AT, Speed CA, Hazleman BL. Diagnosis and relation to 22. Juel NG. Norsk fysikalsk medisin. Oslo: Fagbokforlaget; 2007. general health of shoulder disorders presenting to primary care. Rheumatol. 23. Waddell G, Newton M, Henderson I, Somerville D, Main CJ. A Fear-Avoidance 2005;44:800–805. Beliefs Questionnaire (FABQ) and the role of fear-avoidance beliefs in chronic low back pain and disability. Pain. 1993;52:157–168. 5. Walker-Bone K, Palmer KT, Reading I, Coggon D, Cooper C. Prevalence and impact of 24. Mintken PE, Cleland JA, Whitman JM, George SZ. Psychometric properties of the musculoskeletal disorders of the upper limb in the general population. Arthritis Fear-Avoidance Beliefs Questionnaire and Tampa Scale of Kinesiophobia in Rheum. 2004;51:642–651. patients with shoulder pain. Arch Phys Med Rehabil. 2010;91:1128–1136. 25. Ostelo RW, de Vet HC. Clinically important outcomes in low back pain. Best Pract 6. Juel NG, Natvig B. Shoulder diagnoses in secondary care, a one year cohort. BMC Res Clin Rheumatol. 2005;19:593–607. Musculoskelet Disord. 2014;15:89. 26. Von Korff M, Jensen MP, Karoly P. Assessing global pain severity by self-report in clinical and health services research. Spine (Phila Pa 1976). 2000;25:3140– 7. Faber E, Kuiper JI, Burdorf A, Miedema HS, Verhaar JA. Treatment of impingement 3151. syndrome: a systematic review of the effects on functional limitations and return 27. Ekeberg OM, Bautz-Holter E, Keller A, Tveita EK, Juel NG, Brox JI. A questionnaire to work. J Occup Rehabil. 2006;16:7–25. found disease-specific WORC index is not more responsive than SPADI and OSS in rotator cuff disease. J Clin Epidemiol. 2010;63:575–584. 8. Green S, Buchbinder R, Hetrick S. Physiotherapy interventions for shoulder pain. 28. Engebretsen K, Grotle M, Bautz-Holter E, Sandvik L, Juel NG, Ekeberg OM, et al. Cochrane Database Syst Rev. 2003;(2):CD004258. Radial extracorporeal shockwave treatment compared with supervised exercises in patients with subacromial pain syndrome: single blind randomised study. BMJ. 9. Brox JI, Staff PH, Ljunggren AE, Brevik JI. Arthroscopic surgery compared with 2009;339:b3360. supervised exercises in patients with rotator cuff disease (stage II impingement 29. Maenhout AG, Mahieu NN, De Muynck M, De Wilde LF, Cools AM. Does adding syndrome). BMJ. 1993;307:899–903. heavy load eccentric training to rehabilitation of patients with unilateral sub- acromial impingement result in better outcome? A randomized, clinical trial. Knee 10. Brox JI, Gjengedal E, Uppheim G, Bøhmer AS, Brevik JI, Ljunggren AE, et al. Surg Sports Traumatol Arthrosc. 2013;21:1158–1167. Arthroscopic surgery versus supervised exercises in patients with rotator cuff 30. Andersen NH, Sojbjerg JO, Johannsen HV, Sneppen O. Self-training versus physio- disease (stage II impingement syndrome): a prospective, randomized, controlled therapist-supervised rehabilitation of the shoulder in patients treated with ar- study in 125 patients with a 2 1/2-year follow-up. J Shoulder Elbow Surg. throscopic subacromial decompression: a clinical randomized study. J Shoulder 1999;8:102–111. Elbow Surg. 1999;8:99–101. 11. Kuhn JE. Exercise in the treatment of rotator cuff impingement: a systematic review and a synthesized evidence-based rehabilitation protocol. J Shoulder Elbow Surg. 2009;18:138–160.

Journal of Physiotherapy 61 (2015) 125–134 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Research Inspiratory muscle training facilitates weaning from mechanical ventilation among patients in the intensive care unit: a systematic review Mark Elkins a, Ruth Dentice b a Sydney Medical School, University of Sydney; b Physiotherapy Department, Royal Prince Alfred Hospital, Sydney, Australia KEY WORDS ABSTRACT Systematic review Question: Does inspiratory muscle training improve inspiratory muscle strength in adults receiving Respiratory muscle training mechanical ventilation? Does it improve the duration or success of weaning? Does it affect length of stay, Mechanical ventilators reintubation, tracheostomy, survival, or the need for post-extubation non-invasive ventilation? Is it Weaning tolerable and does it cause adverse events? Design: Systematic review of randomised trials. Intensive care Participants: Adults receiving mechanical ventilation. Intervention: Inspiratory muscle training Physiotherapy versus sham or no inspiratory muscle training. Outcome measures: Data were extracted regarding: inspiratory muscle strength and endurance; the rapid shallow breathing index; weaning success and duration; duration of mechanical ventilation; reintubation; tracheostomy; length of stay; use of non- invasive ventilation after extubation; survival; readmission; tolerability and adverse events. Results: Ten studies involving 394 participants were included. Heterogeneity within some meta-analyses was high. Random-effects meta-analyses showed that the training significantly improved maximal inspiratory pressure (MD 7 cmH2O, 95% CI 5 to 9), the rapid shallow breathing index (MD 15 breaths/min/l, 95% CI 8 to 23) and weaning success (RR 1.34, 95% CI 1.02 to 1.76). Although only assessed in individual studies, significant benefits were also reported for the time spent on non-invasive ventilation after weaning (MD 16 hours, 95% CI 13 to 18), length of stay in the intensive care unit (MD 4.5 days, 95% CI 3.6 to 5.4) and length of stay in hospital (MD 4.4 days, 95% CI 3.4 to 5.5). Weaning duration decreased in the subgroup of patients with known weaning difficulty. The other outcomes weren’t significantly affected or weren’t measured. Conclusion: Inspiratory muscle training for selected patients in the intensive care unit facilitates weaning, with potential reductions in length of stay and the duration of non-invasive ventilatory support after extubation. The heterogeneity among the results suggests that the effects of inspiratory muscle training may vary; this perhaps depends on factors such as the components of usual care or the patient’s characteristics. [Elkins M, Dentice R (2015) Inspiratory muscle training facilitates weaning from mechanical ventilation among patients in the intensive care unit: a systematic review. Journal of Physiotherapy 61: 125–134] ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Introduction difficulty is associated with increased risk of further respiratory Critical illness that requires management in an intensive care muscle weakening, nosocomial infection, airway trauma, and unit (ICU) is common, with between 221 and 595 admissions per death.11–15 Among those who are discharged, inspiratory muscle 100 000 population in developed countries in Europe, North dysfunction is a key prognostic factor for readmission to the ICU.16 America and Australasia.1–3 Critically ill patients who undergo However, reduced inspiratory muscle strength and endurance prolonged mechanical ventilation commonly acquire weakness have been observed in weaned patients,9,17 suggesting that while in the ICU.4–6 This ICU-acquired weakness is typically a respiratory muscle strength and endurance may only need to be mixed polyneuropathy and myopathy that is characterised by symmetrical weakness and deconditioning of the proximal improved to a certain threshold to permit weaning. musculature of the limbs and the respiratory muscles.7,8 Recently, lighter sedation protocols have been investigated as a A key contributor to the weakening and deconditioning of a means of allowing some diaphragmatic activation to persist during patient’s respiratory muscles whilst in the ICU is the replacement the period of mechanical ventilation.18 However, even in patients or support of spontaneous breathing by mechanical ventilation.9,10 The diaphragm, for example, responds to mechanical ventilation who can be managed in this way, mechanical ventilation still with rapid atrophy and changes in myofibre length.10 unloads the diaphragm to some extent so ventilator-induced diaphragmatic dysfunction may still occur.10 Respiratory muscle weakness and deconditioning are risk factors for difficulty in weaning from mechanical ventilation.10,11 Inspiratory muscle training applies a load to the diaphragm and In addition to prolonging the length of stay in the ICU, weaning accessory inspiratory muscles to increase their strength and endurance. Trials of inspiratory muscle training in the ICU have typically applied this load via devices that impose resistive or threshold loads, or via adjustment of the ventilator sensitivity so http://dx.doi.org/10.1016/j.jphys.2015.05.016 1836-9553/ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/).

126 Elkins and Dentice: Inspiratory muscle training in intensive care that patients can only initiate inspiratory flow by generating more Box 1. Inclusion criteria. negative intrathoracic pressure.19 Design A previous systematic review of randomised trials of inspirato-  Randomised trial ry muscle training in mechanically ventilated, critically ill patients identified few eligible trials.20 That review estimated that Participants inspiratory muscle training significantly increases inspiratory  Age > 16 years muscle strength, by a mean of 8 cmH2O (95% CI 6 to 9). However,  Intubated or tracheostomised no statistically significant effects were identified for the remaining  Receiving mechanical ventilation in ICU outcomes, which included measures of weaning duration and success, measures of length of stay, and survival. Although the Intervention results on these other outcomes were not statistically significant,  Inspiratory muscle training via any of the following: many of the 95% CIs excluded substantially harmful effects and included beneficial effects that would be considered clinically  isocapnic/normocapnic hyperpnoea worthwhile. Therefore, the authors of this review recommended  inspiratory resistive training further trials, with the expectation that a subsequent review would  threshold pressure training be able to provide more precise estimates of the effect of  adjustment of ventilator pressure trigger sensitivity inspiratory muscle training on these important clinical outcomes. Outcome measures A search of GoogleScholar (www.scholar.google.com) for  Inspiratory muscle strength papers that have cited the previous systematic review identified  Rapid shallow breathing index two new randomised trials that would be eligible for an update of  Weaning duration the review. A search of the Physiotherapy Evidence Database  Weaning success (www.PEDro.org.au) identified only one systematic review that  Duration of mechanical ventilation had been published on this topic in the interim,21 which did not  Reintubation include any new trials. Therefore, we undertook to systematically  Tracheostomy review the evidence about inspiratory muscle training in ICU,  Length of stay aiming to answer the following specific research questions:  Non-invasive ventilation  Survival 1. Does inspiratory muscle training improve inspiratory muscle  Tolerability strength in adults receiving mechanical ventilation?  Adverse events 2. Does it improve the duration or success of weaning? Comparison 3. Does it reduce length of stay, reintubation, tracheostomy,  Inspiratory muscle training versus sham/no training survival, readmission, or the need for post-extubation non- score out of 10 for overall methodological quality.23 The PEDro invasive ventilation? scale has been extensively validated and used.22–26 The authors 4. Is it tolerable and does it cause adverse events? had previously undertaken formal training in the use of the PEDro scale. Studies were not excluded on the basis of quality. Authors of Method the original papers were contacted to clarify any unclear criteria; therefore, scores may differ from those on the PEDro website, Identification and selection of studies which are based on the published paper only. The electronic databases PEDro, PubMed, CENTRAL, EMBASE Participants and CINAHL were searched from the earliest available records until Studies were eligible for inclusion if they involved people aged January 2015. The full search strategy is presented in Appendix 1 on the eAddenda. It incorporated the following search terms (or over 16 years who were intubated or tracheostomised and were synonyms): randomised controlled trial, inspiratory/respiratory/ receiving mechanical ventilation in an ICU. Weaning from ventilatory muscle training/conditioning, pressure threshold load, mechanical ventilation had to be a stated or implied goal of care. incremental threshold load, isocapnic/normocapnic hyperpnoea, The data extracted from each study about the participants were resistance load, mechanical ventilation, weaning, critically ill, age, gender, cause of admission, and clinical severity at admission. intubated/ventilated/tracheostomy. Baseline data for the following measures were also extracted: ventilation status, ventilation period, and whether ventilation was After the removal of duplicates, the two authors independently via an endotracheal tube or a tracheostomy tube. reviewed the titles and abstracts of all the items retrieved by the search against the eligibility criteria (Box 1). Full-text versions Intervention were obtained when there was ambiguity regarding eligibility. The experimental intervention was inspiratory muscle training. Disagreements about eligibility were resolved by discussion. Studies were not excluded on the basis of language or publication The details of the intervention extracted from each study were: the status. When eligible studies were identified, we attempted to use device used; the ventilation mode while training; the training them to identify further eligible studies by backward citation pressure and its progression; and the duration and frequency of tracking (ie, checking each article in their reference lists) and training and their progression. The apparatus/method used to forward citation tracking (ie, checking each article that had cited apply the inspiratory load was also recorded. The control them, as identified using GoogleScholar). intervention was sham or no inspiratory muscle training. Where sham training was used, the same details of the sham regimen were extracted. Assessment of characteristics of studies Outcome measures The outcome measures for which data were extracted were: Quality The methodological quality of the studies was assessed using inspiratory muscle strength (cmH2O); respiratory muscle endur- ance, as reported by the authors; the rapid shallow breathing index the PEDro scale.22 The PEDro scale consists of 11 criteria that are (also known as the Index of Tobin) (breaths/min/l); weaning rated as achieved or not. Items that are unclear are rated as not duration (days); weaning success; the total duration of mechanical achieved. The first criterion relates to external validity and the ventilation (days); the need for reintubation or tracheostomy; other 10 relate to risk of bias and/or completeness of reporting. length of ICU stay and hospital stay (days); use of non-invasive The ratings on these other 10 criteria can be summed to give a ventilation after extubation; survival; readmission; tolerability;

]GIF$DT)1_erugi([ Research 127 Screening of titles and abstracts of records Total retrieved by the electronic search (n = 1513) (0 to 10) Papers excluded after screening titles 6 and abstracts (n = 1471) 4 7 Screening of full text of potentially relevant 5 papers (n = 42) 6 4 Papers excluded after evaluation of full 8 text (n = 28) 4 7 • ineligible study design (n = 13) 6 ineligible participants (n = 11) Point estimate •• ineligible intervention (n = 4) and variability Papers identified from citation tracking reported (n = 1) Y Y Papers included in systematic review (n = 15) Y Studies included in systematic review (n = 10) Y Y Figure 1. Flow of studies through the review. Y Y and adverse events. Weaning duration was defined as the time Y from identification of readiness to wean, as defined by the authors Y and/or commencement of inspiratory muscle training, to the Y discontinuation of mechanical ventilation. Weaning success was defined as spontaneous breathing without mechanical support for Between-group a given DT_F]t[I$14 ime period of 48 hours or greater. Data about reintubation difference were limited to the same hospital admission. Tracheostomy was reported calculated as the proportion of patients tracheostomised after the Y commencement of training from among those who were not Y already tracheostomised. Readmissions to the ICU or to hospital Y were accepted as dichotomous outcomes if the groups had similar Y surveillance periods. $]FI[_T6DY Y Data analysis Y Y The two authors used a standard form to extract study Y characteristics and baseline and outcome data from the included Y studies. Disagreements were resolved by discussion. Where outcomes had been measured but not reported in a usable format, Intention- data were requested from the original authors. The meta-analyses to-treat were performed using RevMan 5.1 using random-effects models.27 analysis Data for continuous outcomes were pooled to calculate a weighted Y mean difference with 95% CI. Data for dichotomous outcomes were N pooled to calculate a relative risk with 95% CI. Y Y Results Y N Flow of studies through the review Y N The search retrieved 1513 papers. After screening titles and Y abstracts, 1471 papers were excluded and the remaining 42 were N obtained in full text for further evaluation. After evaluation of the full text, 28 papers were excluded as ineligible. Citation tracking of < 15% the remaining 14 eligible papers identified one additional eligible dropouts paper. These 15 papers represented 10 eligible studies,28–37 because one study was reported across five papers34,38–41 and N another study was reported across two papers.28,42 Figure 1 N presents the flow of studies through the review. Where necessary, N authors of the included studies were contacted to clarify Y interpretation of the published papers. Y N Description of studies Y N The criteria achieved by each of the studies on the PEDro scale Y are presented in Table 1. The characteristics of the studies are Y presented in Table 2. Note that Table 2 summarises key eligibility Assessor blinding N N Y N N N Y N Y N Therapist blinding N N N N N N N N N N Participant blinding N N N N I]$DT_F[1N N N N N N Groups Y Y Y N Y Y Y Y Y Y N = No, Y = Yes. similar a Original authors were contacted to confirm criteria that were unclear in the published papers. at baseline Concealed YY YN YY YN YN YN YY YN YN YY allocation Random allocation Table 1 Eligibility Y Caruso et al (2005) 29 Y Condessa et al (2013) 30 Y Y Elbouhy et al (2014) 32 Y Ibrahiem et al (2014) 33 Y Martin et al (2011) 34 Y Mohamed et al (2014) 35 Y Pascotini et al (2014) 36 Y Shimizu et al (2014) 37 Y PEDro scores for included studies (n = 10).a criteria and source Study Cader et al (2010) 28 Dixit et al (2014) 31

128 Elkins and Dentice: Inspiratory muscle training in intensive care Table 2 Summary of included studies (n = 10). Study Design Participants Intervention Outcome measures RCT Exp: Threshold device at 30% MIP in supine 45 deg up MIP Cader et al  Intubated via endotracheal tube RSBI (2010) 28 due to acute respiratory failure 5 min x twice daily x 7 d/wk x until weaned Weaning duration Pressure increased by 10% of initial MIP daily, Weaning success  Starting PS after  48 hr controlled as tolerated Duration of MV ventilation Stopped if adverse signs Tracheostomy Supplemental oxygen if needed Post-extubation NIV  MIP < 20 cmH2O Survival Exp: n = 21 (9 male) Con: No training Tolerability Age (yr) = 83 (SD 3) Exp: Adjustment of ventilator trigger sensitivity to 20% MIP APACHE II = 20 (SD 6) of initial MIP Weaning duration Con: n = 20 (10 male) 5 min x twice daily x 7 d/wk x until weaned Duration of MV Age (yr) = 82 (SD 7) Duration increased by 5 min each session, max 30 min Reintubation APACHE II = 20 (SD 7) Pressure increased by 10% of initial MIP, max 40% MIP Post-extubation NIV Stopped if adverse signs Survival Caruso et al RCT  Intubated due to acute respiratory Tolerability (2005) 29 failure or decreased consciousness Con: No training Adverse events  Receiving controlled ventilation or Exp: Threshold device at 40% MIP in supine 45 deg up MIP PS for  72 hr 10 breaths x 5 sets x twice daily x 7 d/wk x until weaned RSBI Pressure increased 10% of initial MIP daily, as tolerated Weaning duration Exp: n = 20 Stopped if adverse signs Weaning success Completed n = 12 (8 male) Supplemental oxygen if needed Duration of MV Age (yr) = 67 (SD 10) Reintubation APACHE II = 23 (SD 6) Con: No training Tracheostomy Post-extubation NIV Con: n = 20 Exp: Threshold device at 30% MIP in supine 45 deg up Survival Completed n = 13 (9 male) 6 breaths x 5 sets x twice daily x 7 d/wk x until Tolerability Age (yr) = 66 (SD 17) extubated Adverse events APACHE II = 24 (SD 7) 1 min rest between sets MIP Pressure increased after each set to achieve 6 to 8 Weaning duration Condessa et al RCT  Intubated via endotracheal tube on 0-10 RPE (2013) 30 due to acute respiratory failure from Pressure increased by 10% of patient’s MIP each day MIP trauma, medical or surgical causes Weaning success Con: No training Duration of MV  Starting PS after  48 hr controlled Exp: Adjustment of ventilator trigger sensitivity to 20% Length of stay ventilation of initial MIP MIP Exp: n = 45 (23 male) 5 min x twice daily x 5 d Age (yr) = 64 (SD 17) Duration increased by 5 min each session, max 30 min MIP APACHE II = 23 (8) Pressure increased by 10% of initial MIP each session Weaning success Stopped if adverse signs Duration of MV Con: n = 47 (28 male) Survival Age (yr) = 65 (SD 15) Con: No training Tolerability APACHE II = 23 (8) Adverse events Exp: Threshold device at 30% MIP in supine 45 deg up Dixit et al RCT  Intubated via endotracheal tube 6 breaths x 5-6 sets x twice daily x 3 d (2014) 31  Conscious and cooperative Rest between sets as required  Starting to wean after  24 hr Pressure increased by 1-2 cmH2O each session controlled ventilation Stopped if adverse signs  MIP < 35 cmH2O Supplemental oxygen if needed Exp: n = 15 Con: n = 15 Con: No training Elbouhy et al RCT  Intubated due to acute respiratory Exp: Threshold device at maximum tolerated pressure (2014) 32 failure from exacerbation of COPD 6-10 breaths x 4 sets x 5 d/wk x until weaned 2 min rest between sets Ibrahiem et al  Starting PS after a period of Pressure increased daily as tolerated (2014) 33 controlled ventilation Progressively lengthened spontaneous breathing trials Martin et al  MIP 15 to 30 cmH2O Con: Sham: threshold device modified to provide low load (2011) 34 Exp: n = 20 (16 male) Progressively lengthened spontaneous breathing trials Age (yr) = 61 (SD 12) Con: n = 20 (17 male) Age (yr) = 64 (SD 8) RCT  Intubated via endotracheal tube due to acute respiratory failure from trauma, medical or surgical causes  Receiving controlled ventilation for  72 hr Exp: n = 15 (12 male) Age (yr) = 49 (SD 3) Con: n = 15 (11 male) Age (yr) = 50 (SD 4) RCT  Tracheostomy due to acute respiratory failure from trauma, medical or surgical causes  Assist control, PS or SIMV  Failed 72-hr spontaneous breathing trial Exp: n = 35 (16 male) Age (yr) = 66 (SD12) SAPS II = 34 (SD 9) Con: n = 34 (15 male) Age (yr) = 65 (SD 11) SAPS II = 33 (SD 9)

Table 2 (Continued ) Design Participants Research 129 RCT Study  Intubated due to respiratory failure Intervention Outcome measures   48 hr controlled ventilation Exp: Threshold device at 30% MIP in supine 45 deg up MIP Mohamed et al Exp: n = 20 (11 male) Weaning duration (2014) 35 6 breaths x 5-6 sets x twice daily (12 sessions in 1 wk) Duration of MV Age (yr) = 55 (SD 8) Rest between sets as required Length of stay Con: n = 20 (10 male) Pressure increased by 1-2 cmH2O each session Stopped if adverse signs Age (yr) = 56 (SD 3) Supplemental oxygen if needed Con: No training Pascotini et al RCT  Weaning from MV via tracheostomy Exp: Threshold device at 20% MIP in supine 45 deg up MIP (2014) 36 Weaning success due to any cause 10 breaths x 3 sets x once daily x 7 d Tolerability Adverse events  Starting PS after  48 hr controlled 2 min rest between sets ventilation Stopped if adverse signs Exp: n = 7 (0 male) Age (yr) = 67 (SD 14) Con: No training Con: n = 7 (3 male) Age (yr) = 72 (SD 12) Shimizu et al RCT  Intubated due to acute respiratory Exp: Threshold device at 50% MIP in supine 45 deg up MIP (2014) 37 Weaning duration failure from trauma, medical or 10 breaths x 3 sets x twice daily x 4 d Duration of MV surgical causes 1 min rest between sets  Unsuccessful 90-min Stopped if adverse signs spontaneous breathing trial after Intermittent spontaneous breathing periods  48 hr controlled ventilation Exp: n = 5 Con: No training Age (yr) = 65 (SD 11) Spontaneous breathing periods of progressively Con: n = 8 longer duration Age (yr) = 62 (SD 14) APACHE II = Acute physiology and chronic health evaluation II score, Con = control group, COPD = chronic obstructive pulmonary disease, Exp = experimental group, MIP = maximal inspiratory pressure, MV = mechanical ventilation, NIV = non-invasive ventilation, PS = pressure support, RCT= randomised controlled trial, RPE = 0-10 Rating of Perceived Exertion scale, RSBI = rapid shallow breathing index, SAPS II = Simplified Acute Physiology Score II, SIMV = synchronised intermittent mandatory ventilation. ]G[7_TD$IF roup _[DT8$d]FI ata ]Df_9[FTI$ or $DT_aI]F01[ ge DF$a]T[_1I nd clinical scores of participants are presented as mean (SD[_2I1]FT$ D)_31T].[FID$ criteria. Full eligibility criteria for each study are presented in Intervention Appendix 2 on the eAddenda. Two studies29,32 applied the training load by adjusting the Quality threshold of the pressure trigger on the ventilator to 20% of the The mean PEDro score of the included studies was ]_DFI[5T1$5 .7. In all patient’s initial maximal inspiratory pressure. The remaining studies used a threshold pressure device. Most of these studies set studies, randomisation, group data, and between-group compar- an initial pressure of 20 to 40% of the patient’s initial maximal isons were reported. No study blinded participants or therapists, inspiratory pressure and progressed this by 10% or 1 to 2 cmH2O but two studies blinded assessors.30,34 Martin et al34 also used a each day or each session. However, one study37 set the threshold low-pressure sham training regimen in their control group and did trainer to 50% and did not progress it, and one study34 set it to the not inform participants about which regimen they had been maximal tolerated pressure and increased the pressure daily as allocated. Although this does achieve at least some of the benefits tolerated. of blinding, it does not achieve the criterion about participant blinding on the PEDro scale because this requires that participants Most of the studies using a threshold trainer positioned the would not be able to sense a difference if they were to experience patient in supine with elevation of the backrest to 45 deg, whereas both regimens. in the studies using ventilator adjustment, no positioning of the patient was specified. Training sessions were prescribed to last 5 to Participants 30 minutes or to consist of three to six sets of 6 to 10 breaths. The There were 394 participants across the 10 studies. The mean frequency of sessions varied from twice daily to 5 days per week. The training period lasted 3 to 7 days in half of the studies and was age of participants across the studies ranged from 50 to 83 years. continued until the patient had weaned in the other studies. All eight studies that reported gender included both male and Monitoring the patient for adverse cardiorespiratory changes female participants. The reasons for mechanical ventilation during the training was common. Most studies nominated a list of included respiratory, other medical, surgical, trauma, and reduced physiological criteria that would mandate the cessation of a consciousness. Four studies28,30,31,33 included only participants training session during the study (Appendix 4, see eAddenda). with an endotracheal tube, two studies34,36 included only participants with a tracheostomy, and the remaining four studies As discussed above, one study34 provided sham training to the did not specify the type of intubation. Three studies28–30 reported control group. All other control groups received no training. All APACHE II scores, which ranged from means of 20 to 67, and one studies provided usual care to both groups. study34 reported a mean SAPS II score of 34. Most studies reported a minimum duration of ventilation before inspiratory muscle Outcome measures training commenced, which ranged from 24 to 72 hours, but few In all studies, inspiratory muscle strength was measured as studies reported the actual time among their participants.29,34 All studies reported the baseline mean maximal inspiratory pressure maximal inspiratory pressure in cmH2O. In six of the studies,28– of the participants, which ranged from 15 to 51 cmH2O. For 30,34,36,37 this measurement involved the application of a reference, the healthy adults range is 104 to 129 cmH2O for men unidirectional valve for around 30 seconds to ensure that the and 70 to 98 cmH2O for women.43 Most studies nominated a list of pressure was generated from residual volume, as recom- physiological criteria that were required to be met before initiation mended.44 Multiple studies (detailed below) reported data for of a training session during the study (Appendix 3, see eAddenda). the remaining outcomes and the data were in (or could be converted to) the same units. This allowed meta-analyses to be reported as weighted mean differences in the original units of the measurement. The only exceptions were the measures of length of

igure_4)TD$FIG][(130 Elkins and Dentice: Inspiratory muscle training in intensive care stay and of post-extubation non-invasive ventilation use, which Study MD (95% CI) were only reported by single studies. Therefore, simple between- Cader28 Random group differences are reported in the text for each of these Condessa30 outcomes. Pooled Effect of intervention -50 -25 0 25 50 Inspiratory muscle strength and endurance The 10 studies provided data on a total of 366 participants. Favours training (br/min/L) Favours control Inspiratory muscle training significantly improved maximal Figure 4. Mean difference (95% CI) in rapid shallow breathing index (in br/min/L) inspiratory pressure, with a mean difference of 7 cmH2O (95% CI due to inspiratory muscle training, estimated by pooling data from two studies 5 to 9). See Figure 2, or Figure 3 on the eAddenda for a detailed (n = 105). forest plot. Tr_6$)DeIGgiF([u] No studies formally measured respiratory muscle endurance. However, one study conducted progressively lengthening sponta- Study MD (95% CI) neous breathing trials daily in both groups, where progression of Cader28 Random the duration of the breathing trials was contingent upon successful Carsuo29 completion.34 The training group successfully completed 77% of Condessa30 330 trials before weaning, whereas the control group successfully Dixit31 completed 73% of 382 trials before weaning. Although these results Mohamed35 favour the training group, statistical comparison would underesti- Shimizu37 mate the benefit in the training group because they undertook trials of progressively longer duration because of their greater Pooled success rate. Rapid shallow breathing index Two studies28,30 reported data on a total of 105 participants. Inspiratory muscle training significantly improved the rapid shallow breathing index, decreasing it by a mean difference of 15 breaths/min/l (95% CI 8 to 23). See Figure 4, or Figure 5 on the eAddenda for a detailed forest plot. Weaning duration -10 -5 0 5 10 Six studies28–31,35,37 reported the weaning duration of Favours training (days) Favours control F[4_2]TID$ 12 participants. On average, the inspiratory muscle training group had a shorter time to wean by 1.7 days, but this was not Figure 6. Mean difference (95% CI) in weaning duration (in days) due to inspiratory statistically significant (95% CI –0.3 to 3.6). See Figure 6, or muscle training, estimated by pooling data from six studies (n = TD]_4[2IF$ 12)[5.ID_T$]F Figure 7 on the eAddenda for a detailed forest plot. T]G[)Iu8rgFi$_De( Weaning success Subgroup RR (95% CI) Five studies28,30,32,34,36 reported the weaning outcome of Study Random 2F(gure_2)TD$IG]i[ 56 participants. Inspiratory muscle training significantly improved MD (95% CI) Patients with weaning difficulty Random Study Elbouhy32 Cader28 Carsuo29 Martin34 Condessa30 Dixit31 Subtotal Elbouhy32 Ibrahiem33 Patients without weaning difficulty Martin34 Cader28 Mohamed35 Condessa30 Pascotini36 Pascotini36 Shimizu37 Subtotal Pooled Pooled 0.5 0.7 1 1.5 2 -20 -10 0 10 20 Favours control Favours training Favours control (cmH20) Favours training Figure 8. Relative risk (95% CI) of weaning success due to inspiratory muscle training, estimated by pooling data from five studies (n = 256), with subgroup Figure 2. Mean difference (95% CI) in maximal inspiratory pressure (in cmH2O) due analysis by whether patients were known to have weaning difficulty before to inspiratory muscle training, estimated by pooling data from ten studies (n = 366). randomisation.

[(Figure_10)TD$IG] ResearcuiF(r]GI$DT)41_e[gh 131 Subgroup MD (95% CI) Study RR (95% CI) Study Random Cader28 Random Carsuo29 Patients with weaning difficulty Condessa30 Martin34 Elbouhy32 Pooled Martin34 Shimizu37 Subtotal 0.5 0.7 1 1.5 2 Patients without weaning difficulty Favours control Favours training Cader28 Caruso29 Figure 14. Relative risk (95% CI) of survival if weaning with inspiratory muscle Condessa30 training, estimated by pooling data from four studies (n = 242). Mohamed35 Subtotal the unit. Inspiratory muscle training also significantly shortened this period, by a mean of 5.8 days (95% CI 4.5 to 7.1). Pooled Non-invasive ventilation One study30 reported the prevalence of use of non-invasive -10 -5 0 5 10 ventilation after extubation among 77 participants. The likelihood Favours training (days) Favours control of needing non-invasive ventilation was lower in the inspiratory muscle training group with a relative risk of 0.44, but this was not Figure 10. Mean difference (95% CI) in duration of mechanical ventilation (in days) statistically significant (95% CI 0.13 to 1.44). However, the amount due to inspiratory muscle training, estimated by pooling data from seven studies of time spent on non-invasive ventilation was significantly shorter. (n = 305), with subgroup analysis by whether patients were known to have weaning In one study of 28 participants,28 the time spent on non-invasive difficulty before randomisation. ventilation was a mean of 16 hours shorter in the inspiratory muscle training group (95% CI 13 to 18). the likelihood of weaning success, with a relative risk of 1.34 (95% CI 1.02 to 1.76). See Figure 8, or Figure 9 on the eAddenda for a detailed Survival forest plot. Four studies28–30,34 provided data on 242 participants. The Duration of mechanical ventilation likelihood of survival was slightly higher in the inspiratory muscle Seven studies28–30,32,34,35,37 reported the total duration of training group (relative risk = 1.04), but this was not statistically significant (95% CI 0.96 to 1.13). See Figure 14, and also mechanical ventilation (including the pre-training period) for Figure 15 on the eAddenda for a detailed forest plot. 305 participants. The mean difference was 2.3 days shorter in the inspiratory muscle training group, but this was not statistically Readmission significant (95% CI –0.5 to 5.1). See Figure 10, and also Figure 11 on No studies reported any data about readmission to the ICU or to the eAddenda for a detailed forest plot. hospital after discharge. Reintubation Two studies29,30 reported the number of reintubations among Tolerability and adverse events Three studies28,30,36 reported that experimental group partici- 117 participants. Inspiratory muscle training had no significant effect on reintubations overall (relative risk = 1.00, 95% CI 0.38 to pants underwent training on all days during their weaning period. 2.64). See Figure 12 on the eAddenda for a detailed forest plot. Caruso et al29 reported that 88% of scheduled sessions were initiated and 86% of these were completed. Martin et al34 reported Tracheostomy that participants were able to participate in their scheduled Two studies28,30 reported the number of tracheostomies among activity on 84% of study days. 133 participants. The relative risk of receiving a tracheostomy was Among the studies that nominated a list of physiological criteria not significantly affected by inspiratory muscle training (relative that would mandate the cessation of a training session (Appendix risk = 1.31, 95% CI 0.31 to 5.50). See Figure 13 on the eAddenda for a 4, see eAddenda), only one29 reported which of these physiological detailed forest plot. changes occurred: paradoxical breathing, tachypnea, desaturation, haemodynamic instability and supraventricular tachycardia. In addition, several studies stated that there were no major adverse events in either group.30,36 Martin et al34 reported that the complications that occurred during the ICU stay were similar in each group. Length of stay Subgroup analyses One study32 reported the length of stay of 40 participants. In post-hoc subgroup analyses, the studies that enrolled Inspiratory muscle training significantly shortened the length of participants with known weaning diffculty were analysed stay in the ICU, by a mean of 4.5 days (95% CI 3.6 to 5.4). Inspiratory separately from the studies that enrolled participants who were not selected based on demonstrated weaning difficulty. muscle training significantly shortened the length of stay in the We accepted whatever definition the authors used to define hospital, by a mean of 4.4 days (95% CI 3.4 to 5.5). In another study weaning difficulty, which in the three relevant studies all of 40 participants,35 length of stay in ICU was also reported but it was only calculated from the training period not from admission to

132 Elkins and Dentice: Inspiratory muscle training in intensive care involved failing at least one spontaneous breathing trial.32,34,37 ventilation overall. However, the confidence interval certainly still Each definition is detailed in Appendix 2. The subgroup of includes values that would be clinically worthwhile (most studies that enrolled only patients with known weaning optimistically, shortening mechanical ventilation by 5 days). difficulty showed significant benefits in weaning success Therefore, further trials could help to refine whether the training (Figure 8) and duration of mechanical ventilation (Figure 10). has a significant effect and, if so, whether that effect is clinically The studies that did not limit their enrolment to patients with worthwhile. The subgroup analysis indicates that the mechanical known weaning difficulty did not show significant benefits on ventilation period is significantly shortened in patients with these outcomes. However, the statistical tests for subgroup known weaning difficulty. differences were not significant. The mean estimates of the effects on ICU and hospital length of Discussion stay (ie, reductions of just over 4 days) could each outweigh the time, expense and effort of undertaking the training. However, Trials examining the effects of inspiratory muscle training to further studies should assess this formally with economic facilitate weaning from mechanical ventilation in the ICU have evaluation. An important caveat is that these data come from a expanded greatly, with one new trial published in 2013 and six single study with 40 participants, TF[_D$m1]6I oderate methodological new trials published in 2014. Therefore, this review was able to quality, and no statement about usual care.32 Replication and analyse substantially more data than the 2011 review on the same refinement of these estimates with data from further trials is topic.20 The additional data also allow the review to draw warranted. important new conclusions about the potential effects of inspiratory muscle training in the ICU. Although the training did not significantly reduce the need for non-invasive ventilation after extubation, it did significantly Pooling data from all of the included studies, this review re- shorten the time for which it was used by about 16 hours. Again, confirmed that inspiratory muscle training significantly these results are from individual studies, so attempts to replicate improves maximal inspiratory pressure (MD = 7 cmH20, 95% the findings in future trials are warranted. CI 5 to 9). Although the dataset tripled in size, this estimate remained remarkably similar to the estimate in the previous Inspiratory muscle training was generally well tolerated, with review (MD = 8 cmH20, 95% CI 6 to 9). One limitation of this reports of all or most of the scheduled sessions being completed. newer meta-analysis is that four studies did not report any This finding is reinforced by an earlier case series, which methods to ensure that the measurement of maximal inspirato- concluded that inspiratory muscle training is safe in selected ry pressure occurred at residual volume.31–33,35 However, critically ill patients.46 To preserve this safety of the intervention, excluding these four studies makes negligible difference to clinicians are encouraged to collaborate with their multidisiplin- the pooled estimate (MD = 8 cmH20, 95% CI 5 to 10). Another ary ICU team to: agree on which patients are appropriate to issue affecting all of these meta-analyses is that the experimen- undergo the training (using criteria such as those listed in tal group had greater opportunity to practise the maximal Appendix 2); agree on which patients have appropriate physio- inspiratory pressure measurement procedure (eg, during titra- logical criteria to commence each session of training (such as tion of the training load) and to accommodate to the feeling of those listed in Appendix 3); ensure constant supervision of the loaded breathing during training. Therefore, it is possible that patient during training; and agree on clear physiological criteria some of the improvement in maximal inspiratory pressure in the for cessation of the training (such as those listed in Appendix 4). experimental group could be attributed to familiarisation with Furthermore, to assess the progress of the patient, inspiratory the technique. muscle strength must be measured reliably. This can be achieved with a unidirectional valve, which allows maximal inspiratory The exact amount of improvement in respiratory muscle pressure to be measured from residual volume even in uncooper- strength may not be important if the primary physiological ative patients.29,45 A patient’s progress towards readiness for mechanism by which the training improves clinical outcomes is via weaning could also be monitored with regular spontaneous improved respiratory muscle endurance. Unfortunately, the breathing trials.34 included studies did not report data to assess this adequately. However, a mechanism involving endurance would be consistent Although the quality of the included studies is slightly above with the training regimens used and the pooled estimate of the average for trials in physiotherapy,47 several sources of bias are effect of the training on the rapid shallow breathing index present. Participant and therapist blinding are unavoidably (MD = 15 breaths/min/l, 95% CI 8 to 23). difficult to achieve but assessor blinding, which is much easier to achieve, was only used in two trials.30,34 Concealed allocation The average effect of inspiratory muscle training on weaning and intention-to-treat analysis, which eliminate important sources success (ie, 1.34 times as likely to wean successfully) is arguably of bias and are also very easy to implement,25 were not used in clinically worthwhile. However, the lower end of the confidence FD$T_1[]sI7 ome of the studies. interval around this effect (ie, 1.02 times as likely to wean successfully) would not be clinically worthwhile in isolation. The The included studies differed in various ways, such as the estimate of the effect of the training on weaning duration was not background ventilation mode and the training mode, which may statistically significant, but the confidence interval did exclude have contributed to the statistical heterogeneity among the study the possibility of substantial worsening of weaning duration due results. Despite this, clear effects were obtained on several to the training. This is valuable information because prolonged outcomes. This suggests that the effect of inspiratory muscle mechanical ventilation is associated with increased mortality and training is robust enough to be identified as statistically significant other hazards.45 Therefore, while this review indicates that despite the use of a random effects model to account for the high inspiratory muscle training facilitates weaning by improving heterogeneity among the study results. However, it also highlights weaning success without extending the weaning period, further that factors that differ between the studies (such as usual care and trials could help to refine these estimates to clarify the statistical selection of patients) may have a large impact on the effect of and clinical significance of the training’s effect on the weaning inspiratory muscle training. Although clinicians could choose a process. training regimen that is within the range of regimens in the included studies, further research could help to identify the most The duration of mechanical ventilation consists of the weaning effective regimen, as well as determining whether it is effective period, but also the pre-weaning period, which is influenced by the over current best practice in the ICU, such as the bundle of care patient’s recovery from whatever pathophysiology caused them to known as ABCDE (ie, awakening, breathing coordination, delirium require admission to the ICU. It is therefore not surprising that monitoring/management, and early exercise/mobility).48 Until inspiratory muscle training applied during the weaning period such research is available, the results of this review should be alone did not significantly reduce the total duration of mechanical interpreted as indicating that inspiratory muscle training can have an important effect on clinical outcomes such as weaning and

Research 133 length of stay, but that the exact amount of improvement may be with inspiratory muscle training, such as those who have already highly dependent on the background of usual care to which the failed attempts to wean. training is added. Similarly, where only single studies provide data on some outcomes, readers should remember that the magnitude What is already known on this topic: In patients who have of the effect might be influenced by the details of that particular been receiving mechanical ventilation due to critical illness, study. For example, the single study32 that established the effect on inspiratory muscle training via a threshold pressure device or length of stay was conducted in Egypt and provides little adjustment of ventilator pressure trigger sensitivity improves information about the background usual care received by both maximal inspiratory pressure. groups. Furthermore, this study did not mention blinding of the What this study adds: Inspiratory muscle training FIiD$T[19_] mproves physicians who decided on extubation and discharge, which weaning success, with potential reductions in length of stay introduces a source of potential bias. and the duration of non-invasive ventilatory support after extubation. Patients who are having difficulty weaning may Another issue that is important to consider when interpreting particularly benefit from the training, especially in weaning the results of this review is standard sedation and weaning success and the duration of mechanical ventilation. The influ- protocols, which are effective in reducing weaning time.49 The ence of other aspects of intensive care management on the included studies generally provided little mention of such effect of the training is unclear. protocols, so the effects on inspiratory muscle training in units with standard sedation and weaning protocols is unclear. Also, as eAddenda: Figures 3, 5, 7, 9, 11, 12, 13 and 15 and Appendices more ICUs introduce standard weaning protocols, respiratory 1 to 4 can be found online at doi:10.1016/j.jphys.2015.05.016. muscle deconditioning during mechanical ventilation may become less problematic. This may alter the efficacy of inspiratory Ethics approval: No ethical approval was required for this muscle training observed in this review. As discussed above, study. however, mechanical ventilation still unloads the diaphragm to some extent so ventilator-induced diaphragmatic dysfunction Competing interests: Nil. may still occur.10 Source(s) of support: Nil. Acknowledgements: We thank Lisa Moodie and Julie Reeve for Physiotherapists working in an ICU need to decide whether the their work on the 2011 systematic review of this topic. We thank findings of this review indicate that they should use inspiratory Rafael Zambelli Pinto, Jose Meneses and Ali Alshami for their muscle training with their patients. As discussed above, the results assistance with translation. of this review may not be relevant in ICUs where the bundle of care Provenance: Not invited. Peer-reviewed. known as ABCDE is standard practice.48 However, recent research Correspondence: Mark Elkins, Department of Respiratory from Australia and the USA shows that a minority of units in some Medicine, Royal Prince Alfred Hospital, Sydney, Australia. Email: regions have implemented these elements of care.6,50 Physiothera- [email protected] pists must also consider how best to allocate their patient-contact time when other systematic reviews also indicate benefit from References other physiotherapist-delivered interventions in ICU, such as early mobilisation.51 Until more robust evidence is available about these 1. Australia and New Zealand Intensive Care Society: Adult Patient Database. Mel- issues, physiotherapists could use their assessment of individual bourne, Australia: Australia and New Zealand Intensive Care Society; 2007. patients to decide whether their clinical status indicates that they could benefit from respiratory muscle training. For example, 2. Wunsch H, Angus D, Harrison DA, Collange O, Fowler D, Hoste EAJ. Variation in repeated failed attempts to wean from mechanical ventilation may critical care services across North America and Western Europe. Crit Care Med. suggest a potential role for inspiratory muscle training. This is 2008;36:2787-e8. consistent with the post-hoc subgroup analyses, in which the subgroup of three studies that limited their enrolment to patients 3. UK Department of Health. Statistical press notice: monthly critical care beds, with documented failure to wean had statistically significant cancelled urgent operations and delayed transfers of care data, England. (February estimates of the mean effect on duration of mechanical ventilation 2013) https://www.gov.uk/government/news/statisticalpress-notice-monthly- and weaning success.32,34 These subgroup analyses should give critical-care-beds-cancelled-urgent-operations-and-delayed-transfers-of-care- readers more confidence that the training improves these out- data-england-february-2013.[accessed 24/04/2015] comes in patients with known failure to wean, but the lack of statistical significance on the comparison of subgroups indicates 4. Bednarik J, Vondracek P, Dusek L, Moracova E, Cundrle I. Risk factors for critical that we cannot be sure whether the effects are substantially less in illness polyneuromyopathy. J Neurol. 2005;252:343–351. unselected patient cohorts. 5. de Jonghe B, Sharshar T, Lefaucher JP, Authier FJ, Durand-Zaleski I, Boussarsar M, et In conclusion, this systematic review shows that inspiratory al. Paresis acquired in the intensive care unit: a prospective multicenter study. muscle strengthening that is achieved by threshold pressure JAMA. 2002;288:2859–2867. training or ventilator sensitivity adjustment can have significant benefits for patients weaning from mechanical ventilation in ICU. 6. Study TEAM, Investigators. Early mobilization and recovery in mechanically ven- These benefits include an improved breathing pattern, more tilated patients in the ICU: a bi-national, multi-centre, prospective cohort study. successful weaning, potential reductions in length of stay, and Crit Care. 2015;19:81. briefer use of non-invasive ventilatory support after extubation. These benefits are achieved safely when the training is applied to 7. Griffiths RD, Hall JB. Intensive care unit-acquired weakness. Crit Care Med. appropriate patients under constant supervision and with other 2010;38:779–787. safeguards in place. However, marked heterogeneity in these benefits was noted among the included studies. Many aspects of 8. Kress JP, Hall JB. ICU-acquired weakness and recovery from critical illness. New Engl the eligibility criteria, the training regimens and the usual care in J Med. 2014;370:1626–1635. these studies varied, so it is difficult to know the D$8FTIs][1_ ettings in which the average benefits observed in this review will be achieved. 9. Chang AT, Boots RJ, Brown MG, Paratz J, Hodges PW. Reduced inspiratory muscle However, subgroup analyses suggest that patients who have endurance following successful weaning from prolonged mechanical ventilation. already failed an attempt to wean may gain more benefit, Chest. 2005;128:553–559. particularly in the duration of mechanical ventilation and in the success of subsequent weaning attempts. Therefore, physiothera- 10. Petrof BJ, Jaber S, Matecki S. Ventilator-induced diaphragmatic dysfunction. Curr pists may consider using their clinical assessment and reasoning to Opin Crit Care. 2010;16:19–25. identify patients who appear to have greater potential to improve 11. Choi J, Tasota FJ, Hoffman LA. Mobility interventions to improve outcomes in patients undergoing prolonged mechanical ventilation: a review of the literature. Biol Res Nurs. 2008;10:21–33. 12. Boles JM, Bion J, Connors A, Herridge M, Melot C, Pearl R, et al. Weaning from mechanical ventilation. Statement of the Sixth International Consensus Confer- ence on Intensive Care Medicine. Eur Respir J. 2007;29:1033–1056. 13. Epstein SK. Weaning from ventilatory support. Curr Opin Crit Care. 2009;15:36–43. 14. Esteban A, Alia I, Tobin MJ, Gil A, Gordo F, Vallverdu I, et al. Effect of spontaneous breathing trial duration on outcome of attempts to discontinue mechanical ventilation. Am J Respir Crit Care Med. 1999;159:512–518. 15. Gosselink R, Bott J, Johnson M, Dean E, Nava S, Norrenberg M, et al. Physiotherapy for adult patients with critical illness: recommendations of the European Respira- tory Society and European Society of Intensive Care Medicine Task Force on Physiotherapy for Critically Ill Patients. Intens Care Med. 2008;34:1188–1199. 16. Adler D, Dupuis-Lozeron E, Richard J-C, Janssens J-P, Brochard L. Does inspiratory muscle dysfunction predict readmission after intensive care unit discharge? Am J Respir Crit Care Med. 2014;190:347–350. 17. Bissett B, Leditschke A, Neeman T, Boots R, Paratz J. Weaned but weary: one third of adult intensive care patients mechanically ventilated for 7 days or more have

134 Elkins and Dentice: Inspiratory muscle training in intensive care impaired inspiratory muscle endurance after successful weaning. Heart Lung. 35. Mohamed AR, El Basiouny HMS, Salem NM. Response of mechanically ventilated 2015;44:15–20. respiratory failure patients to respiratory muscles training. Med J Cairo Univ. 18. Schweickert WD, Gehlbach BK, Pohlman AS, Hall JB, Kress JP. Daily interruption of 2014;82:19–24. sedative infusions and complications of critical illness in mechanically ventilated patients. Crit Care Med. 2004;32:1272–1276. 36. Pascotini FS, Denardi C, Nunes GO, Trevisan ME, Antunes VP. Treinamento mus- 19. Moodie LH, Reeve JC, Vermeulen N, Elkins MR. Inspiratory muscle training to cular respirato´ rio em pacientes em desmame da ventilac¸a˜o mecaˆnica [Respiratory facilitate weaning from mechanical ventilation: protocol for a systematic review. muscle training in patients weaning from mechanical ventilation]. ABCS Health Sci. BMC Res Notes. 2011;4:283. 2014;39:12–16. 20. Moodie L, Reeve J, Elkins M. Inspiratory muscle training increases inspiratory muscle strength in patients weaning from mechanical ventilation: a systematic 37. Shimizu JM, Manzano RM, Quite´ rio RJ, Alegria VTC, Junqueira TT, El-Fakhouri S, et review. J Physiother. 2011;57:213–221. al. Determinant factors for mortality of patients receiving mechanical ventilation 21. Patsaki I, Padadopoulos E, Sidiras G, Christakou A, Kouvarakos A, Markaki V. The and effects of a protocol muscle training in weaning. Man Ther Posturology Rehabil J. effectiveness of inspiratory muscle training in weaning critically ill patients from 2014;12:136–142. mechanical ventilation. Hospital Chronicles. 2013;8:86–90. 22. Maher CG, Sherrington C, Herbert RD, Moseley AM, Elkins M. Reliability of the 38. Martin AD, Davenport W, Gonzalez-Rothi J, Baz M, Banner J, Caruso L, et al. PEDro scale for rating quality of randomized controlled trials. Phys Ther. 2003;83: Inspiratory muscle strength training improves outcome in difficult to wean 713–721. patients. Eur Respir J. 2006;28:369s. 23. de Morton N. The PEDro scale is a valid measure of the methodological quality of clinical trials: a demographic study. Aust J Physiother. 2009;55:129–133. 39. Martin AD, Caruso L, Banner M, Layon A, Gabrielli A. Inspiratory muscle strength 24. Macedo LG, Elkins MR, Maher CG, Moseley AM, Herbert RD, Sherrington C. training in failure to wean patients: interim analysis of a controlled trial. In: There was evidence of convergent and construct validity of Physiotherapy Proceedings of the American Thoracic Society. 2006 May 16–21. Evidence Database quality scale for physiotherapy trials. J Clin Epidemiol. 2010;63:920–925. 40. Martin AD, Davenport PW, Gonzalez-Rothi RJ, Baz MM, Banner MJ, Layon AJ, et al. 25. Elkins MR, Herbert RD, Moseley AM, Sherrington C, Maher CG. Rating the quality of Inspiratory muscle strength training improves maximal inspiratory pressure in trials in systematic reviews of physical therapy interventions. J Cardiopulm Phys difficult to wean patients. In: American Thoracic Society International Conference. Ther. 2010;21:20–26. 2007 May 18-23. 26. Elkins MR, Moseley AM, Sherrington C, Herbert RD, Maher CG. Growth in the Physiotherapy Evidence Database (PEDro) and the PEDro scale. Br J Sports Med. 41. Martin D, Smith BK, Gonzalez-Rothi R, Harman E, Harshavard-han D, Huang T, et al. 2013;47:188–189. Inspiratory muscle strength training improves weaning outcome in failure to wean 27. RevMan 2011. Review Manager (RevMan) [Computer program]. Version 5.1 Copen- patients. Cardiopulm Phys Ther J. 2009;20:21. hagen: The Nordic Cochrane Centre. The Cochrane Co-llaboration 2011 28. Cader SA, Vale RG, Castro JC, Bacelar SC, Biehl C, Gomes MC, et al. Inspiratory 42. Cader SA, Vale RGdS, Zamora VE, Costa CH. Dantas EHM. Extubation process in bed- muscle training improves maximal inspiratory pressure and may assist weaning in ridden elderly intensive care patients receiving inspiratory muscle training: a older intubated patients: a randomised trial. J Physiother. 2010;56:171–177. randomized clinical trial. Clin Interven Aging. 2012;7:437–443. 29. Caruso P, Denari SDC, Al Ruiz S, Bernal K, Manfrin GM, Friedrich C, et al. Inspiratory muscle training is ineffective in mechanically ventilated critically ill patients. 43. American Thoracic Society/European Respiratory Society.ATS/ERS Statement on Clinics. 2005;60:479–484. respiratory muscle testing. Am J Crit Care Med. 2002;166:518–624. 30. Condessa RL, Brauner JS, Saul AL, Baptista M, Silva ACT, Vieira SRR. Inspiratory muscle training did not accelerate weaning from mechanical ventilation but did 44. Caruso P, Friedrich C, Denari SD, Ruiz SA, Deheinzelin D. The unidirectional valve is improve tidal volume and maximal respiratory pressures: a randomised trial. J the best method to determine maximal inspiratory pressure during weaning. Physiother. 2013;59:101–107. Chest. 1999;115:1096–1101. 31. Dixit A, Prakash S. Effects of threshold inspiratory muscle training versus conven- tional physiotherapy on the weaning period of mechanically ventilated patients: a 45. Eskander N, Apostolakos MJ. Weaning from mechanical ventilation. Crit Care comparative study. Int J Physiother Res. 2014;2:424–428. Clinics. 2007;23:263–274. 32. Elbouhy MS, AbdelHalim HA, Hashem AMA. Effect of respiratory muscles training in weaning of mechanically ventilated COPD patients. Egypt J Chest Dis Tuberc. 46. Bissett B, Leditschke IA, Green M. Specific inspiratory muscle training is safe in 2014;63:679–687. selected patients who are ventilator-dependent: a case series. Intensive Crit Care 33. Ibrahiem AA, Mohamed AR, Saber HM. Effect of respiratory muscles training in Nurs. 2012;28:98–104. addition to standard chest physiotherapy on mechanically ventilated patients. J Med Res Prac. 2014;3:52–58. 47. Moseley AM, Herbert RD, Maher CG, Sherrington C, Elkins MR. Reported quality of 34. Martin AD, Smith BK, Davenport P, Harman E, Gonzalez-Rothi RJ, Baz M, et al. randomized controlled trials of physiotherapy interventions has improved over Inspiratory muscle strength training improves weaning outcome in failure to wean time. J Clin Epidemiol. 2011;64:594–601. patients: a randomized trial. Crit Care. 2011;15:R84. 48. Balas MC, Vasilevskis EE, Olsen KM, Schmid KK, Shostrom V, Cohen MZ, et al. Effectiveness and safety of the Awakening and Breathing Coordination, Delirium Monitoring/Management, and Early Exercise/Mobility (ABCDE) Bundle. Crit Care Med. 2014;42:1024–1036. 49. Blackwood B, Burns KEA, Cardwell CR, O’Halloran P. Protocolized versus non- protocolized weaning for reducing the duration of mechanical ventilation in critically ill adult patients. Cochrane Database Syst Rev. 2014;11:CD006904. 50. Miller MA, Govindan S, Watson SR, Hyzy RC, Iwashyna TJ. ABCDE, but in that order? A cross-sectional survey of Michigan ICU sedation, delirium and early mobility practices. Ann Am Thorac Soc. 2015. http://dx.doi.org/10.1513/AnnalsATS.201501- 066OC. published online 13 May 2015. 51. Kayambu G, Boots R, Paratz J. Physical therapy for the critically ill in the ICU: a systematic review. Crit Care Med. 2013;41:1543–1554.

Journal of Physiotherapy 61 (2015) 165–167 Appraisal Journal of Research Note PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Intention-to-treat analysis Introduction choice, perhaps because the allocated intervention is perceived as ineffective or the study is given low priority in the participant’s Trials that randomly allocate participants to groups tend to schedule. Adherence to the intervention may also be incomplete if produce groups with similar characteristics, especially when many participants forget doses. In trials of physiotherapy interventions, participants are randomised.1 Because the groups have similar there may be additional reasons such as that the intervention is too characteristics, their outcomes are also expected to be similar. If only exhausting, confronting or boring.7–9 Although rare, participants one of the groups then receives an intervention, the difference in may even receive more than the intended amount of intervention.10 outcomes between the groups can therefore be attributed to that intervention. If each group receives a unique intervention, the Use of other interventions difference in outcomes between the groups instead reflects the difference in the effects of the two interventions.2 Before recruit- Some participants may use interventions other than those they ment of participants into a clinical trial begins, the final protocol for were allocated – either instead of, or in addition to, their allocated the trial should be agreed upon by all the investigators, approved by intervention – perhaps because they are dissatisfied with their an ethics committee, and registered on a publicly available register.3 allocated intervention or with their clinical progress. Some may Ideally, the trial is then conducted exactly according to that protocol, even use the intervention that has been allocated to another with all participants who were randomised receiving all doses of group in the trial. For example, in a trial of exercise versus no their allocated intervention and undergoing all scheduled measure- intervention, participants in the control group may feel that they ment procedures.4 Unfortunately, the conduct of many clinical trials are missing out and independently decide to start the exercises deviates from the protocol, to some degree, which can introduce explained in the consent form for the trial.11 This could occur part unwanted systematic differences (ie, bias) between the groups after way through the trial or for the whole intervention period. randomisation. Investigator error The first two sections of this Research Note describe two types of deviations from the protocol: those where the intended intervention On rare occasions, an investigator may accidentally allocate a is not received, and those where the intended measurements are participant to a group that was not the group dictated by the not made. The third section describes a method of analysing the data randomisation process. For example, in a trial of dynamic splints for known as intention-to-treat analysis, which is designed to minimise distal radial fracture,12 an allocation was either misread or misheard the bias caused by some of these deviations from the protocol. and one participant received the wrong intervention. Alternatively, Intention-to-treat analysis involves analysing participants accord- an investigator may recruit a person who does not meet the ing to their initial group allocation (ie, randomisation) and not by the eligibility criteria for the trial, either by applying the eligibility treatment they actually receive. The final section explains how criteria incorrectly or by more information about the participant intention-to-treat analysis can be easily incorporated into physio- becoming available after randomisation.13 When ineligible people therapy trials. are mistakenly randomised, it may be possible to exclude them from the intention-to-treat analysis without adding bias, as long as there Protocol deviations in which participants do not receive their are few cases and the decision to remove participants is adjudicated allocated intervention by a committee blinded to treatment.14 Poor tolerance Protocol deviations in which participants do not undergo scheduled measurements Some interventions may not be fully tolerated by people with a particular characteristic. Sometimes it is easy to see the link, such Not all participants who enter a clinical trial complete all of the as a localised reaction to strapping tape5 or fainting during tilt scheduled measurements. For example, participants may be unco- table standing for neurorehabilitation.6 However, it is not always operative15,16 or too busy to attend outcome measurement.15 clear whether there is a link. For example, in a trial involving Others may not tolerate their allocated intervention and falsely people with spinal cord injury, one participant missed some sitting believe that further measurement of their outcome data will give balance training sessions due to an exacerbation of his back pain,7 unhelpful data or will expose their lack of adherence. In trials with which may or may not have been due to the training. Regardless of long follow-up, some participants may emigrate, making them the relationship between the intervention and the intolerance, unavailable to complete the measurements, despite willingness to such gaps in the intended treatment schedule could bias the trial’s continue. In trials involving elderly or seriously ill people, some estimate of the effect of the intervention. participants may die before completing the study.17 Because the outcomes of these participants are not measured, these all Poor adherence constitute loss to follow-up. Even if an intervention is perfectly tolerable, some participants Sometimes the reason for loss to follow-up may seem strongly may not adhere to some (or all) of their allocated intervention by related to the intervention. Imagine a trial in which the experimental http://dx.doi.org/10.1016/j.jphys.2015.05.013 1836-9553/ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

166 Appraisal Research Note intervention is very effective at preventing an increase in symptom Assessing effectiveness (as opposed to efficacy) is not the only severity that typically occurs over time in the disease of interest. advantage of the intention-to-treat approach. Another advantage With increases in symptom severity, many participants in the is that groups that were similar due to randomisation are control group might fail to attend follow-up visits because their prevented from becoming unbalanced by unequal loss to follow- symptoms are too severe for them to leave home. It is easy to see how up. As an example, assume in our hypothetical trial of stretching only measuring the remaining participants in the control group before swimming training that 20 participants are highly could introduce bias – that is, the only participants who contribute motivated and competitive, 60 have average motivation and follow-up data for the control group are those with unrepresenta- 20 are lazy. Because the trial is randomised, the groups are likely to tively mild symptom severity. This could prevent the analysis from be similar at baseline – with roughly 10 motivated, 30 average and identifying that the intervention reduces symptom severity. 10 lazy swimmers in each group. During the trial, most swimmers Conversely, the reason for loss to follow-up may seem random do their allocated intervention (stretching or no stretching). and unrelated to the intervention. As an example, consider However, half of the lazy swimmers in the stretching group decide participants who change their work hours and can no longer attend not to bother stretching. If we were to swap these five swimmers to scheduled measurement visits, such as occurred in a trial of neural the control group, the control group would have 15 lazy swimmers tissue management of neck pain.18 If this were truly random loss to and the experimental group would have only five. This is follow-up, it would reduce the power of the study but not introduce sometimes called an ‘as treated’ analysis. Differences in the group bias. However, causes that seem unrelated may be subtly and outcomes would become very difficult to interpret. Perhaps lazy systematically related to the intervention. For example, the swimmers don’t try hard enough during training to cause a cramp, participants may have chosen to increase their work hours because which would spuriously reduce the trial’s estimate of the benefit of their symptoms had improved. Without outcome data for these stretching. Conversely, perhaps lazy swimmers are less fit and participants, it is impossible to assess whether their absence therefore more prone to cramps, which might exaggerate the trial’s introduces bias. Therefore, regardless of the reason that participants estimate of the benefit of stretching. Even if we could measure are lost to follow-up, we must assume that the absence of their data laziness accurately, we don’t know enough about the relationship from the analysis can bias the estimate of the effect of the between laziness and cramps to know what the effect of swapping intervention. participants might be. Even if there is no effect on cramps, stacking the control group with lazy swimmers could still affect a secondary Intention-to-treat analysis outcome like a time trial. Another alternative, known as ‘per protocol’ analysis, is to omit participants from the analysis if An intention-to-treat analysis of a clinical trial includes all data they did not receive the allocated treatment, but this can create that can be obtained from each participant and analyses each similar problems of imbalanced groups. Overall, messing with the participant’s data in the group to which that participant was balanced groups created by randomisation is likely to introduce randomised, regardless of whether that participant received the unmeasurable and unpredictable bias. Intention-to-treat analysis allocated intervention.19–22 To demonstrate this, imagine a con- avoids this problem by preserving the original groups. trolled trial of stretches before swimming training sessions to prevent foot and calf cramps in 100 male swimmers. All available Doesn’t intention-to-treat analysis underestimate the data from each swimmer that was originally randomised to the intervention’s effect? stretching group would be analysed in that group, including swimmers who did only some (or even none) of the stretches. If some randomised participants receive the intervention Similarly, all available data from swimmers who were originally opposite to their intended intervention, then intention-to-treat allocated to not stretch would be analysed in the control group, analysis will produce a more conservative estimate of an effective including any who undertook stretching. Even if a swimmer from experimental intervention’s effect (ie, the between-group difference either group undertook additional interventions that were not part will be smaller in magnitude). However, to analyse participants by of the trial, his data would still be analysed in the group to which he what intervention they undertook risks introducing other biases, was originally randomised. such as disrupting the balance of motivated and lazy participants, as discussed above. Crucially, these biases are of unknown direction Why is intention-to-treat analysis recommended? and magnitude, making it impossible to interpret the result obtained with confidence. At least the bias from intention-to-treat analysis is Intention-to-treat analysis may seem counter-intuitive to some in a known direction, so readers can be confident that if a significant readers who think it is more appropriate to analyse participants effect is observed then the finding is not spurious and it may even be according to what intervention they actually received. For conservative. Therefore, intention-to-treat analysis is preferable to example, if swimmers were meant to stretch but they didn’t, it the alternative. may seem more appropriate to analyse their data in the control group (and vice versa). It may seem appropriate to exclude Are there any alternatives to intention-to-treat analysis? swimmers who undertook additional interventions. Such an analysis would provide some indication of the efficacy of the There are many techniques for estimating effects of treatment intervention.23 Efficacy refers to the effect of an intervention when (as distinct from effects of allocation) in clinical trials with non- it is administered exactly as intended, which often requires highly compliance. These methods attempt to preserve the exchange- controlled experimental conditions to eliminate real-world factors ability of groups generated by randomisation. For an introduction (such as a swimmer forgetting or running out of time to do the see the paper by Stuart and colleagues.25 When certain conditions stretches, in the hypothetical example). Determining the efficacy are met, these methods provide unbiased estimates of the effects of of an intervention can be an interesting question, and some treatment. Nonetheless, the primary analysis of clinical trials, at authors may choose to include such an analysis in a trial.24 least pragmatic randomised trials, should probably be an inten- Ultimately, however, the clinical usefulness of any intervention tion-to-treat analysis of the effects of allocation. depends on its effectiveness; that is, whether it has an effect when applied in everyday clinical practice where factors such as poor Applying intention-to-treat analysis in physiotherapy research adherence potentially reduce its effect.23 Therefore, it is arguably more important that clinical trials include an intention-to-treat For many trials of physiotherapy interventions, the published analysis. report neither mentions nor demonstrates that intention-to-treat analysis was undertaken. In 2011, only 18% of all trials on the

Appraisal Research Note 167 Physiotherapy Evidence Database (PEDro; www.pedro.org.au) same as those that were randomly generated, readers often assume were rated as having used intention-to-treat analysis.26 Trials that intention-to-treat analysis has been conducted. However, the can fail to meet this criterion by not using intention-to-treat researchers may have swapped equal numbers of participants analysis, by not reporting that intention-to-treat analysis was between the groups – perhaps due to adherence issues. This is why used, or by misinterpreting the definition of intention-to-treat the explanatory notes for tools such as the PEDro Scale specify that analysis. A study of trials published in 1997 showed varied a study should only be accepted as having used intention-to-treat compliance with the definition.23 However, since the CONSORT analysis if: the published report states that the analysis was by Statement27 and key publications about intention-to-treat analysis ‘intention to treat’ and does not provide any evidence that were published,20,21 there has been improved reporting of the participants were excluded unnecessarily; or the published report information required for readers to interpret whether analysis was states that all available participants were analysed in the groups to by intention to treat in physiotherapy trial reports.26 which they were originally randomised.26 The low prevalence of use of intention-to-treat analysis in Summary physiotherapy trials is disappointing because analysing data in this way only requires two simple steps. First, participants who Researchers conducting a clinical trial should first strive to discontinue their allocated intervention for any reason must be ensure that each participant adheres to the intervention as randomly encouraged to undergo the remaining planned outcome measure- allocated and undergoes all the planned measurements of outcomes. ment, if at all possible. Briefly explaining to these participants the However, when a participant does not receive the intervention as importance of obtaining the data may assist in gaining their intended, ongoing participation in all the planned measurements of consent for ongoing data collection. Second, the statistical analysis outcomes should be encouraged. Whatever data can be obtained of whatever data can be obtained must leave the participants’ data should then be included in the analysis, with each participant’s in the groups created by the random allocation process. Of course, data being retained in the group to which that participant was for readers to know that intention-to-treat analysis was used, a originally randomised. Readers of reports of clinical trials should crucial third step is to indicate in the published report that these take some reassurance from the use of intention-to-treat analysis, two steps were taken. especially when other information in the paper is consistent with such an analysis having been undertaken. Minimising poor adherence and loss to follow-up Mark R Elkinsa and Anne M Moseleyb Some causes of poor adherence and loss to follow-up can be aSydney Medical School, The University of Sydney, Australia minimised by carefully planning who enrols into the trial and how bThe George Institute for Global Health and Sydney Medical School, investigators maintain contact with them. Missing data can also be minimised by early checking for missing or invalid data and The University of Sydney, Australia retrieving the data retrospectively, where valid and possible. Regular audits from an external researcher during the course of the References trial may be helpful. To the extent that the researchers can maximise adherence and follow-up, they will reduce the potential 1. Altman DG, Bland JM. BMJ. 1999;318:1209.1. for intention-to-treat analysis to underestimate the effect of the 2. Herbert RD. Aust J Physiother. 2005;51:58–60. intervention. 3. Costa LOP, et al. Braz J Phys Ther. 2012;16:v–ix. 4. Gallo RBS, et al. J Physiother. 2013;59:109–116. Is imputation of missing data part of intention-to-treat analysis? 5. Parreira PCS, et al. J Physiother. 2014;60:90–96. 6. Leung J, et al. J Physiother. 2014;60:201–208. The original papers about intention-to-treat analysis did not 7. Harvey LA, et al. J Physiother. 2011;57:83–90. recommend imputation of missing data.19–21 This is also the stance 8. Trudelle-Jackson E, Smith SS. Arch Phys Med Rehabil. 2004;85:1056–1062. of the Cochrane Handbook,22 the CONSORT statement,27 and the 9. Braun S, et al. J Physiother. 2011;57:27–34. PEDro Scale (http://www.pedro.org.au/wp-content/uploads/ 10. Ralston KE, et al. J Physiother. 2013;59:237–243. PEDro_scale.pdf), which describe intention-to-treat analysis as 11. Courneya KS, et al. Eur J Cancer Care. 2003;12:347–357. simply collecting data from each participant, wherever available, 12. Jongs RA, et al. J Physiother. 2012;58:173–180. and analysing the data in that participant’s original group, 13. Hancock M, et al. Lancet. 2007;370:1638–1643. regardless of what intervention they received. Imputation of data 14. Fergusson D, et al. BMJ. 2002;325:652. can be performed alongside an intention-to-treat analysis,28 15. Van Wely L, et al. J Physiother. 2014;60:40–49. although readers should have more confidence in the imputed 16. Sevenhuysen S, et al. J Physiother. 2014;60:209–216. data when logical assumptions about the missing data can be 17. Condessa RL, et al. J Physiother. 2013;59:101–107. made.6 18. Nee RJ, et al. J Physiother. 2012;58:23–31. 19. Hill AB. Principles of Medical Statistics. 7th ed London: The Lancet; 1961. Does complete follow-up guarantee intention-to-treat analysis? 20. Newell DJ. Int J Epidemiol. 1992;21:837–841. 21. Lewis JA, Machin D. Brit J Cancer. 1993;68:647–650. If the published report of a clinical trial states that all 22. Higgins JPT, Green S, (eds.), Cochrane Handbook for Systematic Reviews of Interven- participants were followed up and the final group sizes are the tions. 2011. www.cochrane-handbook.org. 23. Hollis S, Campbell F. BMJ. 1999;319:670. 24. de Jong LD, et al. J Physiother. 2013;59:245–254. 25. Stuart EA, et al. Prev Sci. 2008;9:288–298. 26. Moseley A, et al. J Clin Epidemiol. 2011;64:594–601. 27. Moher D, et al. BMJ. 2010;340:c869. 28. Krist M, et al. J Physiother. 2013;59:15–23.

Journal of Physiotherapy 61 (2015) 106–116 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Research Massage therapy has short-term benefits for people with common musculoskeletal disorders compared to no treatment: a systematic review Diederik C Bervoets a, Pim AJ Luijsterburg a, Jeroen JN Alessie b, Martijn J Buijs b, Arianne P Verhagen a a Department of General Practice, Erasmus MC, University Medical Center, Rotterdam; b Avans University of Applied Science, Breda, The Netherlands KEY WORDS ABSTRACT Massage therapy Question: Is massage therapy effective for people with musculoskeletal disorders compared to any other Systematic review treatment or no treatment? Design: Systematic review of randomised clinical trials. Participants: Randomised clinical trials People with musculoskeletal disorders. Interventions: Massage therapy (manual manipulation of the Physical therapy soft tissues) as a stand-alone intervention. Outcome: The primary outcomes were pain and function. Results: The 26 eligible randomised trials involved 2565 participants. The mean sample size was 95 participants (range 16 to 579) per study; 10 studies were considered to be at low risk of bias. Overall, low-to-moderate-level evidence indicated that massage reduces pain in the short term compared to no treatment in people with shoulder pain and osteoarthritis of the knee, but not in those with low back pain or neck pain. Furthermore, low-to-moderate-level evidence indicated that massage improves function in the short term compared to no treatment in people with low back pain, knee arthritis or shoulder pain. Low-to-very-low-level evidence from single studies indicated no clear benefits of massage over acupuncture, joint mobilisation, manipulation or relaxation therapy in people with fibromyalgia, low back pain and general musculoskeletal pain. Conclusions: Massage therapy, as a stand-alone treatment, reduces pain and improves function compared to no treatment in some musculoskeletal conditions. When massage is compared to another active treatment, no clear benefit was evident. [Bervoets DC, Luijsterburg PAJ, Alessie JJN, Buijs MJ, Verhagen AP (2015) Massage therapy has short-term benefits for people with common musculoskeletal disorders compared to no treatment: a systematic review. Journal of Physiotherapy 61: 106–116] ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Background included (spinal) manipulation and the use of mechanical devices. Another definition of massage is ‘a systematic manipulation of the A considerable proportion of the population experiences soft tissues of the body with rhythmical pressure and stroking to musculoskeletal disorders.1 The Global Burden of Disease prevent, develop, maintain, rehabilitate, or augment physical 2010 Study shows that the musculoskeletal disorders are the function or relieve pain’.9 Most massage styles consist of one or fourth greatest burden on health throughout the world, causing more of the following actions: effleurage (a gliding or sliding 21.3% of years lived with disability.2 The most affected areas of the movement over the skin), petrissage (lifting, wringing or squeezing body are the low back, neck, shoulder and the knee, with a point of soft tissues in a kneading motion, or pressing or rolling of the prevalence varying between 20 and 50% of the population.1,2 tissues), friction (penetrating pressure applied through the fingertips), tapotement (strike the tissues at a rapid rate) and Massage therapy is one of the earliest therapeutic tools used vibration.3,10–12 to relieve pain.3,4 It has been promoted as a treatment of choice for numerous conditions such as musculoskeletal disorders, The specific mechanisms of action of massage therapy are stress and pregnancy.3 With its popularity for pain relief and unknown, but various physiological responses to massage therapy recovery of function, massage therapy has become a widely have been claimed. These mechanisms include: increased lymph accepted treatment for musculoskeletal disorders.5 In physiother- flow, a shift from sympathetic to parasympathetic response, apy practices, massage therapy plays a major role in the treatment prevention of fibrosis, increased clearance of blood lactate, and of patients with musculoskeletal disorders. In a large cohort study, effects on the immune system, cognition and pain.9,13,14 A popular 87% of participants with complaints of the arm, neck and/or claim is that massage therapy can increase blood flow to the shoulder were treated with massage therapy, often in combination muscles. However, this claim has been questioned, as increasing with exercise therapy.6 scientific evidence has shown no influence of massage therapy on blood flow.15 Massage does seem to produce local biochemical Massage therapy can be defined in different ways. Recently, the changes, which might lead to increased neural activity at the spinal Ottawa panel defined massage as ‘soft tissue and joint manipula- cord level and subcortical nuclei, which might affect mood and tion using the hands or a handheld device’.7,8 This definition also http://dx.doi.org/10.1016/j.jphys.2015.05.018 1836-9553/ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/).

Research 107 pain perception.5,16 Therefore, massage therapy could potentially blinding of participants and personnel, blinding of outcome reduce anxiety, depression and pain through the increase of assessment, incomplete outcome data, selective outcome report- serotonin and endorphins.13 ing and ‘other issues’. Each item was rated as being at ‘low’, ‘unclear’ or ‘high’ risk of bias. Two review authors (DB, PL) Previous systematic reviews of the effectiveness of massage independently assessed the risk of bias. The discrepancies were have shown mainly short-term positive effects on low back pain, resolved through discussion and disagreements were discussed neck pain and shoulder pain.5,7,8,17 However, these reviews have with a third review author (AV). A study was defined as being at combined studies that used a variety of massage, spinal low risk of bias when it fulfilled six or more of the criteria. manipulation and mobilisation techniques, often as part of complementary and alternative medicine interventions.7,8 The Participants present review aimed to evaluate the currently available evidence The inclusion criteria for participants are shown in Box of massage (ie, manual manipulation of soft tissues) as a stand- alone treatment compared to no intervention or other interven- 1. Studies were excluded if the participants had severe pathology tions on pain and functional status for people with musculoskeletal such as a fracture, nerve damage, psychological disorders (eg, disorders. depression) or sport injuries. Therefore, the research question for this systemic review was: Intervention Is massage therapy effective for people with musculoskeletal Studies were excluded if the intervention involved joint disorders compared to any other treatment or no treatment? manipulation, energy manipulation (eg, Reiki or polarity), or Methods mechanical devices (eg, roptrotherapy). The massage therapy had to be a stand-alone treatment; trials were excluded if massage Identification and selection of studies therapy was combined with another intervention (eg, massage plus joint mobilisation compared to no treatment) or additional to PubMed, PEDro and CINAHL were searched from inception until other active interventions (eg, massage plus exercise compared to October 2014, using medical subject headings (MeSH) and key exercise alone). The comparison therapy could not be an words including anatomical terms, disorder or syndrome terms, alternative form of massage. and treatment terms. The full search strategy is presented in Appendix 1 on the eAddenda. There were no language restrictions. Outcome measures The references of the systematic reviews and (quasi-) randomised The outcomes of interest were pain and function. Outcome data trials identified by the electronic searches were also scanned for potentially relevant articles. were categorised as short term (post treatment up to 12 weeks) or long term (12 weeks or over). Published, randomised controlled trials that studied the effect of massage as a stand-alone intervention (compared to no Data analysis treatment or to another active intervention) in people aged over 18 years with common musculoskeletal disorders (Box 1) were One review author (DB) extracted data using a standardised, included. Two review authors (DB, PL) independently performed piloted data extraction form. A second review author (AV) checked the selection. First, titles and abstracts were screened for possible this process by performing data extraction (independently) on a eligibility. Next, the full-text articles were independently screened random set of studies and comparing the results. Any disagree- for definite inclusion. The review authors resolved discrepancies ments were resolved by discussion. In cases of more than 5% of through discussion or by a third author (AV). disagreements with the random sample, two review authors performed the data extraction of all studies. Data were extracted Assessment of characteristics of studies on patient population, experimental and control interventions, and outcomes. All original data on outcomes were converted into effect Quality estimates, which were reported as: a mean difference (MD) when a To assess the risk of bias, the tool from the Cochrane Back continuous outcome was measured on comparable instruments in the included studies, a standardised mean difference (SMD) when a Review Group was used. This tool describes seven domains, continuous outcome was measured on different instruments in the including 12 items: sequence generation, allocation concealment, included studies, or relative risk (RR) when the outcome was dichotomous. Each of these estimates was reported with a 95% Box 1. Inclusion criteria. confidence interval (CI) whenever possible. An effect of 15% or more was considered to be clinically relevant. Design  Randomised trial Statistical analyses were performed using Review Manager  Published in any language 5.2.18 Statistical heterogeneity was determined using I2 tests, which were interpreted as follows: 0 to 40% no heterogeneity; Participants 40 to 70% moderate heterogeneity; and 70 to 100% considerable  Adults with a common musculoskeletal disordera heterogeneity.19 For statistical pooling, the random effects model was used. Intervention  Massage, defined as systematic manual manipulation of The quality of the evidence was assessed using the GRADE the soft tissues of the body with rhythmical pressure and approach.20 The quality of the evidence starts at high when at least stroking two trials provide results for an outcome. The quality is reduced by one level for each of the following domains not met: limitations of Outcome measures the study design, defined as > 25% of the participants from studies  Pain with a high risk of bias; inconsistency, defined as statistical  Function heterogeneity (I2 > 40%) or inconsistent findings among studies (< 75% of the participants reported findings in the same direction); Comparisons indirectness, defined as generalisability of the findings; impreci-  Massage versus no treatment (wait list control, sham, rest sion of results, defined as total number of participants < 300 for a or usual care) dichotomous outcome and < 400 for continuous outcome; and  Massage versus other active treatments (exercise therapy, ‘other’, such as publication bias, flawed design or massive dropout. joint manipulation, relaxation therapy) Single randomised trials (n < 400) were considered to be a Common musculoskeletal disorders were defined by the International Classification of Primary Care (ICPC) codes chapter L: locomotor system.46

108 Bervoets et al: Massage for musculoskeletal disorders inconsistent and imprecise, and provided low-quality evidence, used.22–25 Ten studies described the use of co-interventions (ie, which could be further downgraded to very-low-quality evidence. interventions that were not a part of the planned treatment and control interventions), which were similar across groups. Subgroup analyses were conducted according to the different conditions: low back pain, neck pain, shoulder pain or knee pain. Twenty studies allocated participants to a no-treatment, placebo or inactive-treatment control group, which included: a Results waiting list, no treatment, usual care, sham laser, light hand touch, medication, self-care educational materials/book, corset or passive Flow of studies through the review physical treatments (eg, ultrasound therapy, hot packs or phonophoresis). Usual/standard care could include pharmacologi- The initial search identified 1473 records. After removing cal therapy (muscle relaxants, pain medication) or advice on duplicates, 1313 unique hits were retained for further assessment. exercise or posture, but this was not always clearly described. Of these, 1249 records were excluded on the basis of titles and Treatment periods varied from one session (n = 3) to 10 weeks abstracts, and 64 full-text studies were read to assess for eligibility. (with four to ten sessions). Of these, 38 studies were excluded and 26 studies were judged to be eligible for inclusion in this review (Figure 1). Eleven studies included one or more active treatment control groups: (progressive muscle) relaxation therapy (n = 5), exercises Description of studies (n = 2), acupuncture (n = 2), Alexander technique lessons, joint mobilisation and manipulation. All studies were published in English. One trial11 had a crossover design and only the data from the first period were Outcomes used in the review. The mean sample size among the studies was Pain (intensity, threshold) was assessed in 22 studies and 95 participants (range 16 to 579). Fourteen studies recruited fewer than 25 participants per arm and the smallest study arm recruited function in 14 studies. Although data were not extracted for this eight participants.21 The characteristics of the included studies are review, the other outcome measures presented were range of presented in Table 1. motion, muscle tension, stiffness and psychosocial outcomes (anxiety, depression, bothersomeness). No studies assessed Participants recovery or sick leave. Thirteen studies did not have any follow- Eight studies included participants with (chronic) low back pain up measurement; in other studies, the follow-up period varied between 12 to 52 weeks. (Table 1). Other studies included participants with shoulder pain (n = 4), fibromyalgia (n = 3), osteoarthritis of the knee (n = 3), Risk of bias in included studies chronic musculoskeletal pain (n = 2), neck pain (n = 2) or chronic patellar tendinopathy, carpal tunnel syndrome, hand pain and Randomisation hand osteoarthritis (one study each). Of the 26 included studies, 10 were assessed as having low risk Interventions of bias (Figure 2). Fifteen studies provided detailed information on Despite the strict definitions in the present review, the studies allocation sequence generation, which were regarded as adequate. In one study, the randomisation procedure was not regarded as used a broad variety of massage techniques, durations and adequate and the remaining 10 studies did not report sufficient frequencies. The massage therapies included: Swedish massage details on the randomisation procedure (unclear risk). Ten studies (n = 5), Thai massage (n = 4), self-massage (n = 1) or a combination reported adequate allocation concealment methods. of techniques (eg, therapeutic and structural massage) (n = 12) Blinding (F([ure_i1)TD$IG]g Table 1). Four studies did not clearly describe the type of massage When the participant was not blinded and self-assessed the Titles and abstracts screened primary outcomes, the outcome assessment was scored as not (n = 1313) blinded (n = 9). Blinding of personnel was rated as high risk, as it is impossible to blind therapists. Two studies were scored as low risk Papers excluded after screening in regards to participant blinding; participants were told that they titles/abstracts (n = 1249) might receive a sham treatment,24 or were blinded to type of massage,26 and 15 studies did not mention blinding (unclear risk). Potentially relevant papers retrieved for evaluation of full text (n = 64) Follow-up Nine studies gave unclear information on withdrawals and loss to follow-up. An intention-to-treat analysis was presented in 11 studies. Effects of interventions Papers excluded after evaluation of Seventeen studies adequately reported continuous data in full text (n = 38) order to be able to calculate effect estimates. Two studies reported dichotomous data (Table 1). • ineligible study design (n = 15) • experimental intervention was not Massage versus no-treatment control massage alone (n = 12) Pain • control intervention was massage All studies measured pain on a visual analogue scale (VAS); (n = 6) therefore, the results are presented as mean differences on a 0 to • ineligible participants (n = 5) 100 mm scale. Eight studies presented short-term pain da- ta,11,24,27–32 of which one had high risk of bias.11 The included Studies included in review (n = 26) disorders were: neck pain,24 osteoarthritis of the knee,11,30 shoulder pain28,29,32 and low back pain.27,31 Three studies11,24,30 Figure 1. Flow of studies through the review. presented long-term data, which varied from 12 to 24 weeks (Table 1).

Table 1 Participants Intervention Characteristics of the included studies. Exp: Swedish massage, n = 11 Study Con 1: Usual care, n = 13 Con 2: Usual care and phone Alnigenis (2001)38 Fibromyalgia, n = 37 High risk of bias Age (yr) range = 21 to 65 calls (to increase Gender = 100% F compliance), n = 13 United States (rheumatology clinic) loss to follow-up = 21 Atkins (2013)39 Knee OA, n = 40 Exp: Self-massage (20 min x High risk of bias Age (yr) > 50 2/wk x 12 wk), n = 21 Buttagat (2011)27 Gender = 89% F Low risk of bias United States (general population) Con: WLC, n = 19 Buttagat (2012)29 Back pain and myofascial trigger Exp: Thai massage (30 min x 1 High risk of bias points, n = 36 session), n = 18 Buttagat (2012)28 Age (yr) mean = 23 yr Con: Rest on bed (30 min x 1 Low risk of bias Gender = 56% F session), n = 18 Thailand (general population) Cherkin (2001)34 Exp: Thai massage (30 min x 3/wk x Low risk of bias Shoulder pain (scapulocostal 3 wk), n = 10 syndrome), n = 20 Cherkin (2011)26 Con: Physiotherapy modalities Low risk of bias Age (yr) mean, range = 25, 18 to 50 incl. ultrasound and hot Gender = 85% F packs (3/wk x 3 wk), n = 10 Thailand (general population) Exp: Thai massage (30 min x 1 Shoulder pain (scapulocostal session), n = 20 syndrome), n = 40 Con: Physiotherapy modalities Age (yr) mean, range = 27, 18 to 50 incl. ultrasound and hot Gender = 80% F packs (30 min x 1 session), Thailand (general population) n = 20 Low back pain, n = 262 Exp 1: Therapeutic massage Age (yr) mean, range = 45, 20 to 70 (10 wk), n = 78 Gender = 80% F United States (primary care) Exp 2: Traditional Chinese medical acupuncture Chronic low back pain, n = 402 (10 wk), n = 94 Age (yr) range = 20 to 65 Gender = 63% F Con: Self-care educational United States (general population) materials, n = 90 Exp 1: Structural massage (10 wk), n = 132 Exp 2: Relaxation (10 wk), n = 136 Con: Usual care (10 wk), n = 133

Outcomes Results Effect size Research Pain and mobility mean (SD) or n/N (%) SMD (95% CI) unless noted (AIMS) 4 wk (change) Post session Pain, stiffness, and Pain = Exp: 0.10; Pain = Exp minus physical functional Con 1: –0.02; Con: –1.3 (–2.0 to –0.6) disability (WOMAC) Con 2: –0.1 Pain (VAS) Mobility = Exp: –0.6; 3 wk Pain (VAS) Con 1: 2.7; Pain = Exp minus Con 2: 0.9 Con: –1.3 (–2.3 to –0.3) Pain (VAS) 28 wk (change) 5 wk Function (RDQ) Pain = Exp: 0.9; Pain = Exp minus Con 1: –1.0; Con: –1.5 (–2.5 to –0.5) Function (RDQ) Con 2: –0.3 Mobility = Exp: 0.6; Post session Con 1: 0.0; Pain = Exp minus Con 2: –1.1 Con: –0.4 (–1.0 to 0.3) No usable data 10 wk Function = Exp 1 minus Post session Con: –0.4 (–0.7 to –0.09); Pain = Exp: 26 (19); Exp 1 minus Exp 2: –0.3 Con: 48 (14) (–0.6 to 0.05) 3 wk 52 wk Pain = Exp: 7 (13); Function = Exp 1 minus Con: 28 (17) Con: 0.08 (–0.2 to 0.4); Exp 1 minus Exp 2: 5 wk –0.2 (–0.5 to 0.1) Pain = Exp: 7 (6); Con: 34 (21) 10 wk Function = Exp 1 minus Post session Con: –0.6 (–0.8 to –0.3); Pain = Exp: 39 (24); Exp 1 minus Exp 2: 0.1 Con: 47 (18) (–0.1 to 0.4) 10 wk 52 wk Function = Exp 1: 6.3 (5.4); Exp 2: 7.9 (6.9); Function = Exp 1 minus Con: 8.8 (6.8) Con: –0.04 (–0.3 to 0.2); Exp 1 minus Exp 2: 0.3 52 wk (0.01 to 0.5) Function = Exp 1: 6.8 (5.9); Exp 2: 8.0 (6.7); Con: 6.4 (3.9) 10 wk Function = Exp 1: 6.5 (4.0); Exp 2: 6 (4.4); Con: 9 (4.5) 52 wk Function = Exp 1: 7.2 (4.3); Exp 2: 6 (4.8); Con: 7.4 (4.7) 109

Table 1 (Continued ) Participants Intervention Study Field (2002)36 Fibromyalgia, n = 20 Exp: Combination of massages High risk of bias Age (yr) mean = 51 incl. Swedish and Shiatsu Gender = NS (2/wk x 5 wk), n = 10 Field (2004)21 United States (general population) High risk of bias Con: Progressive muscle Carpal tunnel syndrome, n = 16 relaxation therapy (2/wk x Field (2007)40 Age (yr) mean, range = 47, 20 to 65 5 wk), n = 10 High risk of bias Gender = 93% F United States (general population) Exp: Massage therapy incl. Field (2011)22 self-massage (4 wk), n = 8 High risk of bias Hand arthritis, n = 22 Age (yr) mean, range = 47, 20 to 65 Con: Usual care (4 wk), n = 8 Hasson (2004)23 Gender = 93% F High risk of bias United States (general population) Exp: Massage therapy incl. self-massage (1/wk x Hernandez-Reif (2001)37 Hand pain, n = 46 4 wk), n = 11 High risk of bias Age (yr) mean = 50 Gender = NS Con: Usual care (4 wk), n = 11 Irnich (2001)24 United States (general population) Low risk of bias Exp: Massage therapy (4 wk), Chronic musculoskeletal pain, n = 23 Little (2008)41 n = 129 Low risk of bias Age (yr) range = 11 to 77 Con: Usual care (4 wk), n = 23 Gender = 84% F Lund (2006)42 Sweden (primary care) Exp: Massage therapy (5 wk), High risk of bias loss to follow-up = 28 n = 62 (41 completed) Chronic low back pain, n = 24 Age (yr) mean = 40 Con: Mental relaxation (5 wk), Gender = 52% F n = 55 (48 completed) United States (primary care) Exp: Massage therapy incl. Chronic neck pain, Swedish and clinical techniques n = 177 (2/wk x 5 wk), n = 12 Age (yr) mean, range = 52, 18 to 85 Gender = 62% F Con: Progressive relaxation incl. Germany muscle relaxation exercises loss to follow-up = 14 (2/wk x 5 wk), n = 12 Low back pain, n = 579 Exp1: Massage (3 wk), Age (yr) mean = 46 n = 60 (57 completed) Gender = NS England (primary care) Exp2: Acupuncture (3 wk), n = 56 (49 completed) Fibromyalgia, n = 19 Age (yr) mean = 51 Con: Sham laser (3 wk), Gender = 100% F n = 61 (57 completed) Sweden (secondary care) Exp 1: Massage (1/wk x 6 wk), n = 75 Exp 2: Massage (1/wk x 6 wk) + exercises, n = 72 Exp 3: Alexander technique (6 lessons in 4 wk), n = 144 Exp 4: Alexander technique (24 lessons in 9 mth), n = 144 Exp 5: Exercise prescription and behavioural counselling, n = 72 Con: Control (normal care), n = 72 Exp: Massage, incl. effleurage, petrissage, friction and shaking (30 min x 2/wk x 6 wk), n = 10 Con: Guided relaxation (30 min x 2/wk x 6 wk), n = 9

Outcomes Results Effect size 110 Bervoets et al: Massage for musculoskeletal disorders Pain (VAS) mean (SD) or n/N (%) SMD (95% CI) unless noted 5 wk 5 wk Pain (VAS) Pain = Exp: 37 (29); Pain = Exp minus Function (grip strength) Con: 63 (30) Con: –0.8 (–1.8 to 0.08) Pain (VAS) Function (grip strength) 4 wk 5 wk Pain (VAS) Pain = Exp: 10; Pain = Exp minus Function (grip strength) Con: 53 Con: 0.3 (–0.1 to 0.7) Pain (muscle pain scale) Grip strength = Exp: 9.0; Con: 6.0 12 wk Pain (VAS) Pain = Exp minus 4 wk Con: –0.2 (–0.6 to 0.2) Pain (VAS) Pain = Exp: 13; Con: 28 5 wk Pain (days with pain Grip strength = Exp: 9.4; Pain = Exp minus during the past 4 wk) Con: 6.1 Con: –0.5 (–1.3 to 0.4) Function (RDQ) 4 wk 4 wk Pain Pain = Exp: 13; Pain = Exp1 minus Con: 28 Con: 0.3 (–0.04 to 0.7); Grip strength = Exp: 8.5; Exp 1 minus Con: 6.7 Exp 2: 0.6 (0.2 to 1. 0) 5 wk 12 wk Pain = Exp: 60 (25); Pain = Exp 1 minus Con: 52 (30) Con: 0.1 (–0.3 to 0.5); Exp 1 minus 12 wk Exp 2: 0.1 (–0.3 to 0.5) Pain = Exp: 45 (27); Con: 50 (28) 5 wk Pain = Exp: 17 (23); Con: 29 (28) 4 wk (change) Pain = Exp 1: –8 (28); Exp 2: –24 (28); Con: –17 (28) 12 wk (change) Pain = Exp 1: –14 (32); Exp 2: –17 (30); Con: –17 (26) No usable data No usable data

Mackawan (2007)35 Chronic low back pain, n = 67 Exp: Thai massage (1 session), n = 35 High risk of bias Age (yr) range = 20 to 60 Con: Joint mobilisation (1 session), Gender = 61% F Melancon (2005)25 Thailand (general population) n = 32 High risk of bias Low back pain, n = 60 Exp: Massage therapy, n = 30 Perlman (2006)11 Age (yr) mean = 38 Con: Usual care, eg, muscle relaxants, Cross-over trial Gender = 45% F High risk of bias United States (primary care) NSAIDS, n = 30 Perlman (2012)30 Knee OA, n = 68 Exp: Swedish massage therapy Low risk of bias Age (yr) mean = 68 (8 wk), n = 34 Gender = 78% F United States (primary care, Con: WLC (8 wk), n = 34 specialised care) Exp 1: Swedish massage (30 min x 1/wk x 8 wk), n = 25 Knee OA, n = 125 Age (yr) mean = 64 Exp 2: Swedish massage (30 min x Gender = 70% F 2/wk x 8 wk), n = 25 United States (mixed population) loss to follow-up = 6 Exp 3: Swedish massage (60 min x 1/wk x 8 wk), n = 25 Exp 4: Swedish massage (60 min x 2/wk x 8 wk), n = 25 Con: Usual care (current treatment), n = 25 Pope (1994)31 Subacute low back pain, n = 164 Exp 1: Swedish massage (3 wk), High risk of bias Age (yr) mean = 32 n = 37 Gender = 62% F Sherman (2009)43 United States (secondary care) Exp 2: Manipulation (3 wk), Low risk of bias loss to follow-up = 20 n = 70 Stasinopoulos (2004)44 Chronic neck pain, n = 64 Con 1: Transcutaneous muscle High risk of bias Age (yr) mean, range = 69, 20 to 64 stimulation (3 wk), n = 28 Gender = 69% F United States (primary care) Con 2: Corset (3 wk), n = 29 loss to follow-up = 6 Exp: Massage incl. Swedish and Chronic patellar tendinopathy, n = 30 clinical techniques (10 wk), Age (yr) range = 21 to 33 n = 32 Gender = 40% F Greece (secondary care) Con: Self-care book, n = 32 Exp 1: Transverse friction (4 wk), n = 10 Exp 2: Exercise programme (4 wk), n = 10 Con: Pulsed ultrasound (4 wk), n = 10

Pain (VAS) Post session Post session Pain = Exp: 25 (18); Pain = Exp minus Pain (VAS) Con: 34 (17) Con: –0.55 (–1.03 to –0.06) Disability (ODI) No usable data Pain (VAS) 8 wk (change) 8 wk Research Pain, stiffness, and physical Pain = Exp: –23 (26); Pain = Exp minus functional disability Con: –2 (21) Con: –0.9 (–1.4 to –0.4) (WOMAC) WOMAC-function = WOMAC-function = Exp minus Exp: –20 (22.5); Con: –0.7 (–1.2 to –0.3) Pain (VAS) Con: –5.2 (16.4) WOMAC-global = Exp minus Pain, stiffness, and physical WOMAC-global = Con: –0.9 (–1.4 to –0.3) functional disability: Exp: –21.2 (22.5); (WOMAC) Con: –4.6 (15.9) 8 wk Pain = Exp 2 minus 5 8 wk (change) Con: –0.7 (–1.2 to –0.08) Pain = Exp 1: –14 (26); WOMAC-function = Pain (VAS) Exp 2: –26 (27); Exp 2 minus Exp 3: –40 (21); Con: –0.5 (–1.1 to 0.04) Exp 4: –31 (21); WOMAC-global = Con: –10 (22) Exp 2 minus WOMAC = Exp 1: –17.4 (19.0); Con: –0.6 (–1.2 to –0.04) Exp 2: –18.4 (22.9); Exp 3: –24.0 (20.2); 24 wk Exp 4: –24.0 (22.2); Pain = Exp 2 minus Con: –6.3 (16.1) Con: –0.1 (–0.5 to 0.7) WOMAC-function = 24 wk (change) Exp 2 minus Pain = Exp 1: –14 (28); Con: –0.1 (–0.9 to 0.7) Exp 2: –14 (27); WOMAC-global = Exp 3: –19 (26); Exp 2 minus Exp 4: –23 (32); Con: –0.05 (–0.6 to 0.5) Con: –12 (24) WOMAC = Exp 1: –14.3 (20.6); 3 wk Exp 2: –7.0 (21.5); Pain = Exp 1 minus Exp 3: –14.2 (23); Con 2: –0.05 (–0.6 to 0.5); Exp 4: –15.1 (25.5); Exp1 minus Exp 2: Con: –6.0 (16.4) 0.3 (–0.2 to 0.7) 3 wk (change score) Pain = Exp 1: –17 (25); Exp 2: –24 (27); Con 1: –10 (30); Con 2: –16 (27) Function (improvement 10 wk 10 wk of  10% on NDI) NDI = Exp: 12/31 (39%); NDI = Exp versus Con: 4/28 (14%) Con: RR 2.6 (0.95 to 7.2) Pain (‘no pain or much better’ versus ‘slightly better, same 26 wk 26 wk or worse’) NDI = Exp: 17/30 (57%); NDI = Exp versus Con: 9/28 (31%) Con: RR 1.7 (0.9 to 3.1) 4 wk 4 wk Pain = Exp 1: 2/10 (20%); Pain = Exp 1 versus Exp 2: 8/10 (80%); Con: RR 2.0 (0.21 to 18.69); Con: 1/10 (10%) Exp 1 versus Exp 2: RR 0.25 (0.07 to 0.90) 16 wk Pain = Exp: 2/10 (20%); 16 wk Exp 2: 10/10 (100%); Pain = Exp 1 versus Con: 0/10 (0%) Con: ARR 0.80 (0.38 to 0.94); Exp 1 versus Exp 2: RR 0.20 (0.06 to 0.69) 111

Effect size112 Bervoets et al: Massage for musculoskeletal disorders A negative effect estimate means the effect is in favour of massage. SMD (95% CI) unless noted AIMS = Arthritis Impact Measurement Scales (0 to 10), ARR = absolute risk reduction, CI = confidence interval, Con = control group, Exp = experimental group, F = female, FLEX-SF = self-reported Flexilevel, NDI = Neck Disability Index (0 to 100), NMQ = Nordic 2 wk In the short term, the pooled estimate from the three studies Musculoskeletal Questionnaire, NS = not specified, OA = osteoarthritis, ODI = Oswestry Disability Index (0 to 50), PSFDM = Patient Specific Functional Disability Measure (0 to 10), RDQ = Roland Disability Questionnaire (0 to 23), RR = relative risk, SFMPQ = Short Form McGill Pain VAS = Exp minus involving participants with shoulder pain and with a low risk of Pain Questionnaire (0 to 15), SMD = standardised mean difference, VAS = visual analogue scale (0 to 100), WLC = waiting-list control, WOMAC = Western Ontario and McMaster Universities Osteoarthritis Index. Con: –0.8 (–1.6 to –0.05) bias was a mean difference of –16 mm (95% CI –25 to –7, Disability = Exp minus I2 = 15%).28,29,32 Both studies involving participants with low back Con: –1.0 (–1.8 to –0.2) pain and with a low risk of bias27,31 showed no benefit of massage 4 wk over no treatment, but heterogeneity was considerable (MD = –12, Function = Exp minus 95% CI –32 to 8, I2 = 81%). Two studies included participants with Con: –1.9 (–2.5 to –1.2) osteoarthritis of the knee,11,30 of which one had low risk of bias, with a pooled estimate of MD –19 (95% CI –28 to –10, I2 = 0%). One Results study involving participants with neck pain and with a low risk of mean (SD) or n/N (%) bias24 showed no statistically significant benefits (Figure 3). For a 2 wk detailed forest plot, see Figure 4 on the eAddenda. The effect Pain (VAS) = Exp: 32 (26); estimates of 16 to 19 mm on the VAS were considered to be Con: 54 (26) clinically relevant. Disability = Exp: 17.6 (8.0); Con: 10.4 (5.6) It was concluded that, in the short term, there is moderate-level No data evidence that massage reduces pain compared to no treatment in 4 wk people with shoulder pain but not in those with low back pain Function = Exp: 40.5 (5.2); (both downgraded by imprecision). Furthermore, there is low-level Con: 31.7 (3.8) evidence that massage reduces pain compared to no treatment in people with osteoarthritis of the knee (downgraded by imprecision Outcomes and design) but is ineffective compared to no treatment in those Pain (VAS) with neck pain (single study). Disability (PSFDM) Pain (VAS) Function Function (FLEX-SF) In the studies on shoulder pain,32,33 function was measured Intervention with different instruments; therefore, pooled results are presented Exp: Soft tissue massage (2 wk), n = 15 as standardised mean differences. In the studies on low back Con: WLC (2 wk), n = 14 pain,26,34 function was measured using the Roland Disability Exp: Swedish massage (5 wk), n = 19 Questionnaire (RDQ), which ranges from 0 to 24. In the studies on Con: Usual care incl. pharmacological osteoarthritis of the knee,11,30 function was measured with the Western Ontario and McMaster Universities Arthritis Index therapy, advice on exercise and (WOMAC), which also ranges from 0 to 24. posture (5 wk), n = 10 Exp: Soft tissue massage (4 wk), n = 30 Six studies provided data on short-term function,11,26,30,32–34 of Con: Light hand touch (4 wk), n = 30 which, one had high risk of bias.11 Three studies11,26,34 presented long-term data varying from 12 to 52 weeks (Table 1). Participants Shoulder pain, n = 29 In the short term, the pooled estimate of the two studies with Age (yr) mean, range = 68, 18 to 80 low risk of bias involving participants with shoulder pain was a Gender = 69% F standardised mean difference of –1.5 (95% CI –2.3 to –0.6, I2 = 63%), Australia (secondary care) with substantial heterogeneity (Figure 5).32,33 For a detailed forest Chronic musculoskeletal pain, plot, see Figure 6 on the eAddenda. An effect estimate of 1.5 is n = 29 regarded as a large effect size. The pooled estimate of the two Age (yr) range = 16 to 65 studies with low risk of bias involving participants with low back Gender = 90% F pain was a mean difference of –2.5 (95% CI –3.4 to –1.6, I2 = 0%), Germany (primary care) which is not considered to be a clinically relevant difference.26,34 Posterior shoulder tightness, n = 60 The pooled estimate of the two studies on osteoarthritis of the Age (yr) mean = 54 knee11,30 was a mean difference of –13.0 (95% CI –20.1 to –5.9, Gender = 72% F I2 = 0%), which is considered to be clinically relevant (Figure 7). For Taiwan (secondary care) a detailed forest plot, see Figure 8 on the eAddenda. Loss to follow-up = 8 It was concluded that, in the short term, there is moderate-level evidence that massage improves function compared to no treatment in people with low back pain (downgraded by imprecision). Furthermore, there is low-level evidence that massage improves function compared to no treatment in people with shoulder pain (downgraded by inconsistency and impreci- sion) and those with osteoarthritis of the knee (downgraded by design and imprecision). Massage versus active treatments Pain All studies except one23 measured pain using a VAS; therefore, data are presented as mean differences on a 0 to 100 mm scale. Six studies provided data on short-term pain,23,24,31,35–37 of which one had low risk of bias.24 The disorders examined in the included studies were fibromyalgia,36 general pain23 and low back pain,37 which were all compared to relaxation treatment; low back pain31,35 was compared to mobilisation/manipulation and neck pain24 was compared to acupuncture. Because the studies differed in the musculoskeletal disorder, the comparative treatment, or both, studies were not pooled (Figure 9). For a detailed forest plot, Table 1 (Continued ) Study Van den Dolder (2003)32 Low risk of bias Walach (2003)45 High risk of bias Yang JL (2012)33 Low risk of bias

]GIF$DT)[(igure_2 Researc]GIF$DT)3_erugi([h 113 Random sequence generation Subgroup MD (95% CI) Allocation concealment Study Random Blinding of participants Blinding of personnel Shoulder pain Blinding of outcome assessors Buttagat29 Drop-out rate Buttagat28 Intention-to-treat analysis Van den Dolder32 Selective reporting Subtotal Group similarity at baseline Co-interventions Low back pain Compliance Buttagat27 Timing of outcome assessments Pope31 Low risk of bias Subtotal Unclear risk of bias High risk of bias Knee osteoarthritis Perlman11 Alnigenis (2001)38 Perlman30 Atkins (2013)39 Subtotal Buttagat (2011)27 Neck pain Buttagat (2012)29 Irnich24 Buttagat (2012)28 Subtotal Cherkin (2001)34 Cherkin (2011)26 –50 –25 0 25 50 Favours massage (mm) Favours control Field (2002)36 Field (2004)21 Figure 3. Weighted mean differences (95% CI) in the effect of massage versus Field (2007)40 control on pain measured on a visual analogue scale (0 to 100 mm), pooling data Field (2011)22 from three trials on shoulder pain (n = 89), two trials on low back pain (n = 93), two Hasson (2004)23 trials on osteoarthritis of the knee (n = 118), and one trial on neck pain (n = 114). Hernandez-Reif (2001)37 Irnich (2001)24 neck pain or fibromyalgia. Only the short-term data (ie, post Little (2008)41 treatment up to 12 weeks) could be pooled. Control interventions Lund (2006)42 were conservative treatments such as relaxation, manipulation, Mackawan (2007)35 joint mobilisation and acupuncture. These findings suggest that Melancon (2005)25 massage might be a viable treatment option for several musculo- Perlman (2006)11 skeletal disorders. Perlman (2012)30 Pope (1994)31 This systematic review was limited to clearly defined massage Sherman (2009)43 techniques as stand-alone treatments without joint manipulation Stasinopoulos (2004)44 or mobilisation techniques, as defined by others.7,8 This definition Van den Dolder (2003)32 of massage most closely mirrors the massage treatment that is Walach (2003)45 provided by physiotherapists. The Swedish technique predomi- Yang (2012)33 nates in clinical settings and is comparable with the frequently used technique in physiotherapy practice. One of the challenges in Figure 2. Risk of bias of the included studies assessed using the Cochrane Back conducting a systematic review in the field of massage therapy is Review Group tool. tFIG](igure_5)$TD[ he lack of consistent terminology. Therefore, a standardised see Figure 10 on the eAddenda. Two studies23,24 provided long- term data (12 weeks), as shown in Table 1. Study SMD (95% CI) Shoulder pain Random It was concluded that there is low-level evidence (single study) that acupuncture reduces pain more than massage in people with Van den Dolder32 neck pain. Furthermore, there is very-low-level evidence (single Yang33 study, downgraded by design) that massage reduces pain more than joint mobilisation in people with low back pain, but that there Subtotal is no benefit of massage over manipulation or relaxation therapy in those with fibromyalgia, low back pain and musculoskeletal pain. –2 –1 0 2 2 Favours massage Favours control Function Two studies with low risk of bias presented data on function, Figure 5. Standardised mean differences (SMD) (95% CI) in the effect of massage versus control on shoulder function, pooling data from two trials on shoulder pain which was measured with the RDQ in the short and long term (n = 81). (52 weeks) for people with low back pain, where massage was compared to relaxation treatment26 or acupuncture (Figure 11).34 For a detailed forest plot, see Figure 12 on the eAddenda. There is low-level evidence (two single studies) that massage does not improve function more than acupuncture or relaxation in people with low back in the short term; in the long term, relaxation seems superior to massage. Discussion Overall, low to moderate levels of evidence were found for the benefits of massage over no treatment. Furthermore, low to very low levels of evidence were found for the lack of benefit of massage over other active treatments. The participants in the included studies had low back pain, shoulder pain, osteoarthritis of the knee,

1]GIF$DT)7_erugi([ 14 [igure_9)D$FIG]T(Bervoets et al: Massage for musculoskeletal disorders Subgroup MD (95% CI) Subgroup MD (95% CI) Study Random Study Random Low back pain Versus relaxation Cherkin34 Field36 Cherkin26 Hasson23 Subtotal Hernandez-Reif37 Knee osteoarthritis Versus mob/manip Perlman11 Mackawan35 Perlman30 Pope31 Subtotal Versus acupuncture Irnich24 –20 –10 0 10 20 Favours massage (points) Favours control –50 –25 0 25 50 Favours other Figure 7. Weighted mean differences (MD) (95% CI) in the effect of massage versus Favours massage (mm) control on function measured on 0-to-24 point scales, pooling data from two trials on low back pain (n = 410) using the Roland Disability Questionnaire (RDQ) and two Figure 9. Weighted mean differences (MD) (95% CI) in the effect of massage versus trials on osteoarthritis of the knee (n = 118) using the Western Ontario and other active treatments on pain measured on a visual analogue scale (0 to 100 mm), McMaster Universities Arthritis Index (WOMAC). presenting data from three trials on massage versus relaxation, two trials on massage versus mobilisation and/or manipulation (mob/manip), and one trial on massage versus acupuncture. taxonomy for massage is warranted. This will enable researchers The differences between these results and those of other and massage therapists to more clearly communicate about the reviews can be attributed to the different definitions of massage nature of massage treatment and its effectiveness. Furthermore, it techniques and the fact that studies were only included in this was found that studies only reported the frequency of the review if they evaluated massage as a stand-alone treatment intervention; treatment parameters should also include the type instead of a combined intervention.5,7,8,10 This review found of massage, the duration of a massage session, and the intensity or clearer benefit of massage versus no treatment compared to the grade/depth of pressure. One study did indicate that there might be previous reviews. a dose-response relationship.30 The most recent Cochrane review on massage for people with The evidence identified by this review indicates that the effect neck pain found that the effect of massage remains uncertain.5 This of massage can best be found immediately after treatment and is in accordance with the present results on people with neck pain. when compared to no treatment. Few studies included a follow-up The Cochrane review included a wide spectrum of massage- measurement. Three studies even assessed the outcomes immedi- therapy studies, including muscle stretching techniques and joint ately after a single massage session. In a meta-analysis, the authors mobilisation. Just two of their studies were included in this review. challenged the assumption that biological effects of massage only occur immediately.13 The present review found statistically A weakness of this review is that the primary studies had some significant pain reduction immediately after a single massage limitations. The majority of studies (16 out of 26) had high risk of session in two of these studies. This biological paradigm needs to bias. Because massage often precluded blinding of participants or be further evaluated. In addition, this review found no significant care providers, the majority of the studies were rated ‘unclear’ or trend based on the duration of a massage session.13 Neither the ‘high risk of bias’ in relation to blinding. Blinding of participants, optimal frequency nor duration of massage treatments for pain care providers and outcome assessors was often not achieved. reduction and the ‘decay’ in analgesic effect on pain is known. Consequently, not being able to blind trial participants may lead to Future massage trials will benefit when these parameters are exaggerated treatment-effect estimates. Another limitation was the sample size. Fifteen studies recruited fewer than 25 partici- e$DT)1_erugF([I]Gi stablished. pants per arm and the smallest study arm included eight Subgroup MD (95% CI) Study Random Versus relaxation Cherkin26 Versus acupuncture Cherkin34 –4 –2 0 24 Favours other Favours massage (points) Figure 11. Weighted mean differences (MD) (95% CI) in the effect of massage versus other active treatments on function of low back pain measured on the 0-to-24 point Roland Disability Questionnaire (RDQ), presenting data from two trials.

Research 115 participants. This means that any clinically relevant effects may 8. Brosseau L, Wells GA, Tugwell P, Casimiro L, Novikov M, Loew L, et al. Ottawa Panel not be detected as statistically significant effects. evidence-based clinical practice guidelines on therapeutic massage for neck pain. J Bodyw Mov Ther. 2012;16:300–325. A strength of the review is that it evaluated massage techniques using a strict definition. The definition allowed 9. Dryden T, Baskwill A, Preyde M. Massage therapy for the orthopaedic patient: a selection of those studies relevant to physiotherapists. Further- review. Orthop Nurs. 2004;23:327–332. more, a comprehensive search of multiple databases was conducted, and bias was minimised and internal validity was 10. Netchanok S, Wendy M, Marie C. The effectiveness of Swedish massage and improved by the rigorous methodology. The external validity traditional Thai massage in treating chronic low back pain: a review of the depends on the broad variation of participant populations and literature. Complement Ther Clin Pract. 2012;18:227–334. massage techniques included. 11. Perlman AI, Sabina A, Williams A. Massage therapy for osteoarthritis of the knee: a With regard to implications for clinicians arising from this randomized controlled trial. Arch Intern Med. 2006;166:2533–2538. review, the results show that massage as a stand-alone treatment reduces pain and improves function. The results – especially the 12. Weerapong P, Hume P, Kolt G. The mechanisms of massage and effects on ones where massage is compared to no treatment – are clinically performance, muscle recovery and injury prevention. Sports Med. 2005;35: relevant, mainly in people with low back pain, shoulder pain or 235–256. osteoarthritis of the knee. In other patient populations, the effectiveness of massage is hardly evaluated. 13. Moyer CA, Rounds J, Hannum JW. A meta-analysis of massage therapy research. Psychol Bull. 2004;130:3–18. With regard to implications for further research, this review highlights the need for better and larger studies of the efficacy and 14. Field TM. Massage therapy effects. Am Psychol. 1998;53:1270–1281. acceptability of massage therapies in other patient populations. 15. Wang Q, Zeng H, Best TM. A mechatronic system for quantitative application and The promising findings need further confirmation. Future studies of massage therapy need to report greater detail about the assessment of massage-like actions in small animals. Ann Biomed Eng. 2014;42: intervention beyond its frequency (sessions per week), and should 36–49. include the type of massage, the duration of a massage session, and 16. Sagar SM, Dryden T, Wong RK. Massage therapy for cancer patients: a reciprocal the intensity or grade/depth of pressure. Furthermore, a standar- relationship between body and mind. Curr Oncol. 2007;14:45–56. dised taxonomy for massage is warranted. This would significantly 17. van den Dolder PA, Ferreira PH, Refshauge KM. Effectiveness of soft tissue massage assist researchers in selecting appropriate techniques and inter- and exercise for the treatment of non-specific shoulder pain: a systematic review preting the results of massage studies. with meta-analysis. Br J Sports Med. 2014;48:1216–1226. 18. Higgins JPT, Green S, eds. In: Cochrane Handbook for Systematic Reviews of What is already known on this topic: Musculoskeletal Interventions Version 5.1.0 [updated 2011 September]. London: The Cochrane disorders cause pain and disability in a substantial proportion Collaboration; 2009. Available from: www.cochrane-handbook.org. Accessed of the population. Existing systematic reviews of massage for April 29, 2015. these disorders do not include several relevant trials and some 19. Higgins JP, Thompson SG. Quantifying heterogeneity in a meta-analysis. Stat Med. do not distinguish the effect of massage from other manual 2002;21:1539–1558. therapies. 20. Guyatt GH, Oxman AD, Vist GE. GRADE Working Group. GRADE: an emerging What this study adds: Massage reduces pain, in the short consensus on rating quality of evidence and strength of recommendations. BMJ. term, in shoulder pain and osteoarthritis of the knee. Massage 2008;336:924–926. improves function, in the short term, in shoulder pain, low back 21. Field T, Diego M, Cullen C. Carpal tunnel syndrome symptoms are lessened pain and osteoarthritis of the knee. Massage is not clearly more following massage therapy. J Bodyw Mov Ther. 2004;8:9–14. or less beneficial that other commonly used treatments for 22. Field T, Diego M, Delgado J. Hand pain is reduced by massage therapy. Complement musculoskeletal disorders. Ther Clin Prac. 2011;17:226–229. 23. Hasson D, Arnetz B, Jelveus L. A randomized clinical trial of the treatment effects of eAddenda: Figures 4, 6, 8, 10 and 12 and Appendix 1 can be massage compared to relaxation tape recordings on diffuse long-term pain. found online at doi:10.1016/j.jphys.2015.05.018. Psychother Psychosom. 2004;73:17–24. 24. Irnich D, Behrens N, Molzen H. Randomised trial of acupuncture compared with Ethics approval: Not applicable. conventional massage and ‘sham’ laser acupuncture for treatment of chronic neck Competing interests: Nil. pain. BMJ. 2001;322:1–6. Source(s) of support: Nil. 25. Melancon B, Miller L. Massage therapy versus traditional therapy for low back pain Acknowledgements: Nil. relief implications for holistic nursing practice. Holist Nurs Pract. 2005;19: Provenance: Not invited. Peer-reviewed. 116–121. Correspondence: Arianne P Verhagen, Department of General 26. Cherkin D, Sherman K. A comparison of the effects of 2 types of massage and usual Practice, Erasmus MC, University Medical Center Rotterdam, care on chronic low back pain. Ann Intern Med. 2011;155:1–9. Rotterdam, The Netherlands. Email: [email protected] 27. Buttagat B, Eungpinichpong W, Chatchawan U. The immediate effects of traditional Thai massage on heart rate variability and stress-related parameters in patients References with back pain associated with myofascial trigger points. J Bodyw Mov Ther. 2011;15:15–23. 1. Picavet HS, Hazes JM. Prevalence of self-reported musculoskeletal diseases is high. 28. Buttagat V, Eungpinichpong W, Kaber D. Acute effects of traditional Thai massage Ann Rheum Dis. 2003;62:644–650. on electroencephalogram in patients with scapulocostal syndrome. Complement Ther Med. 2012;20:167–174. 2. Hoy DG, Smith E, Cross M, Sanchez-Riera L, Blyth FM, Buchbinder R, et al. Reflecting 29. Buttagat B, Eungpinichpong W, Chatchawan U. Therapeutic effects of traditional on the global burden of musculoskeletal conditions: lessons learnt from the Global Thai massage on pain, muscle tension and anxiety in patients with scapulocostal Burden of Disease 2010 Study and the next steps forward. Ann Rheum Dis. syndrome: a randomized single-blinded pilot study. J Bodyw Mov Ther. 2015;74:4–7. 2012;16:57–63. 30. Perlman AI, Ali A, Njike VY, Hom D. Massage therapy for osteoarthritis of the knee: 3. Brummitt J. The role of massage in sports performance and rehabilitation: current a randomized dose-finding trial. PLoS One. 2012;7:e30248. evidence and future direction. Am J Sports Phys Ther. 2008;3:7–21. 31. Pope M, Phillips R, Haugh L. A prospective randomized three-week trial of spinal manipulation, transcutaneous muscle stimulation, massage and corset in the 4. Kong LJ, Zhan HS, Cheng YW. Massage therapy for neck and shoulder pain: a treatment of subacute low back pain. Spine. 1994;19:2571–2577. systematic review and meta-analysis. Evid Based Complement Alternat Med. 32. van den Dolder P, Roberts D. A trial into the effectiveness of soft tissue massage in 2013;2013:613279. the treatment of shoulder pain. Aust J Physiother. 2003;49:183–188. 33. Yang JL, Chen SY, Hsieh CL, Lin JJ. Effects and predictors of shoulder muscle massage 5. Patel KC, Gross A, Graham N, Goldsmith CH, Ezzo J, Morien A, et al. Massage for for patients with posterior shoulder tightness. BMC Musculoskelet Disord. mechanical neck disorders. Cochrane Datab Syst Rev. 2012;9:CD004871. 2012;13:46. 34. Cherkin D, Eisenberg D, Sherman K. Randomised trial comparing Traditional 6. Karels C, Polling W, Bierma-Zeinstra S. Treatment of arm, neck, and/or shoulder Chinese Medical Acupuncture, Therapeutic massage, and Self-care educational complaints in physical therapy practice. Spine. 2006;31:E584–E589. for chronic low back pain. Arch Intern Med. 2001;161:1081–1088. 35. Mackawan S, Eungpinichpong W, Pantumethakul R. Effects of traditional Thai 7. Brosseau L, Wells GA, Poitras S, Tugwell P, Casimiro L, Novikov M, et al. Ottawa massage versus joint mobilization on substance P and pain perception in patients Panel evidence-based clinical practice guidelines on therapeutic massage for low with non-specific low back pain. J Bodyw Mov Ther. 2007;11:9–16. back pain. J Bodyw Mov Ther. 2012;16:424–455. 36. Field T, Diego M, Cullen C. Fibromyalgia pain and Substance P decrease and sleep improves after massage therapy. J Clin Rheumatol. 2002;8:72–76. 37. Hernandez-Reif M, Field T, Krasnegor J. Lower back pain is reduced and range of motion increased after massage therapy. Int J Neurosci. 2001;106:131–145. 38. Alnigenis M, Bradley J, Wallick J. Massage therapy in the management of fibromy- algia: a pilot study. J Musculoskel Pain. 2001;9:55–67. 39. Atkins D, Eichler D. The effects of self-massage on osteoarthritis of the knee: a randomized, controlled trial. Int J Ther Mass Bodyw. 2013;6:4–14. 40. Field T, Diego M, Hernandez-Reif M. Hand arthritis pain is reduced by massage therapy. J Bodyw Mov Ther. 2007;11:21–24. 41. Little P, Lewith G, Webley F. Randomised controlled trial of Alexander technique lessons, exercise, and massage (ATEAM) for chronic and recurrent back pain. BMJ. 2008;337:a884. 42. Lund I, Lundeberg T, Carleson J. Corticotropin releasing factor in urine—a possible biochemical marker of fibromyalgia responses to massage and guided relaxation. Neurosci Lett. 2006;403:166–171.

116 Bervoets et al: Massage for musculoskeletal disorders 43. Sherman K, Cherkin D, Hawkes R. Randomized trial of therapeutic massage for 46. Wood M. The new International Classification of Primary Care: genesis and chronic neck pain. Clin J Pain. 2009;25:233–238. implications for patient care and research. J Fam Pract. 1987;24:569–571. 44. Stasinopoulos D, Stasinopoulos I. Comparison of effects of exercise programme, Further reading pulsed ultrasound and transverse friction in the treatment of chronic patellar tendinopathy. Clin Rehabil. 2004;18:347–352. back.cochrane.org 45. Walach H, Gu¨ thlin C, Ko¨ nig M. Efficacy of massage therapy in chronic pain: a pragmatic randomized trial. J Altern Complement Med. 2003;9:837–846.

Journal of Physiotherapy 61 (2015) 168 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Media New South Wales Agency for Clinical Innovation - Pain Management Network website: a useful resource for patients and clinicians Agency for Clinical Innovation Pain Management Network: track. The design of this planning sheet very much promotes self- http://www.aci.health.nsw.gov.au/chronic-pain determination and self-management. The Australian National Pain Strategy1 highlights the problems The For Youth: PainBytes section also consists of seven episodes of chronic pain in Australia. Among others, the key goals of the and is well presented with shorter videos and tutorials that are National Pain Strategy include empowering and supporting people laid out in more detail than the adult section. It’s a refreshing with pain and ensuring that professionals are skilled and utilise change to see this information being made available for younger best-practice evidence-based care. [3]DF_FTI$ urthermore, access to inter- people and I almost found the IF_]DT$6[children’s stories more engaging disciplinary care at all levels is a priority. As the management of than the adults’ ones. chronic pain is complex and multi-faceted, it requires additional resources to support patient care. With increasing demands on The Spinal Cord Injury[7_T]FD$I Pain section, with eight episodes, includes health resources and issues with access to appropriate healthcare, topics that are specific to pain following spinal cord injury, such as resources such as the Pain Management Network website, neuroplasticity and neuropathic pain, as well as basic information designed and compiled by the New South Wales (NSW) Agency about pain covered in other sections. I found the pain, physical for Clinical Innovation (an Australian state government initiative) - activity and exercise episode inspiring, particularly hearing $DFT[]oI_8 ne Pain Management group, are much needed and welcome, and align woman’s journey back to farming, skiing and kayaking following well with the National Pain Strategy. her spinal cord injury. The website is a useful resource for clinicians and people On the downside, the videos, particularly in the adult section, suffering from chronic pain. It is structured into four sections: For are somewhat repetitive, both in tone and format, which may limit Everyone, For Youth: PainBytes (for younger sufferers), Spinal Cord engagement with the material. Some are also a little long and Injury Pain and Health Professionals. The latter contains a engagement may depend on the individual’s tolerance of the comprehensive bank of resources for healthcare professionals – longer (10 to 15 minute) videos. The footage in the For EveryoneFI]D[9_T$ from copies of questionnaires to educational resources. This section primarily focuses around two older people, which might be section also provides a [F$D]_T4Ihealth ]p5F$D[T_I lan pro-forma, which is directed at less appealing to a younger, more active or ambitious audience. GPs but useful for any primary care practitioner. For each pain- Some of the videos included schematics, which were helpful and related domain (eg, pain intensity or psychological factors), a cFT][_$01ID omplemented the rest of the video footage. However, I think that number of validated questionnaires are provided. A simple click on this site could be enhanced by more creative presentation of the each link provides brief information about when to use this material, better use of graphics and less reliance on interviews and questionnaire, as well as the actual questionnaire and scoring lectures alone. information. There is a lot of information on this site, so for those not familiar with many of these questionnaires, it may appear With the growing number of pain websites available, one overwhelming and further guidance may be useful. wonders whether this mass of information would be overwhelm- ing for some people. Perhaps clearer guidelines on how to use the The main body of the site (For Everyone) consists of a series of information, and the site being more than a series of tutorials (eg, seven short videos (episodes) of 10 to 15 minutes, on various having more action buttons) would improve the utility of the aspects of pain (eg, sleep and pain, medications and pain, physical website for people and clinicians. activity and pain), the latter being presented by a physiotherapist. The first video, An Introduction to Pain, which is presented by pain Nonetheless, the Agency for Clinical Innovation is to be specialist Associate Professor Phillip Siddall, includes some commended for compiling such a useful resource. I hope that excellent graphics that are useful to help people understand the many people and clinicians will benefit from the information, differences between acute and chronic pain and the influence of leading to better outcomesF_].$[1DTI factors such as stress, activity and sleep. The information is very clear and the use of videos is helpful. Each section is supported by Provenance: Invited. Not peer-reviewed. an action plan, links to additional useful resources, many succinct summaries on each subject and links to videos (eg, a YouTube video Niamh Moloney on neuroplasticity). Consumers can download a health plan that Department of Health Professions, Macquarie University,]FID$_1[T Sydney, may help them to prioritise their goals, identify who can help them to achieve their goals and what strategies will help to keep them on Australia Reference 1. National Pain Strategy: Pain Management for all Australians. PainAustralia 2010. www.painaustralia.org.au. http://dx.doi.org/10.1016/j.jphys.2015.05.009 1836-9553/ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 61 (2015) 142–147 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Research Non-invasive ventilation used as an adjunct to airway clearance treatments improves lung function during an acute exacerbation of cystic fibrosis: a randomised trial Tiffany J Dwyer a,b, Lisel Robbins c, Patrick Kelly d, Amanda J Piper a,e, Scott C Bell c,f, Peter T P Bye a,b a Department of Respiratory and Sleep Medicine, Royal Prince Alfred Hospital; b Sydney Medical SchoolD[F_T$I3], University of Sydney; c Adult Cystic Fibrosis Centre, The Prince Charles Hospital; d Sydney School of Public Health, University of Sydney; e Woolcock Institute of Medical Research, Sydney; f QIMR Berghofer Medical Research Institute, Brisbane, Australia KEY WORDS ABSTRACT Cystic fibrosis Question: During an acute exacerbation of cystic fibrosis, is non-invasive ventilation beneficial as an Non-invasive ventilation adjunct to the airway clearance regimen? Design: Randomised controlled trial with concealed allocation Respiratory therapy and intention-to-treat analysis. Participants: Forty adults with moderate to severe cystic fibrosis lung Physical Therapy disease and who were admitted to hospital for an acute exacerbation. Intervention: Comprehensive inpatient care (control group) compared to the same care with the addition of non-invasive ventilation during airway clearance treatments from Day 2 of admission until discharge (experimental group). Outcome measures: Lung function and subjective symptom severity were measured daily. Fatigue was measured at admission and discharge on the Schwartz Fatigue Scale from 7 (no fatigue) to 63 (worst fatigue) points. Quality of life and exercise capacity were also measured at admission and discharge. Length of admission and time to next hospital admission were recorded. Results: Analysed as the primary outcome, the experimental group had a greater rate of improvement in forced expiratory volume in 1 second (FEV1) than the control group, but this was not statistically significant (MD 0.13% predicted per day, 95% CI –0.03 to 0.28). However, the experimental group had a significantly higher FEV1 at discharge than the control group (MD 4.2% predicted, 95% CI 0.1 to 8.3). The experimental group reported significantly lower levels of fatigue on the Schwartz fatigue scale at discharge than the control group (MD 6 points, 95% CI 1 to 11). There was no significant difference between the experimental and control groups in subjective symptom severity, quality of life, exercise capacity, length of hospital admission or time to next hospital admission. Conclusion: Among people hospitalised for an acute exacerbation of cystic fibrosis, the use of non-invasive ventilation as an adjunct to the airway clearance regimen significantly improves FEV1 and fatigue. Trial registration: ANZCTR 12605000437662. [Dwyer TJ, Robbins L, Kelly P, Piper AJ, Bell SC, Bye PTP (2015) Non-invasive ventilation used as an adjunct to airway clearance treatments improves lung function during an acute exacerbation of cystic fibrosis: a randomised trial. Journal of Physiotherapy 61: 142–147] ß 2015 Published by Elsevier B.V. on behalf of Australian Physiotherapy Association. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/). Introduction sputum volume,5–7 was not different with the application of NIV when compared to resting breathing alone4 or standard chest Airway clearance is an integral component of the respiratory physiotherapy.5–7 In addition, there were no differences in lung management of cystic fibrosis (CF).1 During acute exacerbations of their lung disease, people with CF often find it difficult to maintain function, as measured by spirometry, following standard chest effective airway clearance due to increased breathlessness, physiotherapy and NIV-assisted treatment.5–8 lethargy and respiratory muscle fatigue.2 Studies have shown that the addition of non-invasive ventilation (NIV) to a single However, this does not preclude a role for NIV as an adjunct to treatment of physiotherapy techniques for airway clearance assists in overcoming these issues.3 However, it is not known whether NIV airway clearance physiotherapy for people with CF. In several is a beneficial adjunct to physiotherapy treatments for airway studies, participants reported greater ease of expectoration,5 less clearance if used throughout a hospital admission for an acute breathlessness6 and less fatigue5,7 when standard chest physio- exacerbation of CF. therapy treatment was assisted with NIV. Additionally, partici- The efficacy of NIV as an airway clearance technique in CF has been examined in several single-session crossover trials. Mucus pants reported that they preferred treatment with NIV compared clearance, as measured by inhaled radioaerosol4 and expectorated to standard chest physiotherapy.5,6,8 The improvements in breathlessness, fatigue and ease of expectoration may be at least partly explained by the preservation of respiratory muscle strength with NIV-assisted treatment compared to standard chest physio- therapy.5,6 This effect on respiratory muscle strength suggests that http://dx.doi.org/10.1016/j.jphys.2015.05.019 1836-9553/ß 2015 Published by Elsevier B.V. on behalf of Australian Physiotherapy Association. This is an open access article under the CC BY-NC-ND license (http:// creativecommons.org/licenses/by-nc-nd/4.0/).

Research 143 people with CF may be able to perform airway clearance participants performed the active cycle of breathing technique, manoeuvres with less effort or tolerate more effective airway which consists of cycles of deep breathing, relaxed breathing, clearance techniques with the support of NIV, especially when huffing and coughing in order to aid mucus clearance.11 In most required – during an acute exacerbation. addition, the physiotherapist determined whether to incorporate any additional techniques, including manual percussion, vibration, Despite the promising results of single-session applications of postural drainage positioning, autogenic drainage, positive expi- NIV during airway clearance treatment in people with CF, to date, ratory pressure (PEP) and oscillating PEP. no study has examined the efficacy of NIV over a longer period of time.3 Therefore, the research questions for this study were: Non-invasive ventilation treatment Participants assigned to the experimental group underwent an 1. In adults with moderate to severe CF lung disease and who are admitted to hospital with an acute exacerbation, does the acclimatisation session on the day of admission. During this time, addition of NIV to chest physiotherapy improve the rate of the appropriate inspiratory and expiratory pressures were deter- change and discharge values of lung function and subjective mined, and an interface (nasal mask or mouthpiece) was chosen symptom severity? that maximised participant comfort and efficacy (pressure support), whilst minimising leak. On Day 2 of admission, 2. Does it improve the change in respiratory muscle strength participants assigned to the experimental group were provided following chest physiotherapy? with a NIV machinea with inbuilt humidifierb and were instructed to use NIV during all chest physiotherapy sessions for the rest of 3. Does it improve quality of life, fatigue score, exercise capacity their hospital admission. NIV use was recorded with the inbuilt and quantitative sputum microbiology at discharge? storage cardc and downloaded at discharge from hospital. 4. Does it reduce the length of hospital admission and lengthen the time to next hospital admission for an acute exacerbation of CF? Methods Outcome measures Design Spirometry, which was the primary outcome measure, was performed daily to determine the rate of change in FEV1. Subjective A parallel-group, randomised controlled trial was conducted in symptom severity was also recorded daily for breathlessness, two tertiary Australian hospitals with specialist CF units. After expectorated sputum volume and energy levels on 10-cm visual signing consent to participate, patients were assigned to the control analogue scales (breathlessness: 0 = nothing at all, 10 = the group (standard comprehensive inpatient care from the CF team) or maximal I have ever experienced; expectorated sputum volume: the experimental group (standard care plus NIV during chest 0 = none at all, 10 = as much as I have ever had; energy: 0 = full of physiotherapy). Group allocation was determined by computer- beans, 10 = no energy at all). Maximal inspiratory (PImax) and generated block randomisation, which was stratified for hospital expiratory (PEmax) pressure12 were measured immediately before and gender. Randomisation was performed by a person not involved and after chest physiotherapy on Day 2 of admission, 1 week after in the study and stored in sealed, sequentially numbered, opaque admission and at discharge from hospital. At admission and envelopes, which were opened after the participant had signed discharge from hospital, participants also completed the CFQ (a CF- consent. The participants, treating therapists and independent specific quality of life questionnaire),13 the Schwartz fatigue assessors were not blinded to treatment group allocation. Partici- scale,14 and the 25-level modified shuttle test to assess exercise pants in the experimental group used NIV during chest physiother- capacity.15 Sputum samples were collected at admission (prior to apy from Day 2 of admission until discharge from hospital. commencement of intravenous antibiotics), 1 week after admis- Spirometry and subjective symptom severity were recorded daily sion and at discharge from hospital. Samples were couriered on ice throughout the admission. All other outcome measures were to a central laboratory for quantitative microbiological analysis, collected on admission to and discharge from hospital. which was performed by a blinded assessor. Organisms were identified with the use of standard microbiological techniques, Participants, therapists and centres including the API 20 NE systemd and quantification of pathogens was performed with the use of the modifications of Wong and All patients who were aged over 17 years and admitted with an colleagues.16 The length of admission and time to next hospital acute exacerbation of CF to the Royal Prince Alfred Hospital, admission for an acute exacerbation9 of CF were recorded. Sydney, and Prince Charles Hospital, Brisbane, were assessed for inclusion in the study. An acute exacerbation of CF was defined as Data analysis the need for intravenous antibiotics with the presence of four or more signs or symptoms, according to the criteria used by Fuchs The primary outcome measure was rate of change in FEV1 (% of et al.9 Patients with moderate to severe CF lung disease were predicted per day) from admission to discharge from hospital. Data included in the study if their forced expiratory volume in 1 second from 10 patients of similar lung function severity (five from each of (FEV1) on admission < 60% of the predicted value.10 Patients were the hospitals in this study) showed a standard deviation of 0.49% of excluded from the study ifT$[]FID_2 theyT_5F$DI[]: were using domiciliary NIV for the predicted per day. With an anticipated effect size of 0.5, treatment of respiratory failure, had precautions to the use of NIV (eg, 17 participants in each group would give 80% power to detect a pneumothorax, recent severe haemoptysis), were colonised with Burkholderia cepacia complex, were pregnant or had already significant difference (a < 0.05) between the two groups. To allow participated in the study. The physiotherapy staff rostered to the respiratory inpatient wards applied the interventions after discussion for a 15% dropout rate, we sought to recruit 40 people to the study. with a senior physiotherapist not otherwise involved in the study. Repeated-measures analyses were performed using linear Interventions mixed models to determine if the ]rIFTD[6_$ ates of improvement in FEV1 and symptom severity were different between the experimental A senior respiratory physiotherapist, who was not otherwise and control groups. The models included fixed effects for group involved in the study, individually assessed all participants and (experimental or control), time (day of admission), and the group determined the type, frequency and duration of chest physiother- by time interaction. A random effect for the intercept (ie, the apy treatment with the participant. During chest physiotherapy, all participant) was included. Independent-samples t-tests, adjusting for the participants’ admission values, were performed to compare the discharge values between the experimental and control groups for FEV1, symptom severity, quality of life, exercise capacity and quantitative sputum microbiology. Independent-samples t-tests were calculated to determine whether there were differences

[(Fig]GI$DT)1_eru144 Dwyer et al: NIV-assisted chest physiotherapy in cystic fibrosis Figure 1. Design and flow of participants through the trial. B. cepacia = Burkholderia cepacia, CF = cystic fibrosis, FEV1 = forced expiratory volume in 1 second, NIV = non-invasive ventilation. a One participant in the experimental group did not tolerate NIV and withdrew from further intervention and data collection from Day 2, so was not able to be included in the analyses of rate of improvement or discharge, but was included in the calculation of length of stay and time to readmission. b On Day 1, one participant in the control group withdrew without providing a reason, so was not able to be included in the analyses of rate of improvement or discharge, but was able to be included in the calculation of length of stay and time to readmission. c One participant in the experimental group received a lung transplant at Day 19 of the admission, so was able to be included in the rate of improvement analysis but was not able to be included in the discharge analysis or the calculation of length of stay and time to readmission. between the experimental and control groups in the change in analyses of daily rate of improvementID$]T_,7[F length of stay and time to respiratory muscle strength following chest physiotherapy and the next hospital admission. The experimental and control groups length of admission. Cox proportional-hazards regression was were similar at baseline (Table 1 and first two columns of Table 2). calculated to compare the time to next hospital admission in the experimental and control groups. Data were analysed on an Compliance with study protocol intention-to-treat basis and statistical significance was set at p = 0.05 for all calculations. All participants performed the active cycle of breathing technique. In addition, physiotherapists applied techniques of Results manual percussion, vibration and postural drainage positioning, autogenic drainage, PEP and oscillating PEP (Table 3). One of the Flow of participants, therapists and centres through the trial 19 participants in the experimental group did not tolerate NIV and withdrew from the study. The average inspiratory pressure was Over a 26-month period, 40 participants consented to the study 13 cmH2O (SD 2, range 10 to 19), average expiratory pressure was protocol and underwent randomisation (Figure 1). Three partici- 5 cmH2O (SD 1, range 4 to 7) and average pressure support pants withdrew from the study and the data from these (difference between inspiratory and expiratory pressure) was participants were used, as available and appropriate, in the 8 cmH2O (SD 3, range 6 to 15). The average daily NIV use was 65 minutes (SD 31, range 26 to 135).

Research 145 Table 1 Table 3 Characteristics of participants at admission to hospital. Physiotherapy techniques applied during the admission. Characteristic Exp Con Technique, n (%) Exp Con (n = 19) (n = 21) Manual techniques (n = 18) (n = 20) Age (yr), mean (SD) 28 (7) 30 (9) PEP Gender, n female (%) 6 (32) 8 (38) Oscillating PEP 18 (100) 20 (100) BMI (kg/m2), mean (SD) 21.4 (3.1) 20.4 (3.0) Autogenic Drainage 6 (33) 16 (80) FVC (% predicted), mean (SD) 61 (15) 63 (15) 1 (6) PImax (% predicted), mean (SD) 87 (32) 93 (30) 0 (0) 1 (5) PEmax (% predicted), mean (SD) 70 (22) 92 (41) 3 (15) BMI = body mass index, Con = control group, Exp = experimental group, FVC = forced All participants performed the active cycle of breathing technique. In addition, physiotherapists applied manual techniques of percussion, vibration and postural vital capacity, PEmax = maximal expiratory pressure, PImax = maximal inspiratory drainage positioning to all participants. Two participants in the control group capacity.10,12 performed PEP and autogenic drainage. One participant in the experimental group performed independent oscillating PEP without NIV for some physiotherapy Some sputum samples were unable to be collected before treatments. One participant in the experimental group performed independent PEP intravenous antibiotics commenced or no courier was available to without NIV for some physiotherapy treatments. Five participants in the transport samples to the centralised laboratory. Therefore, sputum experimental group performed PEP with NIV. samples were analysed for only 14/19 participants in the experimental group and 18/21 participants in the control group PI(Figure]_2[G)TD$ EP = positive expiratory pressure. at admission and discharge. Effect of the intervention Most of the study outcomes are presented in Table 2, with individual participant data presented in Table 4 (see eAddenda for Table 4). Lung function There was a non-significant trend for the experimental group to have a greater daily rate of improvement in FEV1 (% predicted) than the control group, with a mean difference of 0.13% per day (95% CI – 0.03 to 0.28), as shown in Figure 2. At discharge, the experimental group had a significantly higher FEV1 (% predicted) than the control group (Table 2). Subjective symptom severity and quality of life Figure 2. Daily rate of improvement in FEV1 (% predicted)10 from admission until There was no significant difference in the daily rate of Day 14, and discharge for the experimental group (closed circles) and the control group (open circles). improvement between the experimental and control groups for *p = 0.04. any domain (breathlessness MD 0.05 cm/day, 95% CI –0.07 to 0.18; expectorated sputum volume MD 0.00 cm/day, 95% CI –0.06 to reported significantly less fatigue at discharge than the control 0.06; energy MD –0.07 cm/day, 95% CI –0.20 to 0.06). At discharge, group using the Schwartz fatigue scale (Table 2). there was also no difference between the experimental and control groups in the participants’ rating of breathlessness, expectorated Respiratory muscle strength following chest physiotherapy sputum volume and energy levels (Table 2). On Day 2 of the admission, the PImax and PEmax worsened At discharge, there was no significant difference between the following standard chest physiotherapy (control group) and experimental and control groups in the physical, health and respiratory domains of the CFQ (Table 2). The experimental group _T]F[1ITD$ able 2 Mean (SD) of each group at admission and at discharge from hospital, and mean (95% CI) difference between groups at discharge. Outcome Groups Difference between groups Admission Discharge Discharge Exp (n = 19) Con (n = 21) Exp (n = 17) Con (n = 20) Exp minus Con FEV1 (% predicted) 36.1 (10.4) 39.1 (10.2) 49.5 (14.3) 48.2 (12.1) 4.2 (0.1 to 8.3) MST-25 distance (m) 737 (349) 776 (413) 914 (343) 929 (418) –13 (–86 to 61) P. aeruginosa density (log CFU/g) a 7.44 (2.23) 7.57 (0.78) 6.77 (1.68) 7.71 (0.88) –0.86 (–1.79 to 0.06) CFQ – physical CFQ – health 36 (19) 37 (24) 54 (27) 58 (22) –3 (–14 to 8) CFQ – respiratory 43 (23) 32 (19) 59 (24) 56 (24) –5 (–19 to 8) Fatigue 42 (16) 37 (15) 62 (17) 51 (15) 8 (–2 to 18) Breathlessness – VAS (cm) 43 (12) 44 (12) 37 (13) 43 (8) –6 (–11 to –1) Sputum volume – VAS (cm) 5.6 (2.2) 6.0 (2.4) 1.9 (1.3) 2.1 (2.0) –0.2 (–1.3 to 1.0) Energy – VAS (cm) 5.1 (2.1) 6.4 (1.8) 2.0 (1.4) 2.5 (2.2) –0.4 (–1.7 to 0.9) 6.4 (2.0) 5.7 (1.9) 3.2 (1.7) 3.2 (2.1) –0.1 (–1.5 to 1.2) Comparisons between the experimental and control groups are shown as estimated group mean differences, adjusted for admission values, with 95% CI. Fatigue measured on the Schwartz fatigue scale (minimum score = 7, maximum score = 63, lower values represent less fatigue). CFQ = CF-quality of life questionnaire, with physical, health and respiratory domains (minimum score = 0, maximum score = 100, higher values represent better quality of life), Con = control group, Exp = experimental group, FEV1 = forced expiratory volume in 1 second as percent of predicted normal lung volume, MST-25 = 25-level modified shuttle test, P. aeruginosa = Pseudomonas aeruginosa, calculated as the percentage of participants in each group colonised in sputum and density of P. aeruginosa colony forming units per gram of sputum, expressed as log values on a base 10 scale (log CFU/g), VAS = 10-cm visual analogue scale (higher values represent feeling more symptomatic). a Sputum samples were analysed for 14/19 participants in the experimental group and 18/21 participants in the control group at admission and discharge.

[(Fi]GI$DT)3_erug146 Dwyer et al: NIV-assisted chest physiotherapy in cystic fibrosis Length of hospital admission and time to next hospital admission There was no significant difference between the experimental and control groups in the length of hospital admission. The mean length of stay was 14.7 days (SD 3.9) in the 8[]F_TD$Icontrol group and 14.2 days (SD 4.0) in the DT_e$9[IF] xperimental group (MD –0.4 days, 95% CI –3.0 to 2.1). There was no significant difference between the experimental and control groups for the time to next hospital admission for an acute exacerbation of CF, as defined by the criteria of Fuchs et al,9 with a hazard ratio for re-admission in the experimental group of 0.78, p = 0.44 (Figure 4). Adverse events There were no adverse events in either group. Figure 3. Change in respiratory muscle strength following chest physiotherapy on Discussion Day 2 of admission. On the left is the change in maximal inspiratory pressure (PImax) from immediately before to immediately after chest physiotherapy for the Non-invasive ventilation-assisted chest physiotherapy applied experimental group (black column) and the control group (white column). On the during a hospital admission for an acute exacerbation of CF right is the change in maximal expiratory pressure (PEmax) from immediately resulted in significantly higher FEV1 (% predicted) at discharge than before to immediately after chest physiotherapy for the experimental group (black standard chest physiotherapy alone. However, the group who column) and the control group (white column). received NIV-assistance to the chest physiotherapy only showed a PEmax = maximal expiratory pressure, PImax = maximal inspiratory pressure. non-significant trend towards a greater daily rate of improvement *p = 0.006. in FEV1 (% predicted), which was the primary outcome of the study. At discharge, the improvement in lung function may have been improved following NIV-assisted chest physiotherapy (experi- related to the reduced sputum bacterial load and the preservation mental group). The mean difference in PImax was 8 cmH2O (95% CI of respiratory muscle strength with chest physiotherapy for those 2 to 14) and the mean difference in PEmax was 6 cmH2O (95% CI –3 using NIV during the early stages of the hospital admission. to 15), as shown in Figure 3. When PImax and PEmax were re- measured before and after chest physiotherapy 1 week into the Participants in the experimental group reported significantly admission and on discharge from hospital, there was no difference lower levels of fatigue on discharge from hospital compared to the in the change in respiratory muscle strength between the control group. This reduction in fatigue was also reported in other experimental and control groups. trials of people with CF when NIV-assistance was added to a single session of chest physiotherapy.5,7 Potentially, the lower levels of Exercise capacity fatigue are related to the preservation of respiratory muscle At discharge, there was no significant difference in distance strength following NIV-assisted chest physiotherapy, which has been demonstrated in previous studies5,6 as well as this study. covered in the 25-level modified shuttle test between the two Interestingly, the improvement in respiratory muscle strength groups (Table 2). following NIV-assisted chest physiotherapy in this study was only evident when measured on Day 2 of admission. When respiratory Quantitative sputum microbiology muscle strength was measured before and after chest physiother- Pseudomonas aeruginosa was identified in all sputum samples, apy 1 week into the admission and at discharge from hospital, there was no difference between the experimental and control except in one sample from a participant in the experimental group groups. Perhaps the unloading of the respiratory muscles with NIV at admission. On quantitative microbiology, the experimental is most effective when people with CF are the most unwell and group had a significantly lower Pseudomonas aeruginosa colony they have a higher work of breathing. count than the control group at 1 week after admission (mean difference –1.12 log CFU/g, 95% CI –0.04 to –2.20). At discharge, Despite having greater improvements in lung function and there was also a non-significant trend for the experimental group fatigue, the experimental group reported similar improvements in to have a lower Pseudomonas aeruginosa colony count than the symptom severity to the control group, both in the rate of improvement and discharge values. Perhaps the improvements in c4_er)giF([TD$IG]u ontrol group (Table 2). symptoms and quality of life from admission to discharge from hospital are so great that subjective scoring is unable to detect the Figure 4. Time to next hospital admission for an acute exacerbation of CF for the differences between types of chest physiotherapy. control group (dotted line) and experimental group (solid line). Surprisingly, considering the improvements in lung function and fatigue, the participants in the experimental group had no greater improvement in exercise capacity at discharge than those in the control group. Both groups had marked improvements in exercise capacity, from admission to discharge, with average improvements of 150 m on the 25-level modified shuttle test, where an improvement of 40 m is considered to be clinically significant.17 High sputum bacterial load is a major cause of morbidity in CF. In this study, the experimental group had greater reductions in the sputum Pseudomonas aeruginosa density than the control group, both at 1 week into the admission and at discharge from hospital. It is possible that NIV-assistance augmented the mucus clearance with chest physiotherapy, to help lower the sputum bacterial load, which in turn may have contributed to the greater improvements in lung function for the experimental group. Lung function is often the major determinant for a physician to decide when a patient should be admitted to hospital and when

Research 147 they are well enough to go home. In this study, although Footnotes: a VPAP III, ResMed, Bella Vista, Australia; b H2i, participants in the experimental group had slightly faster rates ResMed, Bella Vista, Australia; c ResLink, ResMed, Bella Vista, of improvement in lung function, there was no difference in length Australia; d BioMerieux Vitek, Durham, USA of hospital stay between the two groups. In addition, despite leaving hospital with significantly higher FEV1, there was no eAddenda: Table 4 can be found online at doi:10.1016/j.jphys. difference in the time to next hospital re-admission for an acute 2015.05.019. exacerbation between the two groups. Ethics approval: The Central Sydney Area Health Service (RPAH Although not formally measured in this study, the addition of Zone) (X03-0213) and The Prince Charles Hospital Health Service NIV to chest physiotherapy was well tolerated, which was similar District (EC 2389) Ethics Committees approved this study. All to previous single-session trials of NIV.4–8 The average daily NIV participants gave written informed consent before data collection use was 65 minutes. Only one of 19 participants in the began. experimental group refused NIV treatment due to poor tolerance, and there were no adverse events associated with NIV. The range in Competing interests: Dr Amanda Piper has been involved in: pressure support applied with NIV in this study (6 to 15 cmH2O) educational activities sponsored by the manufacturers of bilevel was similar to that of previous trials;4–7 however, the average devices (Phillips, Respironics, Australia; ResMed, Australia; Sentec, pressure support was only 8 cmH2O, which may have been Switzerland); industry-sponsored research (ResMed, Australia). insufficient to adequately support inspiratory effort and unload the No other author has a competing interest. respiratory muscles for those receiving lower inspiratory pressure. Perhaps higher levels of inspiratory pressure could have been Source of support: This study was funded by a project grant tolerated as participants became more accustomed to the from the National Health and Medical Research Council of treatment later in the admission, possibly allowing NIV treatment Australia (Project grant #245513). The NHMRC had no involve- to have been more effective, rather than maintaining the same NIV ment in the study design, analysis or presentation or publication. settings as determined in the acclimatisation session on the day of admission. Acknowledgements: The authors would like to thank Robyn Cobb and Kathleen Hall from the Adult Cystic Fibrosis Centre, The This investigation into the efficacy of NIV as an adjunct to chest Prince Charles Hospital, Brisbane, Australia for data collection at physiotherapy during an exacerbation of CF lung disease is a The Prince Charles Hospital; Dr Pholawat Tingpei from the difficult setting in which to identify between-group differences. Department of Infectious Diseases and Immunology, University This is because the standard care received by both groups of Sydney, Sydney, Australia for performing the quantitative (intravenous antibiotics and comprehensive input from the CF sputum microbiology assessments; and Dr Keith Wong from the team) is so effective, as reflected in the marked improvements in Department of Respiratory and Sleep Medicine, Royal Prince Alfred exercise tolerance and lung function for both groups in this study. Hospital, Sydney, Australia for assistance with statistical analyses. Perhaps if the study had targeted more severe CF patients or those with significantly increased work of breathing and intolerance to Provenance: Not invited. Peer-reviewed. standard chest physiotherapy, more differences would have been Correspondence: Tiffany Dwyer, Department of Respiratory found between the treatment groups. Another limitation of this and Sleep Medicine, Royal Prince Alfred Hospital, Sydney, study was that participants, clinicians and independent assessors Australia. Email: [email protected] were not blinded to treatment allocation. Complete blinding of treatment allocation would have been impossible; however, References introducing a third group with sham NIV or continuous positive airway pressure (CPAP) may have provided some blinding to 1. Dodd ME, Prasad SA. Physiotherapy management of cystic fibrosis. Chron Respir Dis. participants and independent assessors. 2005;2:139–149. In summary, the use of NIV as an adjunct to chest physiotherapy 2. Alison JA, Donnelly PM, Lennon M, Parker S, Torzillo P, Mellis C, et al. The effect of a during a hospital admission for an acute exacerbation of CF comprehensive, intensive inpatient treatment program on lung function and significantly improved lung function and fatigue on discharge from exercise capacity in patients with cystic fibrosis. Phys Ther. 1994;74:583–591. hospital. This improvement in lung function, however, was not matched by improvements in symptom severity, exercise capacity, 3. Moran F, Bradley J, Piper A. Non-invasive ventilation for cystic fibrosis. Cochrane length of hospital stay, or the exacerbation-free period after Database Syst Rev. 2013;4:CD002769. discharge from hospital. It is very difficult to show improvements over comprehensive inpatient care; however, the results of this 4. Elkins MR, Eberl S, Alison J, Bye P. The effect of bi-level non-invasive ventilation on study suggest that the addition of NIV to standard chest mucociliary clearance in subjects with cystic fibrosis. Pediatr Pulmonol. physiotherapy is a useful tool to aid airway clearance, especially 2004;38(S27):315. when people with moderate to severe CF lung disease are most at risk of respiratory muscle fatigue. Future trials targeting people 5. Fauroux B, Boule M, Lofaso F, Zerah F, Clement A, Harf A. Chest physiotherapy in with more-severely affected lungs or continuing the addition of cystic fibrosis: improved tolerance with nasal pressure support ventilation. Pedi- NIV to chest physiotherapy after discharge from hospital would atrics. 1999;103:E32. further evaluate the role of NIV in CF care. 6. Holland AE, Denehy L, Ntoumenopoulos G, Naughton MT, Wilson JW. Non-invasive What is already known on this topic: People with cystic ventilation assists chest physiotherapy in adults with acute exacerbations of cystic fibrosis often find it hard to maintain effective airway clearance fibrosis. Thorax. 2003;58:880–884. during acute exacerbations of their lung disease due to in- creased breathlessness and respiratory muscle fatigue. When 7. Placidi G, Cornacchia M, Polese G, Zanolla L, Assael BM, Braggion C. Chest physio- used as an adjunct to a single session of airway clearance therapy with positive airway pressure: a pilot study of short-term effects on techniques, non-invasive 0_][1FvT$ID entilation reduces these symptoms sputum clearance in patients with cystic fibrosis and severe airway obstruction. and improves ease of expectoration. Respir Care. 2006;51:1145–1153. What this study adds: When non-invasive ventilation was used as an adjunct to chest physiotherapy throughout a 8. Kofler AM, Carlesi A, Cutrera R, Leone P, Lucidi V, Rosati S, et al. BiPAP versus PEP as chest hospital admission by adults with an acute exacerbation of physiotherapy in patients with cystic fibrosis. Pediatr Pulmonol. 1998;26(S17):344. cystic fibrosis lung disease, they left hospital with better lung function and less fatigue. 9. Fuchs HJ, Borowitz DS, Christiansen DH, Morris EM, Nash ML, Ramsey BW, et al. Effect of aerosolized recombinant human DNase on exacerbations of respiratory symptoms and on pulmonary function in patients with cystic fibrosis. New Engl J Med. 1994;331:637–642. 10. Quanjer PH, Tammeling GJ, Cotes JE, Pedersen OF, Peslin R, Yernault JC. Lung volumes and forced ventilatory flows. Report Working Party Standardization of Lung Function Tests, European Community for Steel and Coal. Official Statement of the European Respiratory Society. Eur Respir J. 1993;16(Supplement):5–40. 11. Pryor J. The forced expiration technique. In: Pryor J, ed. In: Respiratory Care. Edinburgh: Churchill Livingstone; 1991:79–100. 12. Wilson SH, Cooke NT, Edwards RH, Spiro SG. Predicted normal values for maximal respiratory pressures in caucasian adults and children. Thorax. 1984;39:535–538. 13. Quittner AL, Sweeny S, Watrous M, Munzenberger P, Bearss K, Gibson Nitza A, et al. Translation and linguistic validation of a disease-specific quality of life measure for cystic fibrosis. J Pediatr Psychol. 2000;25:403–414. 14. Schwartz JE, Jandorf L, Krupp LB. The measurement of fatigue: a new instrument. J Psychosom Res. 1993;37:753–762. 15. Elkins MR, Dentice RL, Bye PT. Reliability and repeatability of the MST-25: An extension of the Modified Shuttle Test (MST). Pediatr Pulmonol. 2009;44(S32):445. 16. Wong K, Roberts MC, Owens L, Fife M, Smith AL. Selective media for the quantita- tion of bacteria in cystic fibrosis sputum. J Med Microbiol. 1984;17:113–119. 17. Bradley J, Howard J, Wallace E, Elborn S. Reliability, repeatability, and sensitivity of the modified shuttle test in adult cystic fibrosis. Chest. 2000;117:1666–1671.

Editorial 105 4. De Angelis C, et al. N Engl J Med. 2004;351:1250–1251. 18. Hoffmann T, et al. BMJ. 2014;348:g1687. 5. le Fort SM, et al. Pain. 1998;74:297–306. 19. Loudon K, et al. Trials. 2013;14:115. 6. Snowdon D, et al. J Physiother. 2014;60:66–77. 20. Hollis S, Campbell F. BMJ. 1999;319:670. 7. Butler GS, et al. Pat Educ Counseling. 1996;28:189–197. 21. Phillips AC, et al. BMC Med Educ. 2013;13:9. 8. Dowsey MM, et al. Med J Aust. 1999;170:59–62. 22. Moher D, et al. BMJ. 2010;340:c869. 9. Condessa RL, et al. J Physiother. 2013;59:101–107. 10. Blackmer J, Haddad H. CMAJ. 2005;173:1052–1053. Websites 11. Schulz KF, et al. BMC Medicine. 2010;8:19. 12. Chan L, et al. Arch Phys Med Rehabil. 2014;95:415–417. www.equator-network.org 13. Moseley AM, et al. J Clin Epidemiol. 2011;64:594–601. www.immpact.org 14. Brueton VC, et al. Cochrane Datab Syst Rev. 2013;12:MR000032. www.omeract.org 15. Chan A-W, et al. BMJ. 2013;346:e7586. 16. Boutron I, et al. Ann Intern Med. 2008;148:295–309. http://dx.doi.org/10.1016/j.jphys.2015.05.012 17. Glasziou P, et al. Lancet. 2014;383:267–276. People’s Choice Award for 2014 The Editorial Board is pleased to introduce the annual People’s Choice Award, which recognises the paper published in Journal of Physiotherapy that generates the most interest by readers of the journal. The winning paper is chosen based on the number of times that each paper published in a given year is downloaded in the six months after its day of publication. The winning paper is ‘Current evidence does not support the use of Kinesio Taping in clinical practice: a systematic review’ by Patrı´cia do Carmo Silva Parreira and colleagues from the Universidade Cidade de Sa˜o Paulo in Brazil.1 This systematic review examined the 12 randomised trials that provided published data by June 2013 about the effect of Kinesio Taping on pain, disability, quality of life, return to work and global perceived recovery in people with musculoskeletal conditions. These trials provided data on 495 participants. Studies were excluded if they were conducted on healthy participants or only reported data on physical performance (eg, vertical jump test). The 12 trials covered a range of musculoskeletal conditions. Among these trials, Kinesio Taping had no benefit over sham taping or other active treatments to which it had been compared, the benefit was too small to be clinically worthwhile, or the trials were of low quality. Therefore the evidence did not support the use of Kinesio Taping for musculoskeletal conditions. The Journal of Physiotherapy has subsequently received and published two more high-quality randomised trials comparing Kinesio Taping to sham taping: one for swelling after ankle sprain by Nunes and colleagues2 and one for low back pain, which again has Patrı´cia do Carmo Silva Parreira as the first author.3 These two trials further reinforce the findings of the review by demonstrating that Kinesio Taping was no better than sham taping for these conditions. The winning paper also generated the highest activity on social media among the papers published in 2014. The Editorial Board of Journal of Physiotherapy congratulates Patrı´cia do Carmo Silva Parreira and colleagues on their success. References 1. Parreira PdCS. Costa LdCM. Hespanhol Junior LC, Lopes AD, Costa LOP. Current evidence does not support the use of Kinesio Taping in clinical practice: a systematic review. J Physiother. 2014;60:31–39. 2. Nunes GS, Vargas VZ, Wageck B, dos Santos Hauphental DP, da Luz CM, de Noronha M. Kinesio Taping does not decrease swelling in acute, lateral ankle sprain of athletes: a randomised trial. J Physiother. 2015;61:28–33. 3. Parreira PCS, Costa LCM, Takahashi R, Hespanhol Junior LC, da Luz Junior MA, da Silva TM, Costa LOP. Kinesio Taping to generate skin convolutions is not better than sham taping for people with chronic non-specific low back pain: a randomised trial. J Physiother. 2014;60:90–96. http://dx.doi.org/10.1016/j.jphys.2015.05.010

Journal of Physiotherapy 61 (2015) 161 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers Physiotherapist-led neck-specific exercise improves pain, disability and self-efficacy in chronic whiplash-associated disorders Synopsis Summary of: Ludvigsson ML, Peterson G, O’Leary S, Dedering A, on visual analogue scales from 0 (none) to 100 (extreme). Confidence to Peolsson A. The effect of neck-specific exercise with, or without a perform activities was rated on the Self-Efficacy Scale. Analgesic behavioral approach, on pain, disability, and self-efficacy in chronic medication use and compliance with the allocated interventions were whiplash-associated disorders: a randomized clinical trial. Clin J Pain assessed via diaries. Results: A total of 156 participants completed the [4_DIF]2$T 015;31:294-303. study, with 50 to 71% compliance among the groups. Compared to physical activity prescription, neck exercises significantly improved Question: Does neck-specific exercise, with or without a beha- the Neck Disability Index at intervention end, with slightly larger vioural approach, improve disability, pain and self-efficacy in people effects at follow-up (MD 2.3, 95% CI 0.1 to 4.5 without behavioural with chronic whiplash-associated disorders? Design: Three-arm, intervention; MD 3.5, 95% CI 1.3 to 5.7 with behavioural intervention). randomised trial with concealed allocation and intention-to-treat At follow-up, >FI3[D_5$T] points of improvement in the Neck Disability Index analysis. Setting: Primary-care centres in Sweden. Participants: People was significantly more common with neck exercises and behavioural with whiplash-associated disorders Grade 2 or 3, Neck Disability Index intervention than with activity prescription (RR 2.1, 95% CI 1.2 to 3.9). score  10, and average pain on a Visual Analogue Scale  20/100. At intervention end, pain bothersomeness was significantly lower in Randomisation allocated 76 to neck-specific exercises, 71 to neck- the two neck-exercise groups than the activity prescription group, with specific exercises with behavioural intervention and 69 to a prescrip- significant differences in the likelihood ofFT]ID$_5[ >50% reduction in pain tion of physical activity. Interventions: The neck-specific exercise bothersomeness (RR 9.4, 95% CI 3.0 to 29.1 for neck exercises; RR 5.8, 95% group received: education (anatomy, posture); active neck exercises CI 1.8 to 18.7 with behavioural intervention). The groups did not differ with increasing resistance and repetitions; and prescription of general significantly for the remaining secondary outcomes. Conclusion: physical activity. The behavioural group received the same regimen Compared with physical activity prescription, neck-specific exercises plus: advice (ignore temporary pain intensity, progress exercises); pain reduced disability and pain from whiplash-associated disorders. Adding education; problem solving; and pain-management skills (eg, relaxa- behavioural intervention had unclear benefit. tion). After a short motivational interview and physical examination, the remaining group received a written, individualised prescription for [95% CIs calculated by the Editor.] home or gym activities without any neck-specific exercises, with one follow-up contact. All participants were encouraged to continue Provenance: Invited. Not peer-reviewed. exercising after these 12-week interventions. Outcome measures: Outcomes were measured at intervention end and at follow-up M]ID$[FT_1 ark Elkins 12 weeks later. The primary outcome was the Neck Disability Index, Editor, Journal of Physiotherapy scored from 0 (normal) to 50 (maximal activity limitation). Pain bothersomeness in the prior 24 hours and current pain were each rated http://dx.doi.org/10.1016/j.jphys.2015.05.007 Commentary Many studies have been undertaken to explore the management The uncertain additional benefit of exercise over the educational strategies for whiplash-associated disorders.1,2 As most recovery of approach indicates a need to develop more effective management structural injury occurs within the first 2 to 3 months of injury, it has algorithms for chronic whiplash-associated disorders by including been considered that management strategies for chronic whiplash- understanding of the mechanisms responsible for prolonged symptoms. associated disorders are different to those for acute whiplash-associated We also need to keep exploring the optimal type and dosage of exercises disorders. It is important to incorporate biopsychosocial perspectives and education with the use of medication and psychological approaches. into the management of chronic whiplash-associated disorders.3 For future studies of the effectiveness of physiotherapists’ contributions to whiplash-associated disorders, it may_[D1]F$IT alsoIDF]T_[$5 be important to consider Ludvigsson et al demonstrated a statistically greater reduction of using measures of outcomes other than pain intensity, disability or cost[6TD$IF_]- Neck Disability Index scores and pain bothersomeness in a group d3[_T]DF$I oing effectivenessI_[D2]TF$ such as a 50% reduction in being bothered by pain, or a neck-specific exercises provided by physiotherapists in comparison to broader concept such as long-term happiness]7_TD.F$I[ a group n[4_TD$IF] ot doing the specific exercises (a simple educational approach). While average between-group differences were less than Provenance: Invited. Not peer-reviewed. what would be considered clinically important, individuals in the exercise groups were much more likely to make clinically important Hiroshi Takasaki improvements in these outcomes than individuals receiving education. Division of Physical Therapy, Saitama Prefectural University, Therefore, the clinical significance of the results remains uncertain. A larger randomised trial by Michaleff et al1 found no additional benefit Koshigaya, Japan of treatment algorithms with specific exercises over the educational approach. They also undertook a meta-analysis regarding additional References benefit of specific or non-specific exercises over the educational approach and showed a reduction in short-term disability of –3.3 (95% 1. Michaleff ZA, et al. The Lancet. 2014;384:133–141. CI –5.5 to –1.1) on a 0-to-100 scale,1 which is considered to be a 2. Lamb SE, et al. The Lancet. 2013;381:546–556. negligible difference. Thus, patients could be advised that a treatment 3. TRACsa. Trauma and Injury Recovery, South Australia, Adelaide 2008. algorithm with exercises is more effective than a simple educational approach, but the difference in effects may be slight. http://dx.doi.org/10.1016/j.jphys.2015.05.006 1836-9553/ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.

Journal of Physiotherapy 61 (2015) 161 Journal of PHYSIOTHERAPY journal homepage: www.elsevier.com/locate/jphys Appraisal Critically Appraised Papers Physiotherapist-led neck-specific exercise improves pain, disability and self-efficacy in chronic whiplash-associated disorders Synopsis Summary of: Ludvigsson ML, Peterson G, O’Leary S, Dedering A, on visual analogue scales from 0 (none) to 100 (extreme). Confidence to Peolsson A. The effect of neck-specific exercise with, or without a perform activities was rated on the Self-Efficacy Scale. Analgesic behavioral approach, on pain, disability, and self-efficacy in chronic medication use and compliance with the allocated interventions were whiplash-associated disorders: a randomized clinical trial. Clin J Pain assessed via diaries. Results: A total of 156 participants completed the [4_DIF]2$T 015;31:294-303. study, with 50 to 71% compliance among the groups. Compared to physical activity prescription, neck exercises significantly improved Question: Does neck-specific exercise, with or without a beha- the Neck Disability Index at intervention end, with slightly larger vioural approach, improve disability, pain and self-efficacy in people effects at follow-up (MD 2.3, 95% CI 0.1 to 4.5 without behavioural with chronic whiplash-associated disorders? Design: Three-arm, intervention; MD 3.5, 95% CI 1.3 to 5.7 with behavioural intervention). randomised trial with concealed allocation and intention-to-treat At follow-up, >FI3[D_5$T] points of improvement in the Neck Disability Index analysis. Setting: Primary-care centres in Sweden. Participants: People was significantly more common with neck exercises and behavioural with whiplash-associated disorders Grade 2 or 3, Neck Disability Index intervention than with activity prescription (RR 2.1, 95% CI 1.2 to 3.9). score  10, and average pain on a Visual Analogue Scale  20/100. At intervention end, pain bothersomeness was significantly lower in Randomisation allocated 76 to neck-specific exercises, 71 to neck- the two neck-exercise groups than the activity prescription group, with specific exercises with behavioural intervention and 69 to a prescrip- significant differences in the likelihood ofFT]ID$_5[ >50% reduction in pain tion of physical activity. Interventions: The neck-specific exercise bothersomeness (RR 9.4, 95% CI 3.0 to 29.1 for neck exercises; RR 5.8, 95% group received: education (anatomy, posture); active neck exercises CI 1.8 to 18.7 with behavioural intervention). The groups did not differ with increasing resistance and repetitions; and prescription of general significantly for the remaining secondary outcomes. Conclusion: physical activity. The behavioural group received the same regimen Compared with physical activity prescription, neck-specific exercises plus: advice (ignore temporary pain intensity, progress exercises); pain reduced disability and pain from whiplash-associated disorders. Adding education; problem solving; and pain-management skills (eg, relaxa- behavioural intervention had unclear benefit. tion). After a short motivational interview and physical examination, the remaining group received a written, individualised prescription for [95% CIs calculated by the Editor.] home or gym activities without any neck-specific exercises, with one follow-up contact. All participants were encouraged to continue Provenance: Invited. Not peer-reviewed. exercising after these 12-week interventions. Outcome measures: Outcomes were measured at intervention end and at follow-up M]ID$[FT_1 ark Elkins 12 weeks later. The primary outcome was the Neck Disability Index, Editor, Journal of Physiotherapy scored from 0 (normal) to 50 (maximal activity limitation). Pain bothersomeness in the prior 24 hours and current pain were each rated http://dx.doi.org/10.1016/j.jphys.2015.05.007 Commentary Many studies have been undertaken to explore the management The uncertain additional benefit of exercise over the educational strategies for whiplash-associated disorders.1,2 As most recovery of approach indicates a need to develop more effective management structural injury occurs within the first 2 to 3 months of injury, it has algorithms for chronic whiplash-associated disorders by including been considered that management strategies for chronic whiplash- understanding of the mechanisms responsible for prolonged symptoms. associated disorders are different to those for acute whiplash-associated We also need to keep exploring the optimal type and dosage of exercises disorders. It is important to incorporate biopsychosocial perspectives and education with the use of medication and psychological approaches. into the management of chronic whiplash-associated disorders.3 For future studies of the effectiveness of physiotherapists’ contributions to whiplash-associated disorders, it may_[D1]F$IT alsoIDF]T_[$5 be important to consider Ludvigsson et al demonstrated a statistically greater reduction of using measures of outcomes other than pain intensity, disability or cost[6TD$IF_]- Neck Disability Index scores and pain bothersomeness in a group d3[_T]DF$I oing effectivenessI_[D2]TF$ such as a 50% reduction in being bothered by pain, or a neck-specific exercises provided by physiotherapists in comparison to broader concept such as long-term happiness]7_TD.F$I[ a group n[4_TD$IF] ot doing the specific exercises (a simple educational approach). While average between-group differences were less than Provenance: Invited. Not peer-reviewed. what would be considered clinically important, individuals in the exercise groups were much more likely to make clinically important Hiroshi Takasaki improvements in these outcomes than individuals receiving education. Division of Physical Therapy, Saitama Prefectural University, Therefore, the clinical significance of the results remains uncertain. A larger randomised trial by Michaleff et al1 found no additional benefit Koshigaya, Japan of treatment algorithms with specific exercises over the educational approach. They also undertook a meta-analysis regarding additional References benefit of specific or non-specific exercises over the educational approach and showed a reduction in short-term disability of –3.3 (95% 1. Michaleff ZA, et al. The Lancet. 2014;384:133–141. CI –5.5 to –1.1) on a 0-to-100 scale,1 which is considered to be a 2. Lamb SE, et al. The Lancet. 2013;381:546–556. negligible difference. Thus, patients could be advised that a treatment 3. TRACsa. Trauma and Injury Recovery, South Australia, Adelaide 2008. algorithm with exercises is more effective than a simple educational approach, but the difference in effects may be slight. http://dx.doi.org/10.1016/j.jphys.2015.05.006 1836-9553/ß 2015 Australian Physiotherapy Association. Published by Elsevier B.V. All rights reserved.